Steve Eisman — Stock Picking in Difficult Times with George Noble

0:04 A fellow asset manager turned podcaster

• Noble ran money at Fidelity and now podcasts and writes a Substack newsletter — "kind of like me." A return guest.

1:20 "Weeks go by and years happen" — when macro matters

• Noble opens with Lenin's line: we are in one of the weeks where years happen (oil, AI, the deficit). Stock pickers can ignore macro when it's quiet; "when you have a gale force hurricane in your face… it's hopeless."
• His framework, credited to "Michael Cananter" (probably Michael Kantrowitz): most stocks' behaviour is a mix of factor influences — rates up or down, the dollar up or down, the economy slowing or accelerating.

2:57 Bessent has no good move

• Asked what advice he'd give the Treasury Secretary: "resign." Eisman: like chess, "in a position where you have no move." The accumulated deficits are bipartisan.

4:00 AI debt is crowding out Treasuries

• Eisman: "the explosion in debt issuance from AI is… crowding out treasuries"; without the AI story, "rates would probably be 40 basis points lower. I just made that number up." Noble's email phrase: "capex boom meets runaway deficits."

5:03 Stop staring at monthly CPI — the regime changed

• Noble likes Truflation, but mocks the "macro yentas" fixated on a CPI of 3.8 vs 3.6: "they're totally missing the plot."
• Post-GFC, printing "didn't really matter," almost enough to make MMT look legit. Now, per Louis-Vincent Gave, policy acts differently with "binding constraints" — stimulus meets bottlenecks and becomes "real inflation."
• Credited to Jim Bianco: the market overestimated inflation until ~2021 and "since 2022… has been constantly underestimating inflation because we have fiscal dominance." Noble's Yale senior paper found forward markets predict rates badly — "no better than the weatherman."

8:03 A whale with a BB gun

• Eisman: upward pressure on rates is "mostly because of AI, not so much the deficit"; against $40T of debt, Treasury buying "is like trying to kill a whale with a BB gun." If rates hit 5% in weeks, "his credibility is shot."
• Noble: it's already happening — rates dipped "for a millisecond" and are "meaningfully higher now."

9:26 "Not the Scott Bessent I once knew"

• A former institutional salesperson who covered Bessent's hedge fund is "shaking his head": at first carrying water for Trump, now he seems to believe it — "the smugness, the arrogance, and… the incompetence." In a pickup game of Bessent vs Druckenmiller, "I know who I'm going with."
• Noble's charge: the statements on oil through the Strait and gas prices are "one exaggeration… after another."

11:28 Bessent as Norman Lamont

• The role reversal: Bessent worked for Soros when Soros and Druckenmiller broke the Bank of England; now he is the official defending "an artificial rate" — and "when there's a wrong price, markets have a way of attacking." (Noble: as a 28-year-old junior, Bessent "had nothing to do with" that trade.)
• Without firepower it's "open mouth operations," and intervention "only really works when the fundamentals are with you."

13:22 AI: "show me the ROI"

• "It's all being funded with funny money": Nvidia's $250B package for OpenAI "tells me all I need to know." The AI boom is ".com and subprime all merged in one."
• Models leapfrog weekly (50% faster, 50% cheaper; Chinese models "95% less"); people "in the weeds" miss that "AI is set up to fail."

15:06 24× the dot-com malinvestment

• Julien Garran (MacroStrategy Partnership) puts AI malinvestment at "24 times what we saw during .com." Noble sees a narrow application ("maybe the market's worth 50 or hundred billion dollars"), with the Chinese model "good enough for 95% of us at 95% less cost."
• Peter Berezin (BCA): the 1999–2000 internet bulls were right — traffic compounded 43% a year for 25 years — and it "didn't stop Global Crossing and all the rest of them from going bankrupt."

16:26 Eisman's chain: two labs hold it up

• Nvidia selling chips and hyperscalers buying them is fine; "where it starts to get squishy" is that ~70% of hyperscaler AI revenue is OpenAI + Anthropic, 25–35% of total cloud revenue.
• OpenAI "is in trouble"; Anthropic needs its S-1 out "fast because token maxing is ending," to list on peak second-quarter numbers — but it is "in better shape than open AI."
• (18:01–20:57: Upwork and Webroot sponsor reads.)

21:00 Role reversal: the homebuilder that lends to its buyers

• Noble puts Eisman in the hot seat: a builder lending to buyers at 0–1% when mortgages cost 7%, not accruing the difference, and booking it all as income — "this car crash" is the AI related-party picture.
• Eisman: "the circular financing kind of drives me crazy," but if the labs became "insanely profitable, it wouldn't matter"; it "just tells me that there's probably something wrong." You wouldn't "blindly say oh look at the earnings, look at the PE."

23:32 Nvidia: 70% of receivables from five accounts

• Eisman: why has almost no one but "a few substackers" commented? Revenue +100%, yet five accounts are 70% of receivables — "very frightening."
• Noble: if lenders thought these were good credits "they'd make loans, but they're not doing it… They're nervous."

24:22 Dollar-yen explained: the widowmaker and Mrs. Watanabe

• Eisman's "mental block." Noble: shorting JGBs was "the widowmaker trade" for years (it has finally worked); the yen "continues to go from weakness to weakness."
• The carry trade: a Japanese life insurer or "Mrs. Watanabe" earning ~0.5% at home buys US assets (e.g. AvalonBay) for the yield — and a rising dollar doubles the gain. The yen went from 80 to 160.

26:56 Robin Brooks' chart: JGBs should yield ~7%

• Brooks plots debt/GDP against borrowing costs: Switzerland and Sweden low, Europe mid, the US higher, Japan highest. Japan's 10-year "just crossed 4% for the first time in 30 years. It really should be around 7%." (US debt/GDP ~125–130%, 10-year 4.8%; Japan ~240%, though ~130% net of domestic holdings.)
• Forward markets aren't infallible, but "the market right now is saying 10 years forward the 10-year US rate is at 6%" — a comment on "the sustainability of the path that we're on," and "the Japanese are going to get there first."

29:20 The money that built the Nasdaq could go home

• US bonds and the Nasdaq "benefited enormously from an influx of money from abroad" — Japan, and European pensions with outsized US-tech allocations.
• The 10-year JGB–Treasury gap has narrowed to ~180bp; "at some point… could be tomorrow, could be 5 years from now," Japan says "bring the money home." It is caught between a falling yen (rice prices, inflation) and higher rates on its own debt — and as the world's largest creditor, its choice matters.

31:29 The swap line: "total insanity"

• Japan spent ~$95B intervening in one week. To buy yen it would sell dollar assets — Treasuries ("nice bond market you got there…"). Instead the US "set up a swap line": Bessent lent Japan the dollars. Eisman: "Now I understand."
• Japan has been selling Treasuries for a while, China for years; asset seizures after 2022, tax and trade policy put foreigners off "at a time when we need their funding more than ever." So "who's going to buy the bonds?"

33:45 Investors want stories, not analysis

• David Einhorn says his model no longer works; Jeff Vinik found people "just want to know the narrative." Eisman: Game of Thrones' finale — what motivates people? "Stories." Guy Adami: people "want to be told a story that'll make them feel good."

35:06 "Narrative dominance"

• Noble's coinage (dated Sep 2): the stock-market counterpart to fiscal dominance — "Don't tell me about the fundamentals… Tell me a story." Add the passive bid for index heavyweights and "there's no price discovery," which "creates threats and opportunities."
• "More money than fools" makes a bull market. Example: a BNPL lender with too-low reserves at six times book — "Oh, but you don't understand. This is SoFi."

37:11 Southwest, and the one-page report

• Noble's Substack is "11 for 12"; the miss is Southwest (cheap, Elliott, bag fees, no middle seat), derailed by Hormuz and oil — "down a few %."
• He revived Fidelity's six-to-eight-page "investment community reports" (SWOT, statements, catalysts, insider buying) at three or four pages; subscribers asked for one: "Give me a story and a page."

39:03 When capital costs something, stock pickers matter again

• At a zero cost of capital "pigs fly"; once it matters, lenders will say "show us the economics of this data center."

39:43 SpaceX: shorts at trillion-dollar scale, and the float

• "I've never seen such shorts at scale" — $1.7T and $1.4T caps. SpaceX is "built to fail": beyond 90× revenue, the float is rising from ~5% to 20–25%, with 7% more every few weeks. Endowments and family offices that bought at $200–800B valuations "have to sell."
• A "Cathie Wood type of story": predictions too far out to disprove (asteroid mining). Data centers in space are "ridiculous" — Eisman: in the sun "it's really really hot."

42:10 The public short, and "a third tier AI company"

• Noble's May report (~2M views on X) came before the IPO; "check back with me at year end or the end of next year… others speculate, we invest."
• The S-1's $29T TAM is $22T AI: "a weak AI company… with a rocket thing attached." Starlink $200–300B; the AI arm worth "a fraction" of the $200–250B paid in stock; total "300… 500 billion" vs $1.7T.

44:48 Tesla: sales imploding, capex exploding

• US sales −26% in July (Gordon Johnson's data), Europe and China down; Q3 laps last year's pre-buying. Capex from ~$6–8B to $30B for AI; "seriously cash flow negative"; ROE never above the cost of capital.
• Robotaxi in geofenced Austin: rides per million people extrapolated nationally value it at "800 million not 800 billion." Optimus "a complete freaking joke."
• SOTP for both Tesla and SpaceX: $30–60 a share, vs ~$360 and ~$140.

47:49 Japan in the '80s, and five flat years

• Both were "jam[med]… into the S&P" — like Japan's momentum market (60× earnings, banks at 100× earnings); once "the worm turned," avoiding those and indexing the rest won.
• Tesla's price is "unchanged over 5 years… a function of liquidity and narrative. And I think the clock's about to strike 12."

49:09 Eisman's wrap

• Bessent's rate repression is "a complete waste of time" against $40T; rising Japanese rates might make Japanese institutions and retail "sell dollar assets"; Noble "may be more negative on AI than I am"; he is short SpaceX (worth "200 to 300 billion" vs $1.7T) and Tesla, where the falling auto sales were "the most interesting data point."