Actionable insights — AI Terminator Fears Grow & Rates Breach 4.9%

14:30 1. Steelman the opposing thesis, then test it on sequencing, not on logic

1. When a strong counter-argument arrives, restate it in its best form first, and say plainly whether its logic holds: "That was a very interesting comment, and I take his viewpoint very seriously… this could be correct."
2. Separate the destination (where the industry ends up) from the path (what has to survive to get there). Most bull/bear disagreements that feel irreconcilable are about the path.
3. Identify the load-bearing parties of the current structure — the few whose failure would break it before the new structure is ready: "the current, and I do emphasize current, AI ecosystem is dependent, almost completely dependent, on the health of Anthropic and OpenAI."
4. Ask whether the replacements can absorb those parties' commitments in time: "one of them could fail before these new companies are capable of replacing the commitments."
5. Find the historical case where the destination was right and the path still crashed: "The first generation of dot-com companies failed miserably, and it was only the second generation companies like Google that went on to glory."
6. State your position as a dated conditional, so it can be checked: "if OpenAI fails within a year, there will be a massive correction."

• A load-bearing counterparty cutting prices, raising at a flat or lower valuation, or renegotiating commitments; the replacement cohort (neoclouds, open-weight models) visibly taking over contracted demand rather than just usage.

5:34 2. "What am I missing?" — when a capable actor does something that looks dumb, look for the other shoe

1. Size the stated action against the problem it claims to solve. If the ratio is absurd, the stated action is probably not the real plan: $4B, then $6B, of buybacks "is just not anywhere close to being able to do the job. After all, US federal debt stands at 40 trillion."
2. Check the market's verdict on the day, in the instrument's own units: the 10-year went 4.80% → 4.845% on the upsizing, "which in Treasury world is a big move" — in the wrong direction.
3. Assume competence before incompetence: "when someone who you think has a brain does something you think looks dumb, it's always best to ask, 'What am I missing?'"
4. Restate the actor's real constraint and list which parties could satisfy it. Rule out the obvious ones: "Bessent needs a large buyer of Treasuries, and it can't be the Fed or Treasury."
5. Carry the gap as an open watch item rather than forcing a guess: "I think there's some other shoe that could drop here, but I don't know quite what it is yet."

• Rule changes that create forced buyers of long Treasuries (bank capital or leverage-ratio exemptions, pension or insurer mandates, stablecoin reserve rules); the buyback being upsized again; the term premium falling without the Fed acting.

9:12 3. Back out the belief a multiple requires — then check it against the latest evidence

1. Compute the forward P/E from consensus, not trailing numbers: 2026 EPS $1.66 → "the 2026 PE is 220 times."
2. Put it next to the closest business peer's multiple to show how much is not the business: "GM's 2026 PE is 6.5 times."
3. Name the single belief that closes the gap. If only one explains it, the stock is a one-question bet: "The only explanation for Tesla's crazy multiple is that some investors believe that Tesla's robotaxi business will conquer the world."
4. Judge each event about that one question against management's record of promises versus delivery: "a history of over-promising and under-delivering. He seems to have done so again."
5. Declare your exposure so the analysis is not mistaken for a position: "I have no skin in this game. Just count me a skeptic."

• Consensus EPS revisions against the multiple; dated robotaxi milestones (cities, fleet size, driverless miles) being hit or slipping; the stock's reaction to each milestone as a read on how much of the belief is already priced.

9:47 4. Right on the fundamentals is not the same as a profitable short — measure the gap first

1. Track earnings from peak to the current estimate: EPS $4.07 in 2022 → $1.66 for 2026, "59% lower than the peak."
2. Compare the price path over the same window: "the stock has done nothing for 5 years despite the massive decline in earnings."
3. If fundamentals collapse and price does not, the holders are paying for a story, not for earnings. A falling E has simply pushed the multiple higher.
4. Treat that as a warning about the short, not a reason to add: "the people who were short were right on the fundamentals," yet "this shows once again how difficult it can be to short a cult stock."
5. Only short once the story itself is failing on evidence, not when the earnings are.

• Evidence against the story (missed milestones, a competitor's deployment at scale) rather than another weak earnings print; how the holder base behaves after a bad event, e.g. a stock that shrugs off bad news.

17:36 5. Read the backlog, not the beat — growth quarter-on-quarter and who owes it

1. Note the headline growth, then take out the adjustments you don't accept. Here non-GAAP drops stock comp: "a common practice in the tech world, which I just think is wrong."
2. Go to the forward indicator. For a contract business that is remaining performance obligations (RPO), "a form of backlog."
3. Measure its sequential change, not the year-over-year level: $664B vs $600B "is not that much greater than the 600 billion of the prior quarter."
4. Find the share owed by the largest counterparty: "approximately 50% of its RPO coming from OpenAI."
5. Judge that counterparty's ability to pay separately from the vendor: "a company that we all know has massive negative cash flow."
6. Read the price reaction against how far the stock is off its high: +4% after hours "is okay, but the stock is way off its highs" because the market already discounts the concentration.

• RPO growth stalling or reversing; the top-customer share of RPO moving; contract terms disclosed as cancellable or "flexible"; the counterparty's own funding rounds and pricing behaviour (this week's OpenAI price cut).

16:24 6. When a company gives away equity it doesn't need to, ask what it is buying

1. Split the deal into its parts and react to each separately: "Not by the partnership, by the warrants."
2. Value the non-cash consideration at today's price: 25M shares struck at $161.26 with the stock over $175 → "in the money and… worth over $4 billion."
3. Rule out the standard reasons. Is it circular financing? Is the issuer raising capital? "No cash changes hands until later and Qualcomm does not need any cash anyway."
4. Then list what the equity could be paying for: customer acquisition ("trades equity upside to guarantee a 60 billion revenue pipeline and secure AWS as a flagship anchor customer") or a missing capability ("their expertise in XPUs is not established. Maybe Amazon's contribution of IP will be important").
5. Either answer is a disclosure: the issuer is paying to enter a market where it cannot yet compete on its own, and shareholders fund it through dilution.
6. Crowdsource the question when you can't resolve it yourself, and keep the answers as hypotheses, not conclusions.

• Warrant vesting tied to purchase volumes (confirms customer acquisition); disclosed IP-licensing terms (confirms capability purchase); other chip vendors copying the structure with other hyperscalers, which would show who holds the bargaining power.

19:35 7. Read the fastest-growing division inside a dull total, and name why it is growing

1. Start from neglect. A turnaround is easiest to spot where "no one cares about this retailer anymore at all."
2. Take the like-for-like metric (same-store sales) at the total level and grade it honestly: +2.7% "is not bad, but not great."
3. Split it by banner or segment and find the outlier: "Bloomingdale's, saw a same-store sales growth of 11.3%."
4. Name the cause and check whether it lasts. Share taken from a failing competitor lasts longer than a one-off: "Bloomingdale's benefits from the demise of Saks."
5. Weigh the trend against guidance-driven price moves: the stock fell on next-quarter EPS guidance, "but still, the strength in same-store sales comps was a pretty good sign."

• The outlier banner's comps holding once the competitor's store base is gone or sold; the weak banners stabilising; management moving capital toward the growing segment.

21:34 8. Price every asset against the risk-free yield, and ask what is driving the yield

1. Set the hurdle explicitly: "Why buy stocks when you can get 5% risk-free?" That is "partially why higher rates are generally negative for the stock market."
2. For an asset with no cash flow (gold), drop the yield comparison and find the belief behind demand: "a hedge against inflation and against the debasement of fiat currency."
3. Diagnose why the risk-free yield is rising. If it is inflation fear, the same force can lift the non-yielding hedge: "The 10-year yield has marched higher because the war in Iran has driven oil prices higher and that creates fears of inflation. And those inflation fears drive some investors to buy gold."
4. Make your own judgment on the underlying belief, not the price: "I think our government is stretched but not yet overstretched."
5. Choose on that judgment and say it plainly: "I don't own gold… 5% seems like the better choice to me."

• Whether yields rise on inflation (gold-friendly) or on real growth and term premium (gold-hostile); signs that "stretched" is becoming "overstretched": failed auctions, buyback upsizing, fiscal dominance of Fed decisions.