"The Real Eisman Playbook" podcast · ex-Neuberger Berman portfolio manager · of "The Big Short" — running synthesis of his video/podcast appearances, with per-transcript breakdowns and a stock index.
Fuel cells that turn natural gas into on-site electricity for data centers — "a major beneficiary of the AI boom," +90% YTD. EPS +680%, revenue past $1B for the first time (+166%). Valuation decays fast: 73× 2026 but 37× and 23× for 2027/28 — "if the AI story keeps going, I would expect Bloom stock to continue to perform. But again, the AI story has to keep going."
"Ben's been a great chart" (Verrone, Ep 73), said immediately after listing State Street, T. Rowe and Invesco as the asset-management leadership. Read with care — the auto-caption gives only "Ben"; inside a list of traditional asset managers it reads as Franklin Resources (NYSE: BEN), but the company name is never spoken.
Named with UnitedHealth ("the Signas") among the managed-care names that "put major lows in late last year early this year" — the bottom Verrone admits he missed by being too dogmatic (Ep 73). Bottomed, not leadership; and the PBM middleman role is squarely in the populist-pricing crosshairs.
Dan Ives' top cybersecurity pick — budgets "are going to double the next two or three years" as every AI agent multiplies the attack surface ("if Steve Eisman has three agents… they have to protect three agents"). Best-positioned on product and CEO with Palo Alto; the March fear that Anthropic would eat the sector had the direction backwards.
The credit-market read on the AI cycle — raising $2.6B "priced with a yield of more than, get this, 9%. That is expensive debt." Eisman's point is the spread, not the company: "fixed income investors are discriminating between the large companies like Google that they know can pay the money back and smaller newer companies like CoreWeave that are more risky."
One of Verrone's three "left for dead" life-science reversals (Ep 73) that "haven't been looked at in years" and, on the charts, "put in major major bottoms" — part of the health-care complex he ranks at the top of the defensive sectors.
Named beside Quanta by Verrone (Ep 73) as an "AI-adjacent or build-out type name" down ~30% from its highs that has since responded — the mechanical/electrical construction contractor that installs data-center power and cooling. The rule attached: keep giving the group the benefit of the doubt "until they attempt to rally, don't make new highs, and fail." (Auto-caption garble "Emcore"; the peer set is build-out contractors, i.e. EMCOR Group.)
The second engine of Europe's index high (Verrone, Ep 73) — "names like Siemens Energy turbines." Same structure as Eisman's GE Vernova thesis: when electricity demand outruns supply, own the company selling the generating equipment, not the one selling the power (which this same episode rates negatively).
"One of the better AI power stories" — sells bundled turbines + electrification to the AI build-out. Catalyst: Chevron's 20-yr PPA with Microsoft for a West Texas data center, mostly powered by GEV turbines. The "own the supplier" power play.
Still owned — by him and by Porter Collins (Ep 75, Sep 14): the low-cost California greenhouse grower (~$90–100/lb) whose prize is export — "$200 a pound" in California vs "600, 700 bucks a pound" in Germany, i.e. "50 to 65% EBITDA margins" if interstate/overseas sales open; "a 2027 event," first hemp sold overseas this quarter. Earlier: cannabis name he owns; Trump's medical-cannabis reclassification (Schedule 1→3) "will improve growth and earnings." Premium management interview June 24. Educational, not a recommendation.
Gold, the guests' core holding (Ep 75, Sep 14) — Porter Collins: "that's where the majority of our capital sits." Interest plus entitlements swamp the tax take and nobody will cut or tax, so printing comes in a crisis and in the status quo; central banks swap Treasuries ("someone else's liability") for an asset with no debt. Eisman dissents: the deficit is "not… eminent doom" yet, and he doesn't own gold (Sep 11).
Guests Daniel & Collins "own a lot of" it (Ep 75, Sep 14): four floating-liquefaction ships on ~20-year toll-road contracts, ~$2.5–3.5B capex vs a $30–50B onshore plant; two head to Argentina to "pull the gas out starting next year" — the listed expression of their bullish-Argentina (Milei) thesis.
The guests own it (Ep 75, Sep 14) — Daniel & Collins: "we've owned Google for years," bought "when Google was allegedly being disrupted"; also their evidence of the capex regime change ("Google issues $85 billion in equity"). Eisman himself sold (disclosed Aug 15) — "I sold my Google. I wanted to reduce my exposure to AI. What scares me is that it's all one trade. So it better succeed." Not a verdict on the business: in the same clips Google is one of the hyperscalers with "multiple revenue streams from established businesses which are very unlikely to simply disappear" and a moat built out of sheer capex. An exposure decision, not a downgrade — the layer he defends is the layer he exited.
Named beside Kraft Heinz as the staples "all starting to bottom here… worth a look" (Verrone, Ep 73) — a bottoming call inside the group he ranks last among defensives, not a leadership call.
The deepest-de-rated of Verrone's three life-science bottoms (Ep 73) — a sequencing capital-equipment business a squeezed research budget can simply defer, which is why the reversal is "particularly notable" when funding conditions stop deteriorating.
Third name in Verrone's asset-manager breakout list (Ep 73), with State Street and T. Rowe — "really the leadership in the asset management space." A fee-based manager re-rating as the market turns earnings-driven and dispersion returns.
Two roles in Ep 72 — Schorr's exemplar of the relationship that survives cash optimization ("you don't just go to a great bank like JP Morgan just to park your cash": branches, safety, cybersecurity, investments, research), and the big-bank quarter Eisman calls "actually stunning" (IB +38%, trading +47%, wealth +7% organic, huge ROEs — "almost as good as it gets"). Caveats: spread lending only "okay" with deposit competition lifting cost of funds, and trading historically −17% in the second half. Eisman's tell: "when everything is humming… it's always possibility that tomorrow that's not going to hum."
Named again as an AI diversifier (Sep 18) — "the property and casualty subsector is uncorrelated to AI and to the economy"; one of three ETFs he suggests with LVHD and SPLV. Earlier:The one concrete instrument he names for someone trying to de-risk out of concentrated AI gains — "I could tell you certain ETFs that you could buy that would be lower risk. Like you could buy the KBWP, which is the property and casualty insurance index ETF. Just as one example" (Aug 28). Framed as an illustration, not a recommendation; the point is an earnings driver unrelated to the one trade, offered against the tax lock that keeps most holders from selling at all ("35% plus in capital gains taxes. That is painful").
The one flicker in the worst defensive group (Verrone, Ep 73): after "a just devastatingly bad couple years," "Kraft Heinz, Hershey's all starting to bottom here. I think they're worth a look" — immediately sized down with "I would hardly call the sector leader in the market." Staples stay near the bottom of his defensive ranking.
The monetization exception in a stalled private-equity industry — "record monetizations" the day of taping (Jul 30) while peers "can't sell some of the stuff they own" despite a better IPO market, better M&A and record equity highs. Schorr's dispersion point: private equity has underperformed public markets "for the first time in a very long time, probably since '08 or '09," so an inability to sell into this backdrop is a valuation statement about the 2020-21 vintage, not a timing preference.
The geopolitics trade in defense — 2Q26 sales +11%, EPS 794 vs 720 est, and a record $230B backlog up 24% in three months. "Clearly, the geopolitical situation is benefiting defense companies."
Guest George Noble's first Substack pick (January; Ep 76, Sep 21) and his only miss so far ("we're 11 for 12"): cheap, barely profitable, Elliott involved, bag fees and "getting rid of the middle seat" — "if they earned half of what United or Delta made, it would be big stock." Hormuz and oil "derailed" it, "down a few %. No big deal." Noble's view, not Eisman's.
Franklin U.S. Low Volatility High Dividend Index ETF
One of three ETFs he suggests to diversify away from AI without leaving equities (Sep 18): low-volatility, high-dividend stocks "would provide subdiversification" in a market where "so much of it is correlated to AI."
A neglected turnaround "beginning to work" (Sep 11) — a ~$10B cap "no one cares about" beat and raised; total comps +2.7% but Bloomingdale's +11.3% on "the demise of Saks." Stock fell on next-quarter EPS guidance, "but still, the strength in same-store sales comps was a pretty good sign." Constructive read, no position.
Owned for years — "a duopoly with pricing power" (with S&P) he doesn't believe AI can break. EPS 468 +31%; flat since fall 2024 so the 2026 PE is sub-30×, "the cheapest the stock has been in quite some time."
Bellwether investment bank — "how investors feel about the economy." Rallied as recession fears faded, up 34% in Q2, the biggest of the three big banks.
His recommended homebuilder — pitched low-$70s in Jan, dipped to $62 on war/rates, now $74 at 1.0× tangible book vs M&A at 1.25–1.3× TBV; rallying on Berkshire M&A + the 10-year below 4.5%.
Bought "on the cheap" in the SaaSpocalypse by guests Daniel & Collins (Ep 75, Sep 14) — one of their rare mainstream longs, taken only because the sector was being dumped. Eisman's earlier use: a software-derating proxy — $209 (Jul 2025) to <$90 (−60%), implying private-equity software equity is worth <½ cost; the core of his ARCC/Blue Owl refinancing worry. Illustrative.
The other half of Ives' cybersecurity pair, named with CrowdStrike as best-positioned "from a product perspective and CEOs, what they've done." A platform consolidator in a market where CIOs are cutting from a hundred vendors to thirty.
Gil Luria's top-three winner alongside Microsoft and Micron — "really good software companies like Microsoft, especially Palantir." Sells the insulation against model dependency: own the data and ontology, keep the model swappable, because a provider that sees your data "knows how your business operates" and a model that disappears takes your business with it.
Turnaround that finally turned — 2nd straight quarter of traffic growth, EPS 0.50 vs 0.42 est, SSS +6.2%, raised outlook; stock +8%. "Investors are finally rewarding the company's business changes."
AI-derivative server name raising $7B (equity + equity-linked) off a ~$20B market cap — "no small thing"; stock down 28%. More capital-intensity evidence.
One of three ETFs he suggests to diversify away from AI without leaving equities (Sep 18): the S&P 500's least volatile stocks, alongside LVHD and KBWP.
Re-rated by Ep 73 (Aug 17): Verrone's lead example of the traditional asset managers "starting to break out here for the first time in years" — "everything from State Street to T. Rowe to Invesco… really the leadership in the asset management space" — the tape's confirmation of Strategas' "alpha market" (2026 "has been about the E, not the P/E"). Previously (Jul 24) Eisman grouped it with Bank of New York as a trust/custody bank in "a world all their own," outside his research focus.
One of only two names Verrone credits for the Euro Stoxx's new high (Ep 73): "It's been stocks like Schneider Electric — AI buildout," while European autos and luxury sit in multi-year bear markets on Chinese demand. The uncomfortable read-through: buying Europe to escape the AI trade mostly buys the AI trade in euros.
"Own the supplier, not the operator" — airline suppliers like TransDigm are great businesses vs the awful airlines; his AI analogy is hyperscalers→airlines, suppliers→TransDigm (compare AAL vs TDG 10-yr charts).
Verrone's top defensive call (Ep 73): pharma, biotech and life sciences "have been leadership. They remain the leadership. They act great," and the reversals in the "left for dead" life-science names — "the thermos and the Danaher's and the Illumina's, which haven't been looked at in years" — are "major major bottoms." Health care ranks first among defensives, staples last.
One of the traditional asset managers Verrone says are "starting to break out here for the first time in years" (Ep 73) — the tape's confirmation of Strategas' "alpha market": 2026 has been "about the E, not the P/E," so stock picking and active "matters a lot more than passive" after fifteen years on the wrong side of the passive trade.
The miss Verrone owns (Ep 73): managed care "put major lows in late last year early this year… I got too dogmatic on those and I missed it. We try not to do that." Constructive on the group now, but he says the actual leadership is pharma/biotech. Trennert's overhang: populist politics "against big anything," and government pricing that "worked for them for many years" now works the other way.
Subscriber view (Sep 11), passing: SK Hynix named with Micron as the memory suppliers in the "sell the knives" trade that benefit whoever wins the AI build-out. Not adopted by Eisman.
Sitting out the LLM race "is starting to look like a great decision" — won't be raising capital; new Siri "much better, but the bar was pretty low"; teaming with Google; features gate the iPhone upgrade cycle.
The price benchmark in the Carvana short (Ep 75, Sep 14): one of Carvana's biggest loan buyers, paying "between 102 and 104" — so an average gain on sale of 109–110 implies someone else pays "north of 110." No view on Ally.
Subscriber view (Sep 11), passing: the fourth equipment name in the "sell the knives" list of suppliers that win regardless of AI architecture. Not adopted by Eisman.
The middle term in Luria's chip-multiple comparison — "Micron six times earnings. AMD 50 times earnings. Intel 100 times earnings" — on the CPU side of a market he argues is now worse than memory. Priced closer to the cycle-continues camp than the cycle-is-over camp.
One of the four hyperscalers he says still have something left if the model layer cracks: "the hyperscalers like Google, Amazon, Microsoft, and Oracle have real businesses here. What the returns will look like, I don't know yet, but they have real businesses" — the moat being the sheer cost of entry. The offsetting exposure is the group's: ~70% of hyperscaler AI revenue comes from Anthropic and OpenAI, ≈25–35% of cloud revenue.
Rushing the S-1 before token maxing shows (Ep 76, Sep 21) — "they need to get their S1 out fast because token maxing is ending," to list on peak second-quarter numbers; "in better shape than open AI" regardless. An IPO "maybe in a month or two" (Sep 18) — Amodei's call to slow down "can't" be real: "What would it say to investors on the road show?" The safety alarm is a shell game for regulatory moats ahead of "a price war," and "it's unclear to me how to keep a shell game going in an IPO process that requires transparency." Earlier:The stronger of the two labs — June-quarter revenue ~$11.5B, "up over 100% versus the March quarter" — but the print is discounted before it lands: "token maxing… ended" around late June/July, so "the third and fourth quarters are going to be much more interesting… than the first half." The scheduled verdict is the S-1 ("I'm hoping to get some clarity on Anthropic when they put out their S-1. Whether they'll give it or not, I don't know"); prediction markets price it 93% to list first. Structural flaw unchanged: "they don't have the breadth of revenue streams of Google and Microsoft."
Referenced only as the example in guest George Noble's yen carry-trade explainer (Ep 76, Sep 21): yen into dollars into "Avalon Bay" for yield, plus the dollar's gain — "What could go wrong?" No view.
Soft AI print (strong but below expectations) sent the semis into a Thursday correction — one of the two triggers (with the 10-year) for Friday's selloff.
Named in the "banks are just the AI trade" peer set — at peak valuations with an investment-banking cycle "heavily dependent on AI financing needs," so it offers no diversification from tech.
A trust bank that "operates in a world all their own" and he has "never found that much value to focusing on" — up 30%+ this year, "but you could get the same performance by buying just about any other of the large cap banks."
The institutional leg of Worthington's crypto "big three" — "a more institutional-focused firm called Bullish," i.e. a brokerage for institutions that want to trade crypto, versus Coinbase's retail-plus-everything vertical integration. Context in the first crypto-broker segment the show has ever run; no rating given on air.
Good-but-mixed print (EPS $1.52 vs $1.36 est, $68B raised) against two unresolved sector problems — exit timelines lengthening and private credit's software loans facing next year's refinancing cycle; now also "the biggest financier of AI data centers," which "may or may not prove to be a great bet."
"A successful banking story" — premium deep dive into CEO Jane Fraser's "remarkable turnaround"; the one clear bright spot vs the AI/energy/inflation drumbeat.
Counterparty on the West Texas data-center power deal — signed a 20-yr PPA with Microsoft (a majority of the generation from GEV's turbines). Referenced as the deal party, not rated.
Named only as a peer (Ep 76, Sep 21): with United, the profit yardstick in guest George Noble's Southwest pitch — "half of what United or Delta made." No view.
The named face of the "cheaper open-source Chinese models" enterprises are switching to "in order to control costs" — the substitution that turns the LLM layer's absent moat into a live price-war risk: "the Chinese models are much cheaper, and this could eventually cause a price war." Host exhibit alongside it (Artificial Analysis): DeepSeek V4 Flash at ~3¢ per test versus $1.86 for GPT-5.6 and $3.15 for Claude Fable 5.
Delaware Life (Mark Walter-linked insurer, private)
The suspected Carvana "mystery buyer" (Ep 75, Sep 14): a Mark Walter-linked insurer that "allegedly… owns a heck of a lot of Carvana paper," with related-party transactions up from ~3% to 30–40% in court filings — the catalyst the short needs if it ever fails.
Collateral damage from the FICO fight (Sep 18) — Pulte "also criticized the credit bureaus for price gouging"; Equifax −25% YTD vs TransUnion −15% because it "has a bigger percentage of its profits from mortgage scoring." Earlier: Workforce Solutions (EWS) is "a software business" caught in the SaaSpocalypse, so "there is no room for error" — government revenue down mid-single digits vs flat guidance, Q3/Q4 EPS guides 3% light, stock −4% on top of −17% YTD. "No one is interested in excuses."
Named with MGIC and Radian as the mortgage insurers FHFA's Bill Pulte is now "going after" (Sep 18) for being insufficiently pro-consumer. "We shall see how this evolves." No position.
The cheap-stock turnaround template he wants Charter to follow — $27 (Nov 2023) to $78 on performance; 1Q beat across the board (rev 43.6B, EPS 3.70 vs 2.60 est), raised guidance despite US auto sales −10%.
No longer a meme stock and no longer interesting (Sep 11): profitable and raised its outlook on collectibles, but revenue $790M vs $972M, "EPS growth is slow, and revenue growth is negative"; a ~$9B cap that "has done nothing in years" — "no one seems to really care anymore."
A clean beat read through the noise — EPS 62c vs 43c est, revenue +32% on options, equities and prediction markets. The after-hours drop on crypto revenue −38% was the market mistaking diversification for deterioration; "by Thursday morning, the stock had reversed and was up."
Subscriber view (Sep 11): Nvidia's acquisition of Hugging Face ("I believe, last week") read as arming the open-source/open-weight challengers to its hyperscaler customers — the "Robin Hood" argument against the circular-financing framing.
"Clearly struggling" — in line with its negative pre-announcement (−25% that day), revenue forecast cut to +4–5%, mainframes −42%, infrastructure −7%: soaring tech-equipment prices push customers to lock in hardware and abandon IBM's services.
Subscriber view (Sep 11), passing: one of the upstream "sell the knives" equipment suppliers (with ASML, Lam, Applied) that benefit whoever wins the AI knife fight. Not adopted by Eisman.
The capital-heavy contrast in the Golar pitch (Ep 75, Sep 14): an onshore Louisiana liquefaction build that "cost them… billions of dollars," vs a floating ship. Comparison only.
Named with Schwab as the other retail platform whose best margin is the spread on un-optimized customer cash, facing the identical three-part question: will these firms let third-party agents in, will regulators allow that mobility, and "will Schwab and LPL have their own agents that optimize cash and make it a service that they charge for?" — the outcome that converts a hidden spread into a smaller but far more defensible fee.
Subscriber view (Sep 11), passing: in the "sell the knives" upstream-supplier list — cheaper inference drives more demand for semiconductor manufacturing capacity. Not adopted by Eisman.
Rotate away — rumored to be doing a Google-like equity raise; a capital-raiser investors are "rethinking," vs AI beneficiaries that don't need capital.
The Chinese lab whose Kimi K3 is claimed "as good as any LLM out there, but for a fraction of the cost" — the event that promotes AI's no-moat problem into a live price-war risk: "now, the possibility of a price war looms closer."
Subscriber view (Sep 11), passing: named with Broadcom as the custom-silicon partners hyperscalers use to cut their Nvidia dependence — "but this takes time."
Pulte's next target (Sep 18): after FICO and the bureaus, the FHFA head is "going after the mortgage insurance sector," pushing MGIC and peers to be "much more proconsumer" and tell borrowers when the insurance is no longer necessary. "We shall see how this evolves." No position.
The Q2 scarcity blowout — AI data centers caused a shortage of every kind of semiconductor; prices soared. EPS $2.51 = +1,215% YoY, rev 41.5B = +345% YoY; up 267% into the print yet 2027 PE only 8.7×, and mgmt sees chip supply constrained past 2027. Eisman's cleanest "play scarcity, own the supplier" name.
"Seems to have completely lost its mojo," down 25% in Q2 — "Netflix's growth story no longer seems that powerful." The big drag inside otherwise-positive communication services.
Good results the market wouldn't pay for — EPS 768 vs 682 est, record $105B backlog (+$20B), guidance raised, yet the stock didn't rise: "the cost growth on the company's missile programs seem open-ended and that is hurting current margins."
Europe's #5 (~$260B staples) — "I like chocolate as much as anyone… but it's hardly a compelling investment story," the punchline of the avoid-Europe case.
Sep 18: OpenAI "alone represents 300 billion of Oracle's 600 billion plus backlog" — his proof the labs cannot slow down.The most counterparty-concentrated name in the AI chain, and therefore the cleanest public read on it. Sep 11 print: revenue and non-GAAP EPS both +30% ("in some ways, a very good quarter"), but RPO only $664B vs $600B the prior quarter and ~50% still owed by OpenAI, "a company that we all know has massive negative cash flow" — stock way off its highs on the over-reliance. Earlier: "50% of its $600 billion backlog is just from OpenAI" (Aug 28; "around half" of a "600-plus billion" backlog on Aug 15). Still counted among the hyperscalers with "real businesses," but half of its contracted future revenue is owed by the private, loss-making company he calls the weak sister.
Bewildered by the warrants, not the partnership (Sep 11): custom AI chips + optical networking with Amazon to compete with Nvidia, plus warrants on 25M shares at $161.26 worth >$4B — not circular financing, and Qualcomm doesn't need cash. Subscriber answers: a customer-acquisition cost for a ~$60B pipeline and AWS as anchor, or payment for Amazon IP given unproven XPU expertise.
Named with MGIC and Essent as the mortgage insurers FHFA's Bill Pulte is now "going after" (Sep 18) for being insufficiently pro-consumer. "We shall see how this evolves." No position.
The show's central disagreement. Worthington rates it a Buy — clients can already "click a button" to optimize, the firm need not open up to third-party agents, and regulators won't permit five-basis-point money to stampede between banks. Eisman objects to the shape of the model: "most of its revenue is from the idea that its customers aren't maximizing their own benefits… the business model is dependent upon basically the customer being lazy… not an equilibrium long-term." Both concede the customers "don't know." 2022's +500bp move showed what walks — and that a working-capital floor stays.
Couldn't borrow $10B (then $6B) against its $60B OpenAI stake — banks will take OpenAI public at the valuation but won't risk their own balance sheets.
The worked example of the one-trade problem — long AI hardware, short AI software, levered 4:1, blown up in July. "Directionally he's right. It's just a matter of bad risk management" — "Totally." The analogy is the transferable part: buying every auto company and shorting every horse-carriage company in 1905 is one trade, because "your longs and your shorts are not independent of one another. They move exactly in unison" — so a temporary sentiment reversal ends the fund before the thesis can be proved right.
The other half of the ratings duopoly — some investors assume "AI could eat into the Moody's and S&P duopoly"; Eisman: "I do not believe that is possible." Referenced via the MCO thesis.
The semis-worry tell — Samsung's operating profit rose ~1,800% yet the stock fell 7% "as investors fear that hyperscaler AI growth will slow thereby hurting semi-pricing" (which still "continues to go up"). Eisman uses it as a signal the AI worry is spreading from operators to suppliers, not a Samsung-specific call.
The largest creator of stablecoins (Circle is #2) — named as the incumbent leader as the Stripe/Visa/Mastercard/Coinbase/BlackRock payments consortium moves into the space.
The bureau least exposed to the FICO/FHFA fight (Sep 18): Pulte "criticized the credit bureaus for price gouging," but TransUnion is down only 15% YTD vs Equifax's 25% because less of its profit comes from mortgage scoring. No position.
The first link in Luria's AI value chain — "the companies that make the stuff that makes chips, primarily ASML and TSMC" — a layer that captures value whether the winning model is open or closed source. Ives adds the field note: a Taiwan fab "working 18 hours a day," chip demand:supply 15:1.
Named only as a peer (Ep 76, Sep 21): with Delta, the profit yardstick in guest George Noble's Southwest pitch — "half of what United or Delta made." No view.
Token-pricing-pushback exhibit — "went through its entire AI budget for the year in 4 months" once usage pricing kicked in; evidence customers may use AI less as they get cost-conscious.
The mechanism of his FICO short: now "a 10% market share of new mortgage loans securitized and I expect that percentage will go much higher" (Sep 18), after moving from a 21-lender pilot to available to all lenders (Sep 11), with the FHFA's Bill Pulte publicly attacking FICO's pricing.
Argentina's national oil company, cited by the guests (Ep 75, Sep 14) as planning a $30–50B onshore LNG build — the long-dated alternative Golar's ships bridge. Named inside the bullish-Argentina thesis; no position stated.
The "notoriously bad business" in his airlines-vs-suppliers analogy — capital-intensive, no pricing power; compare its 10-yr chart to TransDigm's. Hyperscalers risk becoming "airlines."
The AI-kills-consulting fear made real — "quite disastrous": rev +3%, bookings −3%, three same-day acquisitions that "look like buying stuff to hide weakness." Stock −18% Friday, −50% YTD, −35% in Q2. "Ugly." Eisman's headline AI victim.
A $29.5B BDC whose 85%/14%/1% software-loan risk buckets Eisman distrusts — "low risk" only means current; using ServiceNow (−60%) the equity beneath is worth <½ cost. "Who refinances the loans when the equity value is less than half?"
Rejected from two directions. Eisman: the only thesis he has ever heard is the fiat-debasement hedge, "and my issue is that crypto trades inversely to its own thesis" — it rallies on risk-on days and falls on frightening ones. Worthington, bullish on the asset class, exempts it: "a single-use case technology… store of value… we aren't building on Bitcoin the same way we're building on Ethereum and Solana," plus the concentration risk that "half the market's Bitcoin" with the least usage — "I don't have a lot of conviction in Bitcoin." Eisman's lighter version: the young speculators moved to Kalshi, so "Bitcoin is for boomers."
The episode's most surprising negative (Verrone, Ep 73). Regulated utilities are making new lows and now "the unregulated ones, the more power producers, the CEGs and the Talons and the Vistra's have also started to weaken" — despite the best demand environment they have ever had. Trennert's diagnosis is political, not fundamental: "the politics are starting to consume the utility sector."
Named with Intel as a valuation that only works "if this cycle is continuing through 2030" — "their current valuations are not otherwise justified." The expensive half of Luria's Micron-vs-Intel dislocation, and the reason owning both sides means holding two incompatible views of the cycle.
Capitulated (Jul 31) — "Bottom line, I give up. Simply put, I made a mistake." The Jan recommendation rested on cheap + improving fundamentals; instead 1Q26 broadband losses deteriorated and 2Q lost 172k, "much worse than expected." "Two really bad quarters in a row is enough for me. I'm selling… I hate thesis creep," and holding for cheapness alone would be exactly that. Told viewers before selling; the week's 15% bounce to $142 is AI-rotation money, "nothing to do with Charter."
The legislative premium stalls (Sep 18) — the expected Clarity Act "explains, I believe, the rally in cryptocurrencies and in the stock price of Circle"; senators blocked it, "a major blow for the crypto industry."A short that "doesn't make any sense" (Ep 75, Sep 14) — offered first when the Big Short partners were asked for a short: "Circle doesn't make any sense to me" / "Zero," yet "the stock has rallied enormously" on "the crypto legislation nonsense." Earlier: down 17.5% on June 30 when a consortium of Stripe, Visa, Mastercard, Coinbase and BlackRock unveiled a rival stablecoin ecosystem — the #2 stablecoin creator (Tether is #1) suddenly facing the payment-rail incumbents. "The importance of having Visa and Mastercard as part of this consortium cannot be overstated."
Short — the mystery buyer of paper north of 110 (guests Daniel & Collins, Ep 75, Sep 14): 75–100% of pre-tax income from gain on sale on largely subprime auto loans; Ally pays 102–104 yet the average is 109–110, allegedly because Mark Walter's Delaware Life buys the rest (related-party share up to 30–40% in filings). Walter says no fraud, "the market believes it" — "I got squeezed." ~30–40× earnings; getting paid needs the buyer to fail.
The largest of the entertainment incumbents who "sold the rope" — they licensed their out-of-syndication libraries to Netflix as "found money with 100% margins," bragged about it on conference calls, and "chose margin over growth and it cost them everything." The scoreboard is the argument: Netflix >$300B vs Disney $178B.
"Facing the same fate as Blockbuster" — the speculative cohort that powered online betting and crypto has migrated to prediction markets (Kalshi). "Being early is no guarantee of survival. Adapting is."
"Lacks a thesis" — consumer solar in a years-long depression ($335 end-2022 to $33), EPS −31%, Q2 guide in line but no improvement; "hard to see why the stock will move up."
The reported evidence that crypto is "no longer cool" — a multi-asset social brokerage whose biggest category is crypto trading, so revenue fell 30% on the trading decline even with okay EPS; −14% on the print, −19% YTD, −47% over 12 months.
Losing share for years, admitted over-earning; new CEO Mike Lyons leaving mid-turnaround to run Truist (−11%). 2026 PE ~6× — "the market does not believe in the turnaround."
"I remain short FICO" (Sep 18) — "its monopoly in mortgage scoring is going to break"; down 43% YTD, VantageScore already "a 10% market share of new mortgage loans securitized and I expect that percentage will go much higher," and the FHFA head "agrees with me."All three short (Ep 75, Sep 14) — Daniel & Collins "are short FICO" / Eisman: "Me too," shorted the day Pulte spoke: "companies gouging customers" — prices +1,600% where "inflation plus 2%" would have gone unnoticed, and mortgages charged "3 to 5x" card/auto pricing "just because they can because of the law." Still short (Sep 11) — FHFA's Bill Pulte is attacking FICO for price gouging ("raised prices by something like 1,600%" in 5 years) and VantageScore went from pilot to available to all lenders: "I think FICO's monopoly is going to break and break badly, and I still like this short." Earlier: short for several months on the VantageScore mortgage pricing war — Score 10 T $2,049 vs VantageScore $99 per 100 apps; FHFA approval in months, pilots 2026 / full 2027; mgmt's "won't lose share" call "could be terribly wrong." A monopoly that pulls data from the bureaus it now competes with.
An addiction business model — Magic / D&D reinvented on scarcity-FOMO fuels most of a ~$12B company's profit; manipulating kids and adults over making creative toys.
Ives' yardstick for the US chip lead — "a third-rate Nvidia chip is a year and a half to two years ahead of Huawei in China… any big Chinese tech company would want an Nvidia chip over Huawei." The only way that reverses, he argues, is domestic data-center moratoriums.
A blowout that proves the nervousness rather than curing it — EPS 42c vs a 10c loss, best revenue growth in 15 years, data-center sales +59%. But "given CapEx budgets, it would be surprising if Intel did not have a good quarter"; it's the budgets themselves investors fear.
The SaaSpocalypse's first body — PE-owned software milked instead of reinvested: "it happened in Medallia last week… this is why there's distress around private equity." Luria's proof that the software carnage is a private-tier event, since the listed names "are in a net cash position… software debt is private equity."
The defended layer — a hyperscaler whose capital intensity is the moat: "the amount of money it takes to be a hyperscaler is insane and that expenditure itself is a moat," and "Google and Microsoft have multiple revenue streams from established businesses which are very unlikely to simply disappear." Caveat kept in his own words: "their LLM businesses are also questionable," and ~45% of the backlog is OpenAI.
"A turnaround story for 2 years that has not turned around." Evercore's Michael Benetti cut it buy→hold "because there is just no evidence that the turnaround is taking hold. Quite the opposite." Down 21% in Q2.
"Very frightening" concentration (Ep 76, Sep 21) — "the largest company on planet Earth basically 70% of its revenue comes from five customers" (70% of receivables), and lenders aren't lending to them; guest George Noble: Nvidia's $250B package for OpenAI "tells me all I need to know." Still long, "less long" — and explicitly not short (Aug 28), against Michael Burry: "I am still long Nvidia and several other tech stocks. I'm less long cuz I've gotten nervous. But I am not short." Hedging "possible" but undisclosed. The worry is the customer list, not the product: revenue more than doubled but "70% of its accounts receivable was from five companies." And the headline $500B third-party facility with KKR/Apollo/Blackstone/BlackRock is not signed — per Bloomberg the deal was dragging and "Jensen went out and basically just announced it anyway… right now there is no $500 billion."
"In trouble" (Ep 76, Sep 21) — the ecosystem "is dependent upon two companies who lose… billions… and one of them I think is in trouble, which is open AI"; guest George Noble: "Totally agree," citing Nvidia's $250B funding package as proof the ROI isn't there. Can't afford to slow down (Sep 18) — Altman agreeing the industry should slow down is "false on its face" when OpenAI "alone represents 300 billion of Oracle's 600 billion plus backlog"; with "token maxing" ending and open-weight models taking share, the doomsday talk is a bid for regulation that "will foster an AI duopoly.""The weaker company" — and the safety alarm is a moat play (Sep 17, CNBC): "between the two, OpenAI is the weaker company. I think that's one reason why they postponed their IPO"; the labs have "no moats… whatsoever" as open-weight models take share, so they are "trying to manufacture a crisis" for regulation and a duopoly — "really postpone your IPO.""Open AI is in trouble" (Ep 75, Sep 14) — June-quarter revenue $6.5B (+18%) against $12B of costs: "their revenue went up a billion and their cost went up three"; narrative turning, people leaving, cost of capital rising — "whoever buys OpenAI out of bankruptcy, it's going to be a fantastic deal." "The weak sister" / "the problem child" (Aug 28) — now ranked explicitly below Anthropic on the sequential math: June-quarter revenue ~$6.5B, up only 18% q/q, while costs went to $12.5B, up $3B in three months — "revenue in 3 months was up a billion dollars sequentially and its costs were up 3 billion sequentially. That's not the right direction." No moats ("or at best, the moats are shallow"), enterprises switching to cheaper Chinese open-weight models, and with token maxing over "it's going to be harder for these guys to raise capital." The stakes are macro: "if tomorrow OpenAI failed, the US economy I think would go into an immediate recession."
"Paramount is at a lowly 10 billion" — the extreme endpoint of the margin-over-growth trap in the Netflix/Blockbuster case study, against Netflix's $300B+.
The K-shaped counterweight to Visa and Mastercard — EPS $1.43 vs $1.48 (−3%) and, "perhaps worse, organic revenue growth was 0%." Zero organic growth in an inflationary year means falling volumes at the mass-market consumer while card payment volume grows 8-10%.
Guest George Noble's example of "narrative dominance" in financials (Ep 76, Sep 21): a "crappy BNPL lender" with too-low loss reserves "on six times book" — at FrontPoint "you'd be pulling your hair out… Oh, but you don't understand. This is SoFi." Disparaging, not a stated short.
The damage gauge for the semis leg of the AI trade — "down 23% from its peak on June 22nd and… down 4% this week," against a NASDAQ −7% from its Jan 2 high. The gap shows the selling is concentrated in the AI supply chain, not spread across the market.
Guest George Noble is short (Ep 76, Sep 21) — "built to fail": the float is unlocking from ~5% to 20–25% with "every few weeks 7% more," and it is "really an AI play… with a rocket thing attached" ($22T of the $29T TAM); Starlink $200–300B, SOTP $30–60/share vs ~$140 and a $1.7T cap. Eisman: data centers in space face heat, not cold. Earlier (Eisman): "Not a fan" of the IPO — now an AI bet (85% of TAM), capex jumped 42%→215% of revenue, AI commoditized/no moats; "just not playing."
Named with Constellation and Vistra ("the Talons") among the unregulated power producers weakening in an environment "you would expect to be great" (Verrone, Ep 73) — the market repricing the whole merchant-power category as data-center power bills become politically visible.
Guest George Noble is short (Ep 76, Sep 21) — US sales −26% in July, Q3 laps last year's credit pre-buying, capex to ~$30B for AI ("seriously cash flow negative"), robotaxi worth "800 million not 800 billion" on Austin ride data; SOTP $30–60/share vs ~$360, the stock "unchanged over 5 years." "Don't short a cult" (Ep 75, Sep 14) — guests Daniel & Collins have been short on and off for 5–6 years, right on earnings (Eisman: "there's no question fundamentally you've been right") while the stock went up. A one-question stock (Sep 11): 2026 EPS $1.66 = 220× PE vs GM's 6.5×, so "the future of this company depends on the success of the robotaxi" — and the robotaxi event again showed Musk "over-promising and under-delivering" (−6%). EPS is 59% below its 2022 peak yet the stock has done nothing for 5 years — "how difficult it can be to short a cult stock." No position; "count me a skeptic." Earlier: earnings down four straight years; a capital-intensive, hyper-competitive EV business (China far cheaper) — doesn't want it folded into "X."
The third merchant generator flagged as rolling over (Verrone, Ep 73). Set against the same episode's positive stance on the build-out contractors (Quanta, EMCOR, Caterpillar, Schneider, Siemens Energy), it sharpens the distinction: the equipment and construction layer gets paid on backlogs, the generation layer gets paid on a politically visible consumer price.
At $69B in the incumbent-market-cap tally behind Netflix's $300B+ — one of the studios whose short-term, comp-driven decision to sell Netflix its old shows armed the disruptor. "They sold a rope to Netflix, and Netflix hanged them all."
In one line (Sep 21, Ep 76 with George Noble):a fellow ex-PM is even more bearish than he is — "He may be more negative on AI than I am" — and the two agree Bessent's rate repression is "a complete waste of time." Against $40T of debt, Treasury buying is "trying to kill a whale with a BB gun," and upward pressure on rates is "mostly because of AI, not so much the deficit" (without AI debt, rates "probably 40 basis points lower"). On the AI chain he repeats his own numbers — ~70% of hyperscaler AI revenue from OpenAI + Anthropic, OpenAI "in trouble," Anthropic racing its S-1 out "because token maxing is ending," and Nvidia's 70% of receivables from five accounts "very frightening" — while guest Noble supplies the yen/JGB mechanics (JGB 10-year "should be around 7%," the US swap line that spared Treasuries), his "narrative dominance" coinage, and public shorts in SpaceX and Tesla (SOTP $30–60/share each). Recorded ~Sep 2, published Sep 21. Previously (Sep 18, The Weekly Wrap):the labs' doomsday talk is a shell game — "the entire future of Anthropic and OpenAI depends on there not being any slowdown." Hundreds of billions of hyperscaler commitments (OpenAI alone ~$300B of Oracle's $600B+ backlog) and an Anthropic IPO "maybe in a month or two" make "slow down" "false on its face"; with token maxing ending, open-weight models taking share and "no pricing moats," they foment hysteria to invite regulation that "will foster an AI duopoly" — a gambit Trump's no-regulation stance has already sunk, though the data-center backlash stays. Rates: the Fed hiked to 3.75–4%, and "for now, 5% on the 10-year does seem to be the demarcation line," pushed up partly by ~$500B of AI debt crowding out Treasuries. Still short FICO (−43% YTD; VantageScore at 10% of new securitized mortgages "and I expect that percentage will go much higher"), with Pulte now turning on the mortgage insurers; the blocked Clarity Act is "a major blow" to crypto and Circle's rally. To diversify from AI: staples, healthcare and low-vol/P&C ETFs (LVHD, SPLV, KBWP); to hedge gains, short against the box. Previously (Sep 17, CNBC):the AI-safety alarm is "all nonsense" — the labs are "trying to manufacture a crisis." No evidence of AGI; token maxxing "is over," open-weight models are "taking big market share," and with "no moats around their business whatsoever" the frontier labs want regulation they can shape into a duopoly. The chain from Nvidia through the hyperscalers "all depends on the future health of Anthropic and OpenAI," and "OpenAI is the weaker company… one reason why they postponed their IPO." He has "taken some off the table" and isn't adding. Previously (Sep 14, Ep 75):his old Big Short partners are more bearish than he is — and he says so. Vincent Daniel and Porter Collins hold "the majority of our capital" in gold because interest plus entitlements leave only printing, and read Bessent's buybacks as "theatricality and deception"; Eisman: "I'm not nearly as hung up on the deficit as you guys," though "I'm a little more nervous than I've been." On AI the three agree on the numbers — Nvidia's top five customers are 70% of receivables, OpenAI's revenue +$1B vs costs +$3B in a quarter ("Open AI is in trouble"), enterprises cutting AI spend ~60% by routing to Chinese open-weight models, and SpaceX/Anthropic IPOs as the supply that "killed the market" in 1929 and 2000. Shorts: FICO (all three — "Me too"), Carvana (the gain-on-sale "mystery buyer"), Circle; the Tesla lesson "don't short a cult"; longs Golar LNG and Glass House (which Eisman also owns). Previously (Sep 11):he concedes the rate level and the AI destination, and keeps his whole bear case on timing. The 10-year has breached 4.9% (highest since November 2023) on an Iran war with oil "well above $100" — "I thought 4 and 1/2% was the Rubicon and I'll admit I was wrong" — and the Treasury's $4B→$6B long-bond buyback against $40T of debt is so small he assumes a bigger plan: "it's always best to ask, 'What am I missing?'… there's some other shoe that could drop here." On AI he grants a subscriber's long-run case (Nvidia arming its customers' challengers, compute commoditizing, "sell the knives") but not the path: the current ecosystem is "almost completely dependent, on the health of Anthropic and OpenAI," one could fail before replacements arrive, as dot-com's first generation did, and "if OpenAI fails within a year, there will be a massive correction" — with OpenAI's GPT 5.6 Luna price cut a possible "harbinger of a price war" and Oracle's RPO stalling at $664B, still ~50% OpenAI. Positions: still short FICO ("its monopoly is going to break and break badly"), a skeptic on Tesla at 220× vs GM 6.5×, constructive on Macy's turnaround, and no gold — "5% seems like the better choice to me." Previously (Aug 28):the dependency map now carries a number on every link, the two labs are finally ranked against each other — "between the two, Open AI is the weak sister" — and the position is on the record: still long Nvidia, "less long," and explicitly not short. The chain, all from public disclosure ("I don't think it's hidden… it's not like I discovered the Rosetta Stone"): US GDP up ~2% this year and "half of it is from AI CapEx"; Nvidia's revenue more than doubled but "70% of its accounts receivable was from five companies"; "70% of hyperscaler AI revenue is from Anthropic and OpenAI… about 25 to 35% of their cloud revenue"; "50% of [Oracle's] $600 billion backlog is just from OpenAI." Hence the macro statement of a stock-market thesis: "if tomorrow OpenAI failed, the US economy I think would go into an immediate recession, and the market would have a massive correction" — put to him at dinner-party level, "that's 100% accurate." The metric is second-half revenue growth, not profit ("eventually profitability matters, but near-term these stocks don't care"), because "token maxing… ended" around late June/July, so H1 measures a spending mood that no longer exists and "the third and fourth quarters are going to be much more interesting." The ranking is arithmetic: Anthropic ~$11.5B in the June quarter, +100% q/q, versus OpenAI ~$6.5B, +18%, with costs to $12.5B, "up 3 billion in 3 months… that's not the right direction." Everywhere outside AI he is the contrarian against the doom trades: the Treasury's $4B buyback is "a joke," spending the $1T TGA to cap long rates "absurd," QE "had absolutely zero impact on the US economy — the only impact it had was causing stock prices to go up," Dalio's reserve-currency argument is "nonsense" because "there is no alternative" (the $3T repo market "is all T-bills, nothing else"), and the banking system "has never been this safe in anyone's lifetime." Only housing is genuinely damaged — priced out by taxes, P&C insurance and utilities rather than by the house price. Previously (Aug 17):the bear case survives the cross-examination — but the market's own internals refuse to confirm it, and the level that would end the cycle is "much higher than people think." Ep 73 (rec. early August, pub. Aug 17) puts Strategas' Jason Trennert and Chris Verrone opposite Eisman, and the answer that comes back is a threshold, not a forecast. The leaders already crashed and nobody noticed: semis and hyperscalers fell "30 40%… some 50%" from mid-May, yet breadth improved through it — 50% of the S&P above its 200-day at the June 2 high versus 75% at taping — so "money doesn't want to leave the asset class of equities." Hence the governing question: what rate competes? "We thought it was 450… but now it seems higher… if we're really going to end the cycle, it's a level much higher than people think. It's not 470" — with Japan '89 (JGBs four to eight), Nasdaq '99 (10s four to seven) and '87 (long rates six to nine, stocks +30%) as the calibration, and Trennert's warning inside it: a market that ignores rising long rates is at "the point at which you have the most risk." Eisman's own closing takeaway: "the economy is, without question, quite strong," the 4.5% kibosh "has not happened," so "the 10-year really has to go significantly higher" before this breaks — and absent that, "they're pretty bullish on the market for the rest of the year." On AI the two strategists reach Eisman's own dependency map independently: the hyperscalers "have a real business and there are real moats around it, but it's a lot more capital intensive," while "the Anthropic's or Open AI's of the world… are totally negative cash flow" and carry "70% of the AI hyperscaler revenue" — "if Open AI and Anthropic ever get in trouble, the ecosystem is in trouble." The cash-flow scoreboard is Trennert's: Meta $785M for the quarter, Microsoft ~$19B but −25% y/y, Amazon negative over twelve months, Google's $85B equity raise "shocking," Oracle "just downgraded" — "I don't see how you do it without hitting the bond and equity markets." His sentence — "it's going to be hard to continue to grow AI investments without further dilution of shareholders" — got him fired by a client, itself the episode's sharpest sentiment datapoint. And yet: "I wouldn't short it. We're market-weight technology," and he is "a little more bullish… on the semiconductors." The sector sheet is unusually clean. Leadership: banks everywhere ("every bank stock around the world is at new highs" — Morgan Stanley, Goldman, Japanese and European banks; double-B spreads at new cycle lows; Dan Clifton's "extreme period of financial deregulation"), traditional asset managers breaking out for the first time in years (State Street, T. Rowe, Invesco, plus "Ben's been a great chart") in what Strategas calls an "alpha market" — 2026 "has been about the E, not the P/E" — and the AI build-out contractors that already corrected ~30% and have responded (Quanta "up like 15% on earnings day," EMCOR, Caterpillar), held under a falsifiable rule: give them the benefit of the doubt "until they attempt to rally, don't make new highs, and fail." Defensives ranked:health care first — "left for dead" life sciences (Thermo Fisher, Danaher, Illumina) putting in "major major bottoms," managed care (UnitedHealth, Cigna) already bottomed in a call Verrone admits "I got too dogmatic on… and I missed it"; staples last but bottoming (Kraft Heinz, Hershey "worth a look"); and utilities the outright avoid — regulated names at new lows and even the merchant producers (Constellation, Talen, Vistra) weakening, because "the politics are starting to consume the utility sector." Geography: Europe is "cheap for a reason" (not dynamic, no AI, German GDP flat for four-five years) and its record Euro Stoxx is only Schneider Electric, Siemens Energy and banks — the same AI trade in euros — while Trennert would "rather stay here at home. I like Japan a lot better as an investment, and I'm invested there" (deflation ending, an ROE religion enforced by delisting rules, higher JGB yields as reflation not rupture, with Japanese banks/insurers as the tripwire). And the demand tell that reframes the macro: "copper new high, zinc new high, tin new high, Chinese iron ore is collapsing" — "a China demand problem, not an AI CapEx problem." Before that (Aug 14): the AI thesis got a named single point of failure — and an explicit rule for why he isn't trading it yet: "there is no such data set with respect to AI" (Aug 14). The hyperscalers "at least on the surface are not the problem — the LLM providers are": no moats around LLMs, constant model switching, cheap Chinese open-weight models that are "good enough," so "a price war could break out." The dependency math is the thesis: ~70% of hyperscaler AI revenue and 25-35% of total cloud revenue is Anthropic + OpenAI, and half of Oracle's $600B backlog is OpenAI alone, while both labs "lose billions and are reliant at this point on raising capital for their survival" — "a bigger version of Situational Awareness, a huge one-way bet. What's the hedge? There is no hedge for the LLMs." If they fail, hyperscaler margins compress and "the entire AI chain goes into reverse. This is the potential Achilles heel of AI." But the position is a trim, not a short, and he says exactly why: pre-GFC he bought Moody's securitization database and "every month it reaffirmed our thesis. That's what gave me the confidence to short subprime paper" — for AI, "there is no such data set… Until then, we have supposition… Stop rushing to call the top or the bottom," and "timing is everything… it might occur a year from now and for markets that is an eternity." So he "got cautious a few months ago and sold some," but is "still quite long" (S&P +13%, NASDAQ +14% YTD), working from what is knowable: strong economy, sound employment, benign bank credit, rising capex — plus Nvidia's $500B financing consortium with six asset managers, which he reads constructively as "a tried-and-true way to create a financial infrastructure to support a capex hungry growth industry" (the CRE / toll-road template) that could improve hyperscaler FCF, pays fees to banks and private credit, and proves "there is as yet no slowdown in AI capex." Winners/losers: MSFT/GOOGL/AMZN the hyperscaler winners, ORCL "questionable with its debt barely rated above junk," META squeezed (no cloud, competing head-on with the labs; revenue +28% vs expenses +55%, FCF ~$785M). Alongside it, an upstart-vs-incumbent framework that cuts both ways: the entertainment incumbents sold Netflix "the rope" for "found money with 100% margins," "chose margin over growth and it cost them everything" (NFLX $300B+ vs DIS $178B, WBD $69B, PARA $10B) — but NFLX itself is −21% YTD because "the upstart has grown old"; the counter-case is Visa/Mastercard, who "link billions of consumers with hundreds of millions of merchants… will fight to the death to defend their turf" and beat every challenger, which is why Circle picked the wrong hill ($30→$240→$71, an $18B cap) and "if I was a CEO, I would try to sell the company." Bitcoin fails his falsification test outright — the debasement thesis requires it to rise when NASDAQ collapses; "instead… the fact that it generally tracks the NASDAQ is the clearest indicator that there really is no thesis" (−27% YTD, −46% 12mo) — and it has lost the cohort too: "Bitcoin is no longer the cool toy. Kalshi is," with DraftKings "facing the same fate as Blockbuster" and eToro revenue −30% on the crypto-trading decline. Also: Iran will not be re-bombed and economic pressure alone "will have difficulty succeeding" (the regime "does not care if its people suffer"), so the conflict runs long — "but as long as there's no bombing, the market will probably march higher"; the SaaS-apocalypse narrative "has taken a bit of a vacation" with the private-credit software refi data due "sometime next year"; and boards of directors are "mostly window dressing" — CEOs handpick them, and in the GFC most directors of the major financial institutions "had no background in financial services at all." Before that (Aug 10): the private-credit reckoning he has been circling all year finally has a date — "we are a year away from finding out what we're going to really need to find out" (Ep 72, rec. Jul 30 / pub. Aug 10, with Evercore's Glenn Schorr and JP Morgan's Ken Worthington). The arithmetic that dates it: ServiceNow −50% from its peak means "any software company owned by private equity has to be down something like 50% from… where it was bought," and ~$270B of sponsor-held software maturities in 2028-29 means "it's somewhere in the next two quarters, three quarters where they'd normal course of business be refinancing" — so the lender-vs-sponsor stalemate (very few keys handed back, very few equity cheques written) breaks in early-to-mid 2027. Wealth-channel gross inflows into direct lending are now "infinitesimal," redemptions still sit above the 5%-per-quarter gates (same sellers, not new ones), returns have fallen from "mid-teens plus" to "high single digits, mid single digits," and stress will arrive as terms — opportunistic funds taking the paper 200-300bp wider with LME/EBITDA-adjustment blockers, which "will eventually bring markdowns that you haven't seen yet." Blue Owl is the named technology-heavy direct lender; KKR's record monetizations are the exception in a private-equity industry underperforming public markets "for the first time… probably since '08 or '09." Alongside it: the lazy-cash debate (Schwab/LPL earn their best margin on cash customers don't optimize — a Buy to Worthington, "not an equilibrium long-term" to Eisman, with agents "wrong the majority of the time" so the threat is real but not imminent), crypto re-framed around use cases with Bitcoin explicitly exempted ("half the market's Bitcoin"; it "trades inversely to its own thesis"; "Bitcoin is for boomers"), Circle told to sell because Visa/Mastercard "are not dumb" and already launched a 140-institution consortium stablecoin while the banks build interoperable tokenized deposits, and a near-perfect big-bank quarter (IB +38%, trading +47%) that worries him precisely because "when everything is humming… it's always possibility that tomorrow that's not going to hum." Before that (Jul 31): he capitulated on his own January recommendation — "bottom line, I give up. Simply put, I made a mistake" — and sharpened the AI map into two layers (Jul 31). Charter is sold after two straight bad broadband quarters (172k losses in 2Q, worse than 1Q): "two really bad quarters in a row is enough for me… I hate thesis creep. And continuing to own the stock just because it's cheap, and it is cheap, would be thesis creep" — announced to viewers before the sale, with the week's 15% bounce dismissed as AI-rotation money that has "nothing to do with Charter." At the same time he inverts his own bear point: hyperscaler capital intensity IS the moat — "the amount of money it takes to be a hyperscaler is insane, and that expenditure itself is a moat… there are only going to be a few" — so Google/Amazon/Microsoft/Oracle "have real businesses here," while the LLM providers (Anthropic, OpenAI, and the model arms of Google/Microsoft) have "no moats or at best… shallow" ones against cheap Chinese open-source models, and their health is the named catalyst: "a key thing to monitor to determine a catalyst for a real sustained selloff is the health of Anthropic and OpenAI," since half of Oracle's $600B+ backlog is OpenAI. He also flips on jobs — "job dislocation, but net job creation" — and calls the destruction story possible "propaganda propagated by Anthropic and OpenAI so that the federal government will step in and regulate AI to their benefit." Underneath, he has already acted on the one-trade view — he sold his long-held Google "a couple of months ago… to reduce my exposure to AI" and is sitting in cash (a lightening, not a short; no replacement bought, because "buying something defensive or stapley is kind of a waste of time… they don't go up"), since "it's all one trade. It's literally one" — even a 60/40 mix is fake diversification (>50% of the 60 is tech/AI-related, and "most of the new issuance of bonds is AI related"), which makes the outcome binary: "it better succeed," and if it doesn't, "I think we have a big correction." Behind that, the AI capex reckoning has arrived — the market has flipped from cheering capex raises to punishing them, so free cash flow is now the scoreboard: Google lifted 2026 AI capex $190B→$205B and printed FCF of −$5.9B (−7%), Tesla's capex took FCF negative for the first time in two years (−14.5%), and Nasdaq fell >2% on the pair — while Moonshot's Kimi K3 ("as good as any LLM… for a fraction of the cost") moves the China price war from theory to near-term risk and the Iran/Houthi escalation puts oil at $100 and the 10-year at 4.7%. What still works: real order books and real moats (GE Vernova orders +88% — "I still own the stock"; Moody's duopoly at a sub-30× PE; Lockheed's record backlog), not the spenders; and the narrative victims whose numbers are fine (ServiceNow) stay victims because "fighting the AI narrative is like fighting a ghost." He owns no banks — at peak valuations with an IB cycle "heavily dependent on AI financing needs," a bank is "just one more aspect of the AI trade." Underneath, the whole market is still a single AI bet — 50%+ of the S&P is tech/tech-related, ~half of 2026 GDP growth is AI, and 15%/50% of existing/newly-originated corporate debt is AI-related, so even 60/40 gives "no real diversification"; and AI has turned capital-intensive — a "sea change" — so software cash-machines (Google, then likely Meta/Microsoft) must now raise stock to fund the build; rotate away from the capital-raising hyperscalers toward the suppliers/beneficiaries that don't need capital (power gen, semis, networking — "own the TransDigm, not the airline"); meanwhile AI is a commodity with no moats (weekly leapfrogging; OpenAI already cutting token prices) whose six-point bear case keeps sharpening, only Visa/Mastercard are real moats in payments, addiction models and private-credit illiquidity are spreading, you should avoid Europe (too little tech), and the Iran "deal" is just a 60-day MOU — while the bull leg still stands as long as Nvidia's revenue growth (+85%) holds.
Ep 76 — "Stock Picking in Difficult Times" with George Noble (Sep 21; recorded ~Sep 2).Rates: Bessent's buybacks can't move a $40T market — "a whale with a BB gun" (Eisman) — and Noble casts him as Norman Lamont defending "an artificial rate" with "open mouth operations"; the regime changed, with the market underestimating inflation "since 2022… because we have fiscal dominance," and AI debt issuance crowding out Treasuries (~40bp, Eisman's rough number). Japan: Noble's Robin Brooks chart puts the JGB 10-year's fair yield near 7% vs just over 4%; the yen went 80 → 160; the JGB–UST gap is down to ~180bp, so Japan could "bring the money home" from Treasuries and the Nasdaq; after a ~$95B intervention week the US lent Japan dollars through a swap line rather than let it sell Treasuries — "who's going to buy the bonds?" AI: Noble's "show me the ROI" — Nvidia's $250B package for OpenAI "tells me all I need to know," Garran's 24× dot-com malinvestment, Berezin's Global Crossing lesson; Eisman: the circular financing "tells me that there's probably something wrong," and NVDA's 70% receivables concentration is "very frightening." Names (Noble's): short SpaceX (float 5% → 20–25%, +7% every few weeks; Starlink $200–300B) and Tesla (US sales −26% in July, capex to $30B, robotaxi "800 million not 800 billion"), $30–60/share each; long Southwest (his January pick, down a few %); SoFi as "narrative dominance" at six times book.
The Weekly Wrap — "why Amodei and Altman are faking the AI doomsday crisis" (Sep 18).AI: a departing Anthropic researcher's extinction post, Amodei's call to slow capability gains ("for too long, the industry lied or downplayed the real risks") and Altman's agreement are, to Eisman, Kevin-Mitnick-style hysteria about poorly understood technology, from an industry whose mass-layoff forecast "just don't come true." His theory: both labs "have hundreds of billions of commitments to hyperscalers" (OpenAI "300 billion of Oracle's 600 billion plus backlog") and Anthropic IPOs "maybe in a month or two," so they cannot slow down; business is getting harder ("token maxing is ending and open weight models keep taking market share… no pricing moats") as data-center and capital costs rise, so they want regulation that "will foster an AI duopoly." Trump's refusal to regulate makes it "a complete failure," but "the damage is done" in local data-center politics, and "it's unclear to me how to keep a shell game going in an IPO process that requires transparency." Rates: Fed +25bp to 3.75–4% with another hike pencilled in; the 10-year briefly above 5% on oil near $110 and ~$500B of AI debt "crowding out" Treasuries; Bessent's buyback "worked only for one day," so "he needs a much bigger bazooka or an alternative buyer." Names:FICO — "I remain short," the FHFA "agrees with me," VantageScore 10% of new securitized mortgages; bureaus hit too (Equifax −25% on bigger mortgage mix, TransUnion −15%); Pulte next "going after" MGIC, Essent, Radian; Circle's rally undone by the blocked Clarity Act. Mailbag: P&C stocks trade on pricing, not float income; only staples and healthcare (14% of the S&P) are AI-uncorrelated, so use sub-sector/low-vol ETFs (LVHD, SPLV, KBWP); hedge big embedded gains by shorting against the box (NVIDIA as the example); Europe's big four stay slow despite Sweden and Poland.
CNBC Squawk Box — "the companies are trying to manufacture a crisis" (Sep 17). The Terminator/AI-safety narrative is "all nonsense": "there's no evidence at all that AI has achieved artificial general intelligence… most of the evidence seems to point that it never will." What is really going on: token maxxing "is over," "the open weight models are taking big market share," and the labs "realize that there are no moats around their business whatsoever," so they are "trying to manufacture a crisis that will create regulation… to create the duopoly that they want." He is "a little bit more nuanced" on the "eight bubble signs" list but repeats the chain thesis: Nvidia → hyperscalers → Anthropic/OpenAI, with OpenAI "the weaker company" and its postponed IPO the tell; he has "taken some off the table." On Nvidia's Hugging Face purchase (after OpenAI agents broke into it), Nvidia "might have been willing to pay more… to protect OpenAI." To labs urging a slowdown on safety: "really postpone your IPO."
Ep 75 — the Big Short partners reunite, and the deficit splits the table (Sep 14, Vincent Daniel & Porter Collins).Rates and gold:Collins — interest plus entitlements already swallow the tax take, "you certainly can't cut interest," nobody will cut or tax, so "they're screwed," and "I've had as much conviction in this precious metals [position] as I've had in anything in a long, long time"; gold "doesn't have any debt" while central banks hold "someone else's liability." Daniel puts himself in Bessent's shoes: end the war (oil 86→60) — refused; cut entitlements, raise taxes — no; "so then you start with theatricality and deception by playing twist games," buying 6–9 months for inflation expectations to fall. Eisman dissents: "I'm not nearly as hung up on the deficit as you guys… I don't think… it's eminent doom and destruction at this point," 50bp higher rates are "not the worst thing in the world," and Bessent's error was "putting a line in the sand. Now he has to defend it" — while conceding "I'm a little more nervous than I've been" and that they were "kind of right." Close: they print in a crisis and in the status quo, so gold "is where the majority of our capital sits." AI: Eisman's own 10-Q read — Nvidia revenue +100% but "top five direct customers… 70% of accounts receivable"; Anthropic ~$11.5B (+100% q/q) vs OpenAI $6.5B (+18%) with costs $12B, "revenue went up a billion and their cost went up three… Open AI is in trouble." The partners add a channel check (an enterprise's tokens +30–40%, AI costs −60% by routing to Chinese open-weight models — "how do we get to the returns on invested capital?") and a supply warning: hyperscalers issuing debt and equity, and Anthropic/SpaceX IPOs as "new stock of supply… that's what killed the market" in 1929 and 2000. Shorting: the pod shops (5:1 levered, beta- and factor-neutral, covering "in 3 weeks") crowd every fundamental short, so "what we used to do for a living is brutal"; a short pays only when something "brings light." Shorts named: FICO (all three — "companies gouging customers"), Carvana (gain on sale 109–110 vs Ally's 102–104 implies a buyer "north of 110," allegedly Mark Walter's Delaware Life — "I got squeezed"), Circle ("doesn't make any sense"), and Tesla as the lesson: "don't short a cult." Longs:Golar LNG (floating liquefaction toll roads for Argentina's trapped gas under Milei) and Glass House — Eisman: "I own it too" — on exporting $90–100/lb California cannabis to $600–700/lb markets, "a 2027 event"; plus Google (owned for years) and ServiceNow (bought in the SaaSpocalypse).
Conceding the destination, keeping the timing — and the Rubicon moves again (Sep 11, The Weekly Wrap).Rates: Iran bombing the US and US retaliation put oil "well above $100" and the 10-year through 4.8% and then 4.9%; "the equity markets, in my view, will not be able to tolerate some level of higher long-term rates. What that level is, no one really knows. Now, I thought 4 and 1/2% was the Rubicon and I'll admit I was wrong." Bessent's Treasury buyback ($4B, upped to $6B; the 10-year still rose 4.80%→4.845% that day) "needs a bigger bazooka" — the irony being that Warsh is leaving QE just as "Treasury is stepping into its shoes" — and against $40T of debt the size is so wrong that he inverts the question: "Bessent needs a large buyer of Treasuries, and it can't be the Fed or Treasury. So, I think there's some other shoe that could drop here." He flags Bessent's partisan convention speech as spending gravitas "he could use," expects Trump's threats against a hike to "be ignored" (60% odds priced for September), and thinks Warsh "will do whatever he wants to do." AI: answering subscriber pushback on the Ed Zitron interview (Sep 09) — Nvidia at ~74–80% gross margin "extracting rent from hyperscalers" to fund their challengers (Hugging Face, neoclouds, open weights), hyperscalers building their own silicon with Broadcom/Marvell, and the "sell the knives" list (ASML, KLA, Lam, Applied, Micron, SK Hynix) — he grants the destination: "this could be correct, but with one major caveat… one of them could fail before these new companies are capable of replacing the commitments of Anthropic and OpenAI… like the dot-com bubble bursting," and restates the conditional: "I still believe that if OpenAI fails within a year, there will be a massive correction in the stock market." Supporting evidence the same week: OpenAI's CFO disclosed a price cut on GPT 5.6 Luna with "a tenfold increase in model usage" ("Is this the harbinger of a price war?"), and Oracle's +30% quarter came with RPO only $664B vs $600B, ~50% OpenAI. The Terminator story (a former Anthropic researcher's extinction warning) gets "count me a skeptic — there is still no evidence that AI is ever going to achieve AGI," with Asimov's Foundation offered as the sci-fi that is predictive (mass behaviour is forecastable, "the entire premise of all current loan underwriting"). Names:Tesla at 220× 2026 EPS vs GM 6.5× is a one-question robotaxi bet on a CEO who over-promises, and a lesson that shorts were "right on the fundamentals" (EPS −59% from peak) yet the stock "has done nothing for 5 years"; FICO — with Pulte back on the attack and VantageScore now open to all lenders — "I still like this short"; Qualcomm's $4B+ of in-the-money warrants to Amazon leave him "bewildered… not by the partnership, by the warrants"; Macy's a turnaround "beginning to work" on Bloomingdale's +11.3% comps; GameStop "no one seems to really care anymore." Gold vs Treasuries: gold rallied as the 10-year rose because the yield rise is war-driven inflation fear, but "I don't own gold and I think our government is stretched but not yet overstretched. 5% seems like the better choice to me."
The dependency map, priced and ranked — and the first explicit split between the two labs (Aug 28, The David Lin Report). The most complete statement of the AI thesis in the archive, and the first where every link carries a number: ~2% GDP growth of which "half is from AI CapEx" → Nvidia's "70% of its accounts receivable was from five companies" → "70% of hyperscaler AI revenue is from Anthropic and OpenAI… about 25 to 35% of their cloud revenue" → "50% of [Oracle's] $600 billion backlog is just from OpenAI." The two labs are no longer one exposure: "between the two, Open AI is the weak sister" / "the problem child" — Anthropic did ~$11.5B in the June quarter, +100% sequentially; OpenAI ~$6.5B, +18%, with costs at $12.5B, "up 3 billion in 3 months… that's not the right direction." The tell is second-half revenue growth, not profitability — "eventually profitability matters, but near-term these stocks don't care" — and H1 is discounted in advance because "token maxing… ended" around late June/July, so "the third and fourth quarters are going to be much more interesting than the first half." The scheduled verdict is Anthropic's S-1 (prediction markets: 93% it lists first). Positioning, stated plainly against Michael Burry's short: "I am still long Nvidia and several other tech stocks. I'm less long cuz I've gotten nervous. But I am not short" — hedges "possible," undisclosed — because "that risk could materialize a year from now, and in markets timing is everything." The consequence is stated at macro scale: "if tomorrow OpenAI failed, the US economy I think would go into an immediate recession, and the market would have a massive correction." Also new: the $500B Nvidia capital facility does not exist yet — per Bloomberg the deal was dragging and "Jensen went out and basically just announced it anyway."
The Google sale confirmed, and the three-domino version of the AI case (Aug 15, New Money clip commentary). A third-party video built around Eisman clips — the host narrates, so only the quoted passages are his — but it is the tightest compression of the thesis yet. One: the pure LLM layer has no protection — "there just don't seem to be any moats, or at best, the moats are shallow. Enterprises are switching between models and using cheaper open-source Chinese models in order to control costs," and "the Chinese models are much cheaper, and this could eventually cause a price war." Two: that layer is the hyperscalers' order book — "of Oracle's 600-plus billion backlog, around half is from OpenAI" — so a price war transmits straight into listed capex names' forward revenue. Three: there is nowhere in a conventional portfolio to hide — of a 60/40, "more than 50%" of the equity sleeve is tech/AI-related and "most of the new issuance of bonds is AI related." Hence the single monitor: "a key thing to monitor to determine a catalyst for a real sustained sell-off is the health of Anthropic and OpenAI." Where he would still play, the capital intensity flips from bear case to moat: "the amount of money it takes to be a hyperscaler is insane and that expenditure itself is a moat… the hyperscalers like Google, Amazon, Microsoft, and Oracle have real businesses here," while "Anthropic and Open AI… don't have the breadth of revenue streams of Google and Microsoft." The unresolved tension is his own: the layer he defends is the layer he sold — "I sold my Google. I wanted to reduce my exposure to AI. What scares me is that it's all one trade. So it better succeed" (the Jul 27 disclosure, restated here as the video's premise).
Ep 73 — the bull answer is a threshold, not a forecast (rec. early Aug / pub. Aug 17, Jason Trennert & Chris Verrone of Strategas). Two strategists who cover the whole market rather than a sector are put opposite Eisman's AI worry, and they neither dismiss it nor trade it. The tape first.Verrone's central observation is that the drawdown already happened and the market absorbed it: "we just had a 30 40% decline in some of the leading stocks, whether it was the semis or the hyperscalers… some 50%," from mid-May through the week of taping — and through it the internals improved: "you only had about 50% of the S&P above the 200-day average on June 2nd. Today… 75%… even as the market has churned for the last 8 weeks, the internals have gotten better, not worse," so "money doesn't want to leave the asset class of equities." The historical control on that: "go back to '99 or 2000. If I told you the leading stocks, Cisco or Sun Micro, were going to be down 40 or 50% over an 8-week period, what do you think the S&P or the Nasdaq would be down? A lot." Hence the governing question: "I don't think we have yet to find the rate of interest that gets money to leave equities… it's higher than most people think." Strategas' old number was 4.50% and "now it seems higher… if we're really going to end the cycle, it's a level much higher than people think. It's not 470," calibrated against the bubble analogues — Japan '89 (JGBs "four to eight that year as the Nikkei was melting up"), Nasdaq '99 ("US 10s went four to seven"), '87 ("long rates went from six to nine percent while the stock market was up 30%"). Trennert's warning is folded into the same fact: "that's the worst sign, when stocks are going straight up, and they don't care at the same time long-term interest rates are going up… That's the point at which you have the most risk." Verrone's closing statistic cuts the other way and is the reason he isn't there yet: over "the 400 days since Trump was inaugurated, this is the narrowest range in 10-year yields that we've ever seen… 85 basis points… I just don't think this move in yield is as explosive yet as it's going to need to be if it's going to disrupt the equity market." The Fed: the "new regime" is "an old regime" — no policy statements before 1994, no post-meeting press conferences until 2019 — and Warsh's view, which Trennert shares, is that forward guidance "has been too big a free pass for the financial markets and capital to basically just take as much risk as they can"; participants "have gotten very used to being spoon-fed… Now they're saying, listen, you're on your own." Verrone's twist: "this may be what people have wrong about the Fed… it may wind up being they have to tighten because of growth," not inflation. Trennert expects a hike anyway while arguing against one ("the backup in interest rates has largely been in real rates. It hasn't been in inflation"). AI, arrived at independently: the hyperscalers "have a real business and there are real moats around it, but it's a lot more capital intensive. The Anthropic's or Open AI's of the world, that's where it's most questionable… they're totally negative cash flow," with the Chinese competing — and "something like 70% of the AI hyperscaler revenue is from just those two companies. So if this is the big if, if Open AI and Anthropic ever get in trouble, the ecosystem is in trouble" (Eisman: "I think it's undeniable"). The scoreboard: Meta quarterly cash flow $785M, Microsoft "the best of anybody" at ~$19B but −25% y/y, Amazon negative over twelve months, Google's $85B raise "shocking given the fact that the last time they actually raised equity capital was in the 2000s," Oracle "just downgraded." Trennert's published line — "it's going to be hard to continue to grow AI investments without further dilution of shareholders" — "we got fired by a client… He said I don't want to read the work anymore." And yet the position is market-weight, not short: "I wouldn't short it… I do think the demand is exceeding the supply for artificial intelligence," with a preference for "the semiconductors" and the caveat that the hyperscaler layer "could be a little bit of a race to the bottom." Verrone's counterweight is sentiment: the negativity "reached a fever pitch… when Oracle got to like 120," and since then "Google's basically back at the highs. Amazon just made new highs. The 5-day change in Microsoft is the best we've seen since the IPO in '87" — "at a minimum I think they're tradeable." Where the money is working: the AI build-out contractors that already corrected ~30% and "have all responded" — Quanta ("up like 15% on earnings day last week"), EMCOR, Caterpillar — under a written exit condition, "give these the benefit of the doubt until they attempt to rally, don't make new highs, and fail"; banks everywhere ("every bank stock around the world is at new highs… it seems a little out of step" — Morgan Stanley, Goldman, Japanese and European banks, double-B spreads at new cycle lows, "an extreme period of financial deregulation" per Dan Clifton, and Trennert's curve-steepening/minimal-credit-losses read); and the alpha market — "2026… has been about the E, not the P/E," so "stock picking and active matters a lot more than passive," confirmed by State Street, T. Rowe, Invesco (and "Ben's been a great chart") "starting to break out here for the first time in years." Defensives, ranked:health care top — "pharma and biotech and life sciences have been leadership… these reversals in some of these left for dead life science stocks are particularly notable. The thermos and the Danaher's and the Illumina's… I think they put in major major bottoms"; managed care (UnitedHealth, Cigna) already bottomed — "I got too dogmatic on those and I missed it. We try not to do that"; staples last but with "hints of life" (Kraft Heinz, Hershey "worth a look," though "I would hardly call the sector leader"); and utilities the avoid — regulated names "making new lows" and even the merchant producers, "the CEGs and the Talons and the Vistra's," weakening, because "the politics are starting to consume the utility sector" — which sits inside Trennert's wider populism read: "there is a reaction against big anything. Big media, big academia, big pharma, big banks," and that morning, big oil. Geography: Europe is "cheap for a reason… they're not dynamic economies. They don't have AI… just trying to stay in the game" (German GDP "has basically not grown for the last four five years, not a dollar"), and its record Euro Stoxx is onlySchneider Electric, Siemens Energy and banks while autos and luxury sit in "prolific bear markets" on Chinese demand — i.e. the same AI trade in euros, not diversification. Trennert: "I would rather stay here at home. I like Japan a lot better as an investment, and I'm invested there" — deflation ending and an ROE conversion enforced by "delisting companies that don't make certain minimums in terms of price to book or return on equity" (the old market-share-over-ROE Japan being "a little bit reminiscent of what's happening in tech here in the US"). On the JGB scare: higher yields "entirely reflect this reflationary animal… until the Japanese insurance stocks or… bank stocks start to really weaken here, I'm not concerned," and the closing datapoint — "Japanese 10-year yields are about to cross German 10-year yields for the first time in decades." Macro tells: the metals split — "copper new high, zinc new high, tin new high, Chinese iron ore is collapsing right here… not a China message here at all. It's AI and CapEx… a China demand problem, not an AI CapEx problem" (iron ore at 52-week lows) — plus a tepid crude rally "back under 80" that corroborates Eisman's oversupply guest. Two portable lessons: the buggy-whip trade — long every automaker and short every buggy-whip maker in 1900 is a correct thesis that still dies at four times leverage on one bad headline, "because you're long X and you're short Y, but it's the same thing" (his explanation of how Situational Awareness blew up with the VIX under 20) — and "the lawsuit is the punishment" (Posner on IBM): IBM won its 12-year antitrust case "but they really lost because it consumes so much of their time and effort that they missed personal computing," the template Trennert applies to Meta's hundreds of social-media suits. Meta itself is the episode's clearest negative on Eisman's own numbers — "revenue grew 28%… and the expenses grew 55%… we want the reverse," with depreciation "starting to really explode… from like 4 billion to 6 billion, it'll go to 8, then 10, then 12. It's like a weight on your shoulders" — while Verrone respects only the tape ("3 days later it's recovered the entire loss"). Eisman's lessons learned: "the economy is, without question, quite strong… the big issue is whether all this money being spent on AI is going to prove to be good, but we won't know that for a while," and "the surprise to them was that they thought that if the 10-year got above 4.5% that would really put a kibosh on the stock market… that has not happened, and so their view now is that the 10-year really has to go significantly higher… If those two things don't happen, I think they're pretty bullish on the market for the rest of the year."
AI's Achilles' heel — the dependency map, and the discipline of not calling it yet (Aug 14). The clearest single statement of the AI view in the archive, and it comes with its own epistemics attached. The method first: the reason he could short subprime in 2006 was not conviction but a refutable dataset — his team bought Moody's securitization database, where "every month, each securitization reported 30-day, 60-day, 90-day delinquencies and real estate owned and losses," and from summer 2006 "every month it reaffirmed our thesis. That's what gave me the confidence to short subprime paper." For AI, "there is no such data set. When OpenAI and Anthropic go public, we will have some real data. Until then, we have supposition… Stop rushing to call the top or the bottom. Accept that we don't have enough information to make the exact right calls and work with the information that we do have." So the expression is a trim, not a short: "a few months ago, I too got cautious and sold some of my positions, but I'm still quite long" with the S&P +13% and NASDAQ +14% YTD, and "timing is everything… this bear case… might occur a year from now and for markets that is an eternity." What is knowable: the economy is strong, employment sound, mid-July bank credit data benign, hyperscaler capex still rising — plus the week's one big new fact, Nvidia's $500B financing consortium with six large asset managers, treating compute infrastructure "much like the commercial real estate toll roads or other assets you can borrow against." His read is deliberately unexcited and constructive: "this is not a brand new concept. It's a tried-and-true way to create a financial infrastructure to support a capex hungry growth industry" — it moves data-center funding off hyperscaler balance sheets into securitizations, so "hyperscaler free cash flow could, I emphasize could, improve," it pays fees to "the banks and private credit companies that are participating," and it proves "there is as yet no slowdown in AI capex." (Note this cuts against the circular-financing reading of the same deal elsewhere in the hub — worth cataloguing as a live disagreement.) The thesis: "the hyperscalers at least on the surface are not the problem. The LLM providers are the potential problem — specifically Anthropic and OpenAI." No moats around LLMs, "users switch between models all the time," Chinese open-weight models are much cheaper "and enterprises seem to be using the Chinese models more and more" — "overcharging for tokens is going to be very difficult with good enough competition," so "a price war could break out." The numbers that make it systemic: research firms put "70% of hyperscaler AI revenue… and 25 to 35% of total cloud revenue" on those two, and "Oracle has a $600 billion backlog and half that backlog is from OpenAI alone" — while both "lose billions and are reliant at this point on raising capital for their survival." Hence: "it feels like a bigger version of Situational Awareness, a huge one-way bet. What's the hedge? There is no hedge for the LLMs," and should they fail "the margins of the hyperscalers will compress… the entire AI chain goes into reverse. This is the potential Achilles heel of AI. That is what I am watching most carefully" — but "until this Anthropic and OpenAI risk metastasizes, the AI story will continue." The sheet:Microsoft, Google, Amazon the hyperscaler winners ("the hyperscalers have existing franchises"); Oracle "questionable with its debt barely rated above junk"; Meta squeezed — no cloud, competing head-on with the labs and the Chinese providers, revenue +28% against expenses +55% and free cash flow of only ~$785M. Earnings confirm the flow layer: CoreWeave revenue +112% and +19% on the day with "demand and pricing… very strong," Super Micro revenue +91% / EPS $1.70 +315% and also +19% (competitor Dell), Cisco EPS $1.22 +23% with raised guidance yet down after hours on whisper numbers (peer Arista) — "as long as the hyperscalers keep spending, AI infrastructure players… can't help but benefit." The second half is a standalone framework on upstarts vs entrenched franchises. Netflix bought the incumbents' out-of-syndication libraries; "the incumbent entertainment businesses said yes with glee. This was their titanic moment" — "found money with 100% margins" they bragged about on conference calls, because "CEOs are compensated annually… stock results drive compensation." Lenin's line supplies the verdict: "we will hang the capitalists with the rope they will sell us… they sold a rope to Netflix, and Netflix hanged them all." They knew streaming was lower-margin and "were right about that. They chose margin over growth and it cost them everything" — NFLX >$300B vs Disney $178B, Warner Bros $69B, Paramount "a lowly 10 billion." Yet the winner isn't exempt: growth is slowing, "growth investors don't like investing in companies where growth is deteriorating. The upstart has grown old," hence NFLX −21% YTD. The mirror image is Visa/Mastercard: challengers "show up every few years… and they fail every time," because "Visa and Mastercard link billions of consumers with hundreds of millions of merchants. Go recreate that," and critically "they are not going to get complacent like the incumbent entertainment companies… they will fight to the death to defend their turf." Which is why Circle ($30 IPO → $240 in sixteen days → $71) picked the wrong hill — investors "could not be more wrong" that payments would be easy, inroads "will occur by teaming up with Visa and Mastercard, not by fighting them," and at an $18B cap "I don't think it has the financial strength to play this game for too long. If I was a CEO, I would try to sell the company." Same test kills Bitcoin: no articulated thesis, and the debasement hedge is falsified by its own tape — it should rise when inflation soars and NASDAQ collapses, but "the fact that it generally tracks the NASDAQ is the clearest indicator that there really is no thesis" (−27% YTD, −46% over 12 months) — plus the cohort has left: "Bitcoin is no longer the cool toy. Kalshi is," with DraftKings "facing the same fate as Blockbuster" ("being early is no guarantee of survival. Adapting is") and eToro printing revenue −30% on the crypto-trading decline, −14% on the day. Elsewhere: Iran — no resumed bombing, economic pressure only, and after the premium Steven Cook (CFR) interview he concludes that "will have difficulty succeeding" because "the Iranian regime does not care if its people suffer. They care about the regime's survival," so the conflict runs long — "but as long as there's no bombing, the market will probably march higher." Software: "the SaaS apocalypse narrative seems to have taken a bit of a vacation as software stocks have all rallied off the bottom"; winners are AI-value-add newcomers plus incumbents who reinvest, and the private-credit refi of PE-owned software is "around the corner sometime next year" — "we need data and that data will emerge soon." And a closing governance sermon: boards are "mostly window dressing" because "CEOs handpick the members of their board," and during the GFC "most members of the boards of the major financial institutions had no background in financial services at all. They were professors. They were once highly ranked government officials… I was horrified."
Ep 72 — private credit gets a calendar, and the lazy-cash model gets challenged (rec. Jul 30 / pub. Aug 10, Glenn Schorr & Ken Worthington). Two sell-side analysts covering the same balance sheets from opposite ends turn Eisman's weekly private-credit worry into arithmetic. Schorr's flow read: wealth-channel gross inflows into direct lending are so small "you can almost not see the line… the gross sales inflows into the funds are infinitesimal," redemptions remain "over the 5% per quarter limit" in most funds but it is "the same people asking… not as many new people," and nothing has broken because the borrowers "are still cash flowing. They still have margins." Eisman's mark: "ServiceNow is down over 50% from its peak. So, any software company owned by private equity has to be down something like 50% from… where it was bought" — which makes every manager's "our borrowers are fine" an answer to the wrong question ("I don't think it matters… it's just that the valuation of the company is half"). The clock: "if there's 270 billion of software maturities held by financial sponsors that have maturities in 2028 and 29, it's somewhere in the next two quarters, three quarters where they'd normal course of business be refinancing," and today the industry is "in a bit of a stalemate" — very few keys handed back, very few equity cheques written. Hence "we are a year away from finding out what we're going to really need to find out." Schorr's escape valve is repricing rather than default: opportunistic credit funds will take the paper "at a 300, 200 basis point higher yield," blocking LMEs and EBITDA add-backs, so it "will eventually bring markdowns that you haven't seen yet" — with returns already down from "mid-teens plus" to "high single digits, mid single digits." Blue Owl is the named technology-focused direct lender ("big exposure and their stock has suffered"); KKR's record monetizations are the counterexample in a PE industry that "for the first time in a very long time, probably since '08 or '09, has underperformed the public markets," where LPs "have guns to their head." Second thread — lazy cash: Schwab and LPL earn their most lucrative margin on cash customers never optimize; Worthington rates SCHW a Buy (clients can already click a button; brokers needn't open up to third-party agents; regulators won't allow five-basis-point money to stampede — "you have runs on the banks"), Eisman objects to the shape ("the business model is dependent upon basically the customer being lazy… not an equilibrium long-term"), and both concede the customers "don't know." Timing comes from the agents themselves: JP Morgan's internal studies say they are "wrong the majority of the time" (they "tell investors what they want to hear based on the prompts"), and the bar for delegating money is "like 99.99%" — so "not a threat immediately." Third — crypto re-framed as use cases, with Bitcoin exempted: Worthington's tokens-follow-chains framework Eisman calls "actually coherent," but Bitcoin is "a single-use case technology… store of value… we aren't building on Bitcoin the same way we're building on Ethereum and Solana," and "half the market's Bitcoin" is the concentration risk — "I don't have a lot of conviction." Eisman's own test: it "trades inversely to its own thesis," plus "Bitcoin is for boomers" now that the young speculators are on Kalshi. Fourth — stablecoins/payments: "I think Circle should sell" to someone with bigger pockets, because unlike the studios that "acted like idiots and let [Netflix] grow," Visa and Mastercard "are among the smartest… not asleep," already shipping a consortium stablecoin with 140 institutions while the banks build interoperable tokenized deposits — "breaking into the payment system is brutal." Worthington's nearer-term case for Circle: dollarization wallets (Tether's turf), cheap cross-border, and a 24/7 settlement layer for weekend trading the traditional back office can't clear, plus interchange unbundling. Finally the banks: IB +38%, trading +47%, wealth +7% organic, "huge ROEs on high capital bases… almost as good as it gets" — Eisman: "actually stunning" — with only second-half seasonality (trading historically −17%) and deposit competition against it, and Schorr's rule that "when everybody agrees on something, trading's not going to be so good." Eisman's closing tell: "everything is just humming, which in some ways worries me because when everything is humming, it's always possibility that tomorrow that's not going to hum."
Charter: the thesis-creep sell — "I give up. Simply put, I made a mistake" (Jul 31). The clearest reversal in the archive. His January long rested on two pillars — "insanely cheap" and "fundamentals would get better" — and the second one failed in sequence: 4Q improving, 1Q26 deteriorating, 2Q26 "more deterioration again with 172,000 broadband losses, which was unfortunately much worse than expected." Management says the worst is over and paid down debt; "however, at this point, I'm skeptical… Two really bad quarters in a row is enough for me. I'm selling." The discipline is the point: "if the fundamentals ever turn, I could come back to this stock, but I hate thesis creep. And continuing to own the stock just because it's cheap, and it is cheap, would be thesis creep" — i.e. once the fundamental pillar breaks, owning it for the multiple alone is a different thesis. Two more transferable habits: he refuses to read the tape as evidence (the stock went $123, a 52-week low, to $142 in a week, but "the rally in the stock in my view has nothing to do with Charter… investors are reallocating out of AI related plays"), and he tells the audience before he trades — "I have not yet sold my position. I recommended the stock to my viewers and I strongly believe that I should not sell until I inform my viewers of my opinion change. I will be selling the stock next week."
The AI reframe — hyperscaler capex is a moat, LLM providers have none, and their health is the catalyst (Jul 31). He splits the trade he used to treat as one: hyperscalers (the data-center builders) versus LLM providers (the model makers), noting Google and Microsoft are both. Capital intensity — the core of his old bear case — becomes the barrier: "the amount of money it takes to be a hyperscaler is insane. And that expenditure itself is a moat… there are only going to be a few," so Google, Amazon, Microsoft and Oracle "have real businesses here. What the returns will look like, I don't know yet." The model layer is the fragile one: "there just don't seem to be any moats or at best the moats are shallow. Enterprises are switching between models and using cheaper open-source Chinese models in order to control costs" — and Anthropic/OpenAI lack "the breadth of revenue streams of Google and Microsoft." Hence the monitor: "a key thing to monitor to determine a catalyst for a real sustained selloff is the health of Anthropic and OpenAI… because so much of the hyperscaler backlogs are from these two companies. For example, of Oracle's 600 plus billion backlog, around half is from OpenAI." The bull leg he keeps: cheaper software → "an age of massive amounts of startups," and AI as net job creation — with the job-loss narrative possibly "propaganda propagated by Anthropic and OpenAI so that the federal government will step in and regulate AI to the benefit of Anthropic and OpenAI. Demanding regulation based on a false narrative would be a very disturbing way to create moats." Markets alongside: Iran re-escalated (oil >$90), the Fed held and "some investors are afraid that the Fed is behind the curve," the 10-year "dangerously close to 4.7%," NASDAQ −7% from its Jan 2 high and SOXX −23% from Jun 22 — and a credit cycle appearing, with CoreWeave paying "more than, get this, 9%" while lenders "discriminate between the large companies like Google that they know can pay the money back and smaller newer companies." Earnings read: Microsoft good (Azure re-accelerating to 43%, FCF −23% accepted) vs Meta "quite bad" (expenses +55% on revenue +28%, R&D $13→22B, FCF $784M "basically nothing," capex floor raised, $279B of off-balance-sheet AI leases, +53% in three months); Visa +10% payment volume and Mastercard +8% against P&G at 0% organic (the K-shape); Quanta +71% EPS on the electricity demand; Bloom Energy +680% EPS at 73×→37×→23× ("if the AI story keeps going"); Meritage mixed but buying back 5% of shares YTD; FICO (his short) −17% because "a company whose entire monopolistic business model is potentially under assault can show no signs of weakness"; PayPal sluggish and playing chicken with a $60 Stripe/Advent bid ("the company should sell"); Apple disappointing on services and China; Amazon strong on AWS +37%. Plus a risk-management sermon: the $20B Situational Awareness fund was "forcibly liquidated to meet margin calls" — "there is more to managing a hedge fund than just being smart and knowledgeable."
Ep 70 — the analysts' answer: hunt dislocations, and the SaaSpocalypse is mostly private (Jul 27, Dan Ives & Gil Luria). Eisman presses his three counterarguments (capital intensity, no moats, a Chinese price war) at two analysts who cover the whole stack. Gil Luria reframes it as a search for inconsistency: "we look for dislocations… where is the market being inconsistent?" — Micron at "six times earnings… as if the cycle is over" against Intel at 100× and Cerebras priced as if "this cycle is continuing through 2030," while memory is now the better market than CPUs. His winners: Microsoft (accelerating on Azure/Office/infrastructure, "the raw end of both deals," low multiple), Palantir, Micron; his loser: Salesforce — "has not been adding value to them in years and it keeps charging them more and more for that less value every year… a bad business that's been declining, regardless of AI." The mechanism is the CIO's budget: fund AI by cutting from 100 software packages to 30, and "guess what I'm cutting? All the small ones" — so the real carnage is the PE-owned private tier that was milked instead of reinvested (Medallia failed that week), while listed software is net cash: "software debt is private equity." On Oracle he tells the September-2025 bandwagon story ("I'm always worried when everybody's on a bandwagon") and lands on the asymmetry: after OpenAI's $122B raise and "code red," the "$630 billion worth of backlog of compute revenue is valued by the market at zero." Google is "a winner, not the winner" — model no longer state-of-the-art, distant second in consumer chat and third in enterprise AI, but search advertising accelerated. Dan Ives holds the bull line: year 3 of an 8–10 year build-out ("the Vegas strip, 1955"), chip demand:supply 15:1 in Asia, Nvidia "their world, everyone else paying rent" with an $8–10 multiplier per dollar and 18–24 months on Huawei, Apple as "the EasyPass on the consumer AI highway" monetizing 2.5B iOS devices, and cybersecurity budgets doubling (CrowdStrike, Palo Alto) — with Adobe and Intuit the names that "miscalculated," the 1995 typewriter company. Both close on Altman and Amodei "pulling the ladder": the job-loss narrative as a bid for regulation that shuts out open source and Chinese labs, "so those two are the only winners." Eisman's verdict: still doubts the moats, "especially Anthropic or OpenAI" — "I would recommend everybody watching it at least once, maybe twice."
He sold Google and is sitting in cash — the one-trade audit turned into a trade (Jul 27). The skepticism finally shows up as a position: "I sold my Google a couple of months ago. I've owned Google. I can't even tell you how long I've owned Google, but I felt I wanted to reduce my exposure to AI." He has not redeployed — "I'm just sitting. I've got cash" — and explicitly refuses the defensive rotation: "buying something defensive or stapley is kind of a waste of time… people either want to buy AI or they don't want to buy AI, but they don't want to shift out of it to buy Clorox." He would still not short ("I would not. I've lightened up"), and has no redeployment trigger: "I don't know yet… I don't think this debate is going to get settled within the next two weeks." The reasoning is the diversification audit made explicit: "it's all one trade. It's literally one" — even 60/40 holders "are actually not diversified," because "more than 50%" of the 60 is tech/AI-related and "most of the new issuance of bonds is AI related." So it's binary: AI succeeds and the market pushes higher, or "I think we have a big correction" (size: "that's a hard question. I don't know"). Crucially he does not expect the pop from disappointing capability — "it's going to be something really good. That doesn't mean that everybody succeeds" — but from the financing: the latest Nvidia deal (backing the financing so OpenAI can buy Nvidia chips) is "so incestuous… totally convoluted. You don't know if there's any real profitability in any of the machinations," though "there probably is for Nvidia. They're not doing it for free"; and Anthropic/OpenAI, "responsible for a lot of the spending that people are doing," run closed models "much more expensive than the competition" — "a potential bottleneck." On the other side of the tape he is calm on oil despite the Iran war he doesn't think ends soon (Trump wants the nuclear fuel surrendered, "the Iranians are not going to give up their nuclear fuel"): "there's an oversupply," China "has basically stopped buying oil," the Strait-bypass pipelines are "pretty close to being finished" (~2027) — "you would have thought oil prices would be 150 at this point." And this week's Fed meeting is "completely irrelevant": "if inflation gets worse, they'll raise rates. And inflation gets better, they'll cut rates."
The capex reckoning — a raise is now a negative catalyst, and FCF is the scoreboard (Jul 24). "A year ago, it was all rah-rah for AI. When companies raised their CapEx budgets, the market cheered… the terms of debate on AI have truly shifted." Proof in one night: Google — revenue +24%, cloud +82%, but (after stripping the investment gains it books in EPS) an adjusted miss, 2026 AI capex $190B→$205B and free cash flow of −$5.9B → −7%; Tesla — big EPS miss on margins (regulatory credits −67%) with capex at $5.8B sending FCF to −$1.1B, "the first time in 2 years" → −14.5%; Nasdaq −2%+. The third leg arrived first: Moonshot's Kimi K3, claimed "as good as any LLM out there, but for a fraction of the cost" — so capital intensity + no moats now carries a live China price war. Also: Iran/Houthi escalation → oil $100, 10-year 4.7%; SpaceX below its IPO price ("does not bode well for the IPO market"). What he still owns / rates: GE Vernova (EPS miss irrelevant in a long-tail business — orders +88% to $24.2B, backlog $176B; "I remain confident"), Moody's (a duopoly with pricing power at a sub-30× 2026 PE, "the cheapest… in quite some time"), Lockheed (record $230B backlog, +24% in 3 months) and Intel's blowout — though that print "won't quell AI nervousness" because it's simply the arithmetic of customers' capex budgets. Narrative victims: ServiceNow executes cleanly (EPS +10%, rev +24%, "no weaknesses") and still falls — "fighting the AI narrative is like fighting a ghost"; Equifax's EWS software business guides 3% light where "there is no room for error" and "no one is interested in excuses"; IBM "is clearly struggling" as soaring hardware prices push customers to lock in equipment and drop its services; Domino's is the K-shaped tell (same-store sales +0.1%, weakest in 5 quarters). Blackstone beat but exits keep lengthening, private credit's software loans face next year's refi cycle, and it now brags of being "the biggest financier of AI data centers." Mailbags: trust banks (BNY/State Street) are "a world all their own"; he owns no banks — peak valuations plus an IB cycle "heavily dependent on AI financing needs," so "buying banks… does not provide diversification from tech" (his last bank was Citi: "right thesis, but the wrong management"); plus a mechanics lesson on the funding short, and a bank-analysis masterclass coming.
The market is one AI bet — diversification no longer exists (Jul 10). Per Apollo's Torsten Slok, AI capex is ~100bps of 2026's ~2% GDP growth — "just about half" of all growth (the rest: ~30bps reshoring, ~90bps tax refunds). Tech + tech-related is 50%+ of the S&P, and 15% of existing / 50% of newly-originated 2026 corporate debt is AI-related — so a 60/40 stock-bond mix gives "no real diversification. It's all one trade. It's all AI. So AI better succeed and succeed big" or "the US is going into a recession and the market is going straight down." He walks the bear case as an evolving argument (his GFC-subprime method): first crack = Oracle's backlog 50% OpenAI (330→140); then the 2026 capex guides (Google $90B→$180B + an $85B equity raise, Meta $135B, Amazon $200B+, Microsoft rumored) turning cash machines into capital-hungry, no-moat businesses at price-war risk (Chinese AI, token-subsidy reversal, blown budgets, usage curbs). Investors migrated to semis — but even that shows worry (Samsung op profit +1,800% yet stock −7%). Also: Circle −17.5% as a Stripe/Visa/Mastercard/Coinbase/BlackRock consortium unveils a rival stablecoin (the incumbents co-opting the disruption — Visa/Mastercard's inclusion "cannot be overstated"); Nike flat on a cautious call; and his portfolio = 16 longs + cash, no bonds/gold/real estate, "hedges are in my shorts" (FICO one of them — an upcoming Unicus Research short deep-dive).
The Q2 report card — play scarcity, not the capital-intensive buyer (Jun 26). The Q1-bottom rally (S&P +15%, NASDAQ +23%) started to correct as "capital intensity + no moats" finally bit. Winners = scarcity: memory (Micron's +1,215% EPS / +345% rev blowout at 8.7× 2027 PE, supply tight past 2027; SanDisk — both +200%) and AI power (GE Vernova on the Chevron/Microsoft 20-yr West Texas PPA). Losers = AI's victims and the no-moat names: consulting (Accenture −50% YTD) and software (Salesforce/Adobe/Intuit/Gartner down 18–39%); Google's Gemini co-head left for OpenAI and a DeepMind engineer for Anthropic (no moats / no non-competes); Nike downgraded buy→hold (a 2-yr turnaround that won't turn). Sector reads: banks rallied (MS +34%, GS/Citi +27%) as recession fears faded; energy gave back its war gains (−13%); payments/crypto stayed weak (Fiserv, PayPal, Coinbase). Capital intensity is fine with a moat — without one it's "a race to the bottom… price wars… low returns on capital."
AI's capital-intensity "sea change." Google raised $85B all-equity — its first since 2005 — as capex jumps ~$80B (2025) → $180–190B (2026), beyond cash flow; Meta and Microsoft are rumored to follow, Oracle added $20B to its plans. Non-capital-intensive software has become capital-intensive hardware; "now shareholders are being asked to foot the bill."
The bull/bear snapshot. Bull: no sign AI capex is weakening, Nvidia +85% (accelerating) — "as long as Nvidia's revenue growth remains elevated, this story is not over." Bear (six sharpened legs): (1) capital intensity, (2) no moats — weekly leapfrogging among Gemini/Anthropic/OpenAI, ~$1T capex for a commodity, (3) OpenAI mulling token price cuts, (4) token-pricing pushback (MSFT Copilot June 1; Uber burned its budget in 4 months), (5) little/no ROI evidence, (6) power/water constraints + community pushback.
The rotation call — supplier, not operator. For investors who stay bullish, shy away from the capital-intensive hyperscalers toward power generation, semiconductors, and tech equipment (Arista, Cisco). The analogy: hyperscalers are becoming "airlines" (capital-intensive, no pricing power) while their suppliers become "TransDigm" — compare AAL vs TDG 10-yr charts.
Payments — only two moats. A "brutal space" where the only "impregnable franchises" are Visa (he owns it) and Mastercard; everyone else (Fiserv — share loss, over-earning, CEO bolting mid-turnaround for Truist, −11%) is a fragile price-taker.
Don't fight the US government. An export-control directive shut down Anthropic's top models — even for its own foreign-national staff — after a jailbreak surfaced (Amazon, an OpenAI investor, reportedly snitched). "Very heavy-handed… hurts the future of Anthropic." A new AI policy risk to price.
Macro & positioning. Warsh's first Fed meeting: hawkish hold, "raising rates is now a possibility." The Iran "deal" is a 60-day MOU, not a treaty (only benefit: Hormuz open 60 days; oil −5%, 10-year below 4.5%). In a stagflation hideout: energy and healthcare, not bonds/utilities; gold "not acting per thesis." Avoid Europe — overregulated, slow-growing, only ~18% tech vs ~38–50% in the US, and no hedge to a US recession. Picks: Meritage (1.0× book vs M&A at 1.25–1.3×), Glass House Brands (cannabis reclassification). Earlier themes persist — addiction business models (Kalshi, Hasbro), private-credit illiquidity (now BlackRock capping HPS redemptions at 5%), index-inclusion herding, and Citi's Jane Fraser turnaround as the bright spot.
The K-shaped consumer (earnings-season read, May 1). The economy is K-shaped: the top (Visa net rev +17%, payment volume +9%; Starbucks SSS +6.2%; GM/Lilly blowouts) is strong while "the bottom of the K consumers are in a deep recession" — Domino's SSS +0.9% vs 2.3% est, −8%, the cheap-staple tell. Read the split via a value brand's same-store sales vs a card network's volume. AI CapEx "pulls all the weight of GDP growth," so the market discounts the consumer weakness — "buy the dip until proven otherwise."
Value & the cheap-stock turnaround (Charter, May 1 — superseded; the position was capitulated on Jul 31, see the top bullet). His January Charter (CHTR) long was the playbook in miniature: CapEx rolling off (11B→7.5–8B) into FCF dollar-for-dollar → buybacks of up to 50% of shares on a $23B cap, 2026 PE ~4×, ~70% FCF yield in outer years; he added through an ugly 1Q (120k broadband-sub loss) because the pillars held — guarding against "thesis creep." GM ($27→$78) is the template. Counterweight: a fallen stock that "lacks a thesis" (Enphase) is a pass, "cheap" ≠ "investable." Still short FICO on the VantageScore pricing war ($2,049 vs $99 per 100 mortgage apps); and the private-credit refinancing question (Ares/ARCC, Blue Owl — who refinances software loans when the equity is worth <½ cost, with ServiceNow −60% as the proxy).
The product — The Real Eisman Playbook
What it is: a podcast/YouTube show hosted by Steve Eisman (ex-Neuberger Berman PM; the investor of "The Big Short"), now with a two-tier structure. Free: the weekly wrap on YouTube plus a free-subscriber email list (release alerts + sneak-peek video/newsletter previews) and free expert interviews (e.g. Dan Clifton / Strategas on the midterms, Jul 6; Torsten Slok / Apollo on AI & the K-shaped economy, Jul 13). Premium (paid, moving to Substack for all premium content): his personal portfolio (released Jul 1), the newsletter, and specialist deep dives (Jul 8 energy/oil/gold with Kathleen Kelly & Nancy of Queen Anne's Gate; Jul 15 biotech with Boris Peaker of Jones Trading). Still not a trades/alerts product — no model portfolio to follow or trade signals; the premium reveal is his own book and expert access, disclosed "as educational, not a recommendation." Added Jul 24: a masterclass on how to analyze banks and how to value them — "a deep dive on the entire sector" — is "releasing… in the near future," the first explicit education product; the cadence is now visible as free Monday interviews (Jul 20 Ben Kallo / Baird on sustainable energy + AI data centers; Jul 27 Dan Ives & Gil Luria on software winners/losers) alongside premium Wednesday deep dives (Jul 22 Ken Suchoski / Autonomous on agentic AI in payments; Jul 29 short-seller Brad Safalow on real-estate brokerage and for-profit education shorts), plus the Friday wrap with a reader mailbag that doubles as teaching (e.g. the funding-short explainer, 2026-JUL-24). Added Jul 31: the Monday slate is broadening beyond markets — Aug 3 is an author interview on Greenspan & Wooldridge's Capitalism in America for the 250th anniversary of the Declaration, and the Aug 5 premium guest is filmmaker/producer Peter Hoffman (Terminator 2, Basic Instinct) on old-school production vs streaming; the mailbag is also being used to teach economic history (the John Cassidy GDP-per-capita chart answering a viewer on the K-shaped economy, 2026-JUL-31). The free Monday interview remains the flagship — Ep 70's Ives/Luria episode (2026-JUL-27) put two sell-side analysts who cover the entire tech stack on one panel, which Eisman closed by telling viewers to watch it "at least once, maybe twice." Added Aug 10: Ep 72 (2026-AUG-10) shows the interview slate is banked, not live — it was recorded Thursday Jul 30 and released eleven days later, so every "today" in it is Jul 30 (worth knowing before trading off a number in one of these episodes). It also establishes the recurring-analyst panel as a distinct format: Glenn Schorr (Evercore) returns and Ken Worthington (JP Morgan) debuts — a former Oppenheimer colleague of Eisman's — with the two covering overlapping books so a disagreement can be staged rather than narrated ("we might do a round two of this podcast just on this topic"). Eisman's own role is the retail-facing one: he stops the analysts to make them spell out mechanics a viewer wouldn't know ("a lot of people don't understand that. Can you just explain the mechanics of that?"), then closes with a "lessons learned" segment stating where he still disagrees — "I think Glenn and Ken are more positive on the sector than I am." (Episode 63, with Bernstein semiconductor analyst Stacy Rasgon, is archived in this hub under stacy-rasgon.) Added Aug 14 (2026-AUG-14) — the product's commercial shape is now stated on air. Pricing: from Labor Day (Sep 7) premium costs $20/month or $200/year for new subscribers, while "all current subscribers monthly and annual are completely locked in at the existing price for as long as you stay subscribed" — a legacy-rate grandfather used explicitly as a subscribe-now inducement. What premium buys, in his own words: "all episodes, ad free, master classes, weekly Wednesday deep dives, and more access." The education product is now dated: the previously-teased banks masterclass ships after Labor Day as a two-part "How to Analyze Banks" — "a deep dive into everything you need to know and understand about bank accounting and bank functionality… with banks playing an opaque role in private credit and the massive AI buildout, understanding the quarterly earnings announcements is critical, and this master class will give you the tools to find the message hidden in the numbers." Cadence: the show takes a two-week break after the Aug 17 interview (Chris Verrone & Jason Trennert of Strategas) — the Wrap resumes Fri Sep 4, interviews Mon Sep 7 (Ryan Tunis, property & casualty insurance) — but "Wednesday premium episodes will continue without a break," i.e. the paid tier is now the continuity tier. The premium Wednesday slate also keeps widening past markets: Aug 12 Steven Cook of the CFR on Iran (whose conclusions feed straight into the free Wrap), Aug 19 Wolfgang Münchau on European stagnation, Aug 26 Becca Platsky of the Corporate Gossip podcast on corporate scandals and boards, Sep 2 Lakshmi Ganapathy "for some short picks." The free Wrap is visibly the shop window for all of it — three of this episode's segments are set up by, or trail, a premium interview. Added Aug 17 (2026-AUG-17) — Ep 73 is the last free Monday interview before the two-week break (the Wrap resumes Fri Sep 4, interviews Mon Sep 7 with Ryan Tunis on property & casualty insurance), so the free tier now goes dark for a fortnight while the premium Wednesday slate continues — the clearest demonstration yet that paid is the continuity tier. It also confirms the Ep 72 finding that the Monday interviews are banked, not live: this one was recorded "the beginning of August" and released Aug 17, so every "this week"/"last week" in it is roughly two weeks stale (state this before trading off a number from one of these episodes). Format-wise it extends the recurring-analyst panel a step further — two people from the same firm (Strategas' founder Jason Trennert and market strategist Chris Verrone) covering the whole market rather than a sector, which lets Eisman run a fundamental view and a technical view against each other on every topic in turn (rates, AI, banks, Europe, defensives) instead of interviewing a single specialist. His own role is unchanged: stop the guests to make them spell out a mechanic ("when you say the market is very concentrated, flesh that out just a little bit"), then close with "lessons learned" summarising where the guests' view differs from his own. The sponsor slate has also expanded — three mid-roll reads in one episode (Long Angle's high-net-worth asset-allocation report, Qwoka's business phone system, DripDrop hydration, the last two with Eisman-specific promo codes), up from the two-plus-one pattern of earlier free episodes; the ad load on the free tier is rising just as the paid tier is being priced up from Labor Day. Added Aug 28: asked on The David Lin Report where to follow him, Eisman states the product's shape himself, in one sentence — "we do three things: an interview that goes out on Monday on YouTube, a market wrap which goes out Friday after the close, and on Wednesday on Substack for premium subscribers, an additional interview or masterclass" — which confirms the cadence this hub had reconstructed from the episodes themselves and settles that Substack is the premium home. He also re-dates the education product: the "how to analyze banks" masterclass is "in a couple of weeks" (i.e. mid-September, after the two-week break), consistent with the Aug 14 announcement of a two-part premium drop after Labor Day. On the show's editorial identity he is explicit that it is not a stock-tip product: "very eclectic — we do a lot of sell-side analysts, we do authors, we do economists. You get a real education from our podcast," citing Wolfgang Münchau on why Europe doesn't grow, law professor Ben Zipursky on the legal theories behind the Meta algorithm suits, and Adrian Wooldridge on Capitalism in America as the range. Added Sep 11 (2026-SEP-11) — the masterclass finally has a date: part one of the two-part "how to analyze banks" masterclass drops on premium Wednesday Sep 16, part two two weeks later, covering "large cap, mid cap, and small cap banks and the investment banks as well." The paid tier's Premium Chat is now a visible input to the free show: "I write a note and chat in Substack almost every day, and I get many comments, which I learn a great deal from," and this Wrap quotes subscriber emails and chat answers at length (the Nvidia "Robin Hood" rebuttal to the premium Ed Zitron episode, two explanations of the Qualcomm warrants) — so the free Wrap now doubles as a showcase for the paid community, with Eisman stepping in to explain "what the commentator is getting at" and then giving his own caveat. The same episode says "subscriptions are free" on Substack via steveeisman.com (the free tier), carries a Granola sponsor read, and trails the Monday Sep 14 free interview with his former Big Short partners Vincent Daniel and Porter Collins, who "each gave a long that they like and own." Added Sep 21 (2026-SEP-21) — Ep 76 introduces a peer format: George Noble, "someone who has a background that's similar to mine, who has been an asset manager and is now a podcaster," so the conversation runs as two stock pickers comparing notes, down to a staged role reversal where Noble puts Eisman "in the hot seat" on the AI related-party financing. It again confirms the Monday interviews are banked: Noble dates a remark "on September 2nd, 2026," so the episode aired roughly three weeks after taping. Sponsor reads: Upwork and Webroot (with an Eisman promo URL).
Format as seen in the archived Ep 63 (2026-JUN-08): Eisman frames the theme, interviews the expert, then returns with his own closing thoughts.
Segment
What it is
How he runs it
Seen in the archive
Expert interviews
Long-form interviews with institutional specialists — the people who normally brief funds, not the public.
Ep 63: Bernstein's senior US semiconductor analyst on the AI semi boom and what could end it.
The stacy-rasgon hub (22 tickers from one episode)
Playbook themes
His recurring framework: AI is a commodity with no moats; tech is turning asset-light → asset-intensive; the token-pricing reckoning.
Laid out in episode framing/closing thoughts and repeated in outside appearances (the CNBC hit archived below).
SpaceX, GOOGL, META, MSFT, TSLA
His own positioning
Stated views, not trades — skeptical of the marquee AI IPOs but explicitly not shorting ("just not playing").
Opinion only; no recommendations attached.
SpaceX ("not a fan"), BRK.B
How it serves retail investors:
Free access to institutional expertise. Sell-side specialists like Rasgon spend their days talking to funds; the podcast puts that hour on YouTube for anyone.
A skeptic's framework for testing hype. The recurring playbook (commoditization, capital intensity, who actually pays) gives retail a checklist for AI-era valuations rather than a stock tip.
No upsell. Ad-supported; he isn't selling a portfolio, which also means the views come with no alert-service incentive to generate trades.
Transcripts
One dated page per appearance — each has its full stock table, talking points, and the saved transcript. Newest first.
Steve Eisman appearances discovered via YouTube search (Steve Eisman), not yet processed — verify publish dates & channels before processing. (None queued yet.) Scheduling (announced 2026-AUG-14): the Mon Aug 17 interview (Chris Verrone & Jason Trennert of Strategas) is processed, and the show is now in its two-week break — the Weekly Wrap resumes Fri Sep 4 and interviews resume Mon Sep 7 (Ryan Tunis, property & casualty insurance). Premium Wednesday episodes continue through the break (Aug 19 Wolfgang Münchau; Aug 26 Becca Platsky; Sep 2 Lakshmi Ganapathy), so expect no free YouTube uploads between Aug 17 and Sep 4. Updated 2026-AUG-28: the break applies only to his own channel — two third-party appearances landed inside it and are now processed: the New Money clip-commentary video (2026-AUG-15) and his interview on The David Lin Report (2026-AUG-28). Worth searching other channels, not just his, during any announced break.