Actionable insights — The Greatest Stock & Earnings Bubble In US History

8:10 1. Strip the capex cycle out of "earnings growth" — find the real number

1. When one spending theme dominates, don't trust headline index earnings growth. Identify the two distortions: (a) one-time mark-to-market gains on equity stakes, and (b) revenue that suppliers book the instant capex is ordered.
2. Recompute: take the reported growth figure, remove the one-time markup gains, then remove the gains from the component suppliers feeding the theme. What's left is organic growth.
3. A tool trick he used: hand the quarterly numbers to an LLM (Gemini) and ask it to strip the named buckets — fast way to pressure-test a bull's headline stat.

• Any market where a handful of capex-supplier stocks drive index EPS; the wider the reported-vs-organic gap, the more fragile the multiple.

6:24 2. Follow the depreciation lag — the buyer's bill arrives later

1. Note the accounting asymmetry: the supplier recognizes revenue on order; the buyer capitalizes the same spend and expenses it slowly (5-6 years for GPUs/DRAM). Early in a buildout, supplier earnings are inflated and buyer costs are hidden.
2. Project the buyer's depreciation curve forward. If the expense line roughly doubles over a few years while the promised revenue is unproven, real earnings sink.
3. Re-rate the "cheap" multiple: compute the PE adjusted for non-success — headline PE assuming the buildout disappoints — instead of the reported PE.

• Rising depreciation as a share of revenue at the buyers; the first hyperscaler whose margin guide-down admits the lag has caught up.

11:30 3. Interrogate the unit economics — cost trajectory vs return

1. Ignore the vision; ask the operator's question: what is the cost of the core input doing, and what is the return on it?
2. If input cost is compounding (doubling every 45 days) while return is flat-to-marginal (~5%), the activity is value-destructive no matter how strategic it sounds — and even insiders start to pull back.
3. Confirm with revealed behavior: watch for spending caps, license cuts, and budget halving — actions speak louder than capex guidance.

• Per-unit cost curves in any hot theme; enterprise budget cuts and usage caps as the leading indicator of a spending peak.

14:30 4. Track the pricing wedge — cheaper substitutes take share

1. For any premium-priced product, watch for a "good-enough" substitute at a fraction of the cost (90% of the job at ~1.5% of the price). That's the wedge that breaks pricing.
2. Quantify adoption with a usage/market-share series, not marketing claims (here: Open Router's token-share data — Chinese models 11% avg → 46%).
3. Check the escape routes are closed: even a ban wouldn't help if domestic players copy the low-cost technique — so the price war is structural, not geopolitical.

• Independent usage/share data on low-cost substitutes; incumbent price cuts and subsidies as the confirmation that pricing power is gone.

29:30 5. The exhaustion checklist — supply of paper and margin debt

1. You can't time the top, so track exhaustion signs instead. First: the buyback→issuance flip — when the biggest share-shrinkers stop buying back and start selling huge new equity/IPOs, demand is being exhausted.
2. Watch the bond market for saturation: a single mega-issue moving the 10-year yield, and a US borrower forced into euros/yen/francs/pounds, means the domestic bid is full.
3. Read margin debt as the speculation gauge: level, year-over-year rate, and % of GDP vs prior tops — a "straight up" line unlike prior spikes is the warning. Add new-speculation tells (levered ETFs, zero-DTE options) and offshore froth.

• The issuance calendar overwhelming buybacks; mega-bonds moving benchmark yields; margin debt / GDP breaking prior-cycle records.

35:00 6. Sell the froth early, hold cash, wait for the fat pitch

1. When a market you own gets "very hot, very speculative," scale out — sell down "as far as you can and still sleep at night," even if it means missing the last leg up.
2. Sit in T-bills with no pressure to perform monthly; missing the top is cheaper than being trapped in the unwind.
3. Pre-commit the buy trigger to the washout, not a forecast: wait for the collapse to begin, then deploy into (a) the survivors and (b) the cheap, hated sectors capital rotates toward. "Keep the bat on your shoulder and wait for the fat pitch."

• Your own position sizing when a theme goes vertical; the start of the drawdown (not a target price) as the signal to begin buying.

58:09 7. Capitulation gauges — is the hated asset washed out yet?

1. Stack sentiment/positioning extremes rather than trusting price: a bullish-percent index collapsing toward zero, futures open interest at a multi-year low, and fund flows in persistent outflow.
2. Check that the froth actually left: was there ever a retail mania to unwind? If not (unlike the prior cycle top), less overhang remains.
3. Identify who is still buying for structural reasons — the durable bid that puts a floor under the price while speculators are gone.
4. Respect the whoosh-down risk: even a washed-out asset can drop in a broad liquidity scramble, so nibble first, size up once the broad selloff is underway.

• Bullish-percent indices near zero, futures OI multi-year lows, ETF outflow streaks; a structural buyer (central banks) holding the floor.

1:09:54 8. Price what the Fed can't do — the interest-expense trap

1. When a hawkish narrative ("next Volcker") drives rate-hike pricing, check the government's carry cost first: interest on the debt (~$1.35T, record) plus ~$2T annual deficits, at an average rate (3.3-3.4%) still below market rates.
2. If hiking blows out the deficit and the economy is weak, treat aggressive hikes as politically/fiscally untenable — the more likely lever is balance-sheet runoff, not higher rates.
3. Combine with the 1970s template: sticky ~4% inflation + a constrained Fed = negative real rates, which is the environment where gold's usual oil-correlation breaks and it works as the inflation hedge.

• Interest-expense-to-deficit math whenever hikes are priced; real-rate direction as the gold driver, not the oil correlation.