Edward Dowd — The AI Crash is Coming (And It's Worse than 2008)

0:00 The call: a 40–50% drawdown, then QE — but only after the damage

• The dot-com bust took the S&P 500 down 50% over two years; the GFC did 50% in ~18 months with most of the damage inside nine. "I suspect we'll have a 40 to 50% drawdown once it's all said and done."
• The Fed then does what it always does — massive QE and money printing, and another attempt at re-inflation. But the fiscal/monetary response comes after the damage, not before it.

1:13 Three pillars of risk — plus the Iran war on top

• Unchanged from the Phinance US economic report published at the start of 2026: the AI bubble, the US housing market, and China's acute economic crisis.
• The Iran war that began 28 February added a fourth element — an energy price shock — which "is only going to pull forward all our risk outcomes that we predicted."

2:05 Signpost 1 — OpenAI floats government financing, then loses insiders

• He watches what the players do. Sam Altman and CFO Sarah Friar floating the idea of government financing was "salvo number one" — you only ask for that when "the engines of financing were starting to crack."
• Private credit, "a big source of funds for the AI buildout, started to freeze… it's the end of the credit cycle." Then OpenAI insiders began leaving — a tell that things aren't going well internally.

2:56 Signpost 2 — Apollo: AI revenue is coming from investors, not customers

• "Notable dignitaries in the finance world" are now questioning it: the head of Apollo said most AI-complex revenue appears to come from investors rather than end customers.
• Classic capex build-out: investors fund the infrastructure, early "parabolic" user growth was largely tyre-kicking, token pricing is falling, and spreads widen a little more with every new AI bond deal.

4:02 Signpost 3 — the semiconductor complex peaked 26 June

• Semis peaked on tremendous orders and good fundamentals; they sold off, and the current attempted rally "seems to be rolling over." The market is only at "the beginning phases" of realising it.
• Two hard stops on the capex cycle: credit markets getting too expensive, and simply not enough power to plug the hyped data centres into the grid. When capex slows, the pick-and-shovel beneficiaries (chips, power generators) roll over — "then it becomes a reflexive problem."

5:24 Double ordering — why the backlog is fake

• An X survey of AI-infrastructure buyers found ~50% admitted to double ordering. The mechanic: you need a million chips, you're allocated 500k, so next round you ask for two million to get your one million.
• "The backlog looks very robust, but a lot of it's double ordering. And that's how these cycles end." It's game theory, and it happens in every capex build-out.

6:53 Shadow inventory, chips as collateral, and MOUs that aren't contracts

• Rumours of Nvidia chips sitting in warehouses waiting for data centres that don't exist yet — "shadow inventory of these chips all over the place" — and chips "being used as collateral in some of these deals, which is insanity to me."
• Nvidia's announced $500B financing from Blackstone and the banks was a memorandum of understanding, not a contract: "it can just vaporize overnight." Same with Stargate's $500B, announced at the start of the Trump administration and still unfunded.

7:51 What surprised him — the war hijacked the fundamentals

• Coming into the year the market was starting to discount economic trouble; the 10-year yield went below 4%. Then the Iran war spiked yields and oil, attention moved to war headlines, Trump announced an MOU, and the market ripped.
• "It was a very narrow rally that took us to new highs. It was all AI or AI adjacent" — and it peaked with the semiconductor index in June. He is not calling a top, but "the bubble may have already burst, but people won't realize that until a couple months from now."

9:20 The real economy is already weak — Walmart's worst comps in six years

• Without the AI build-out "the GDP would be a lot weaker. We may even already be in a recession" — the average consumer is struggling mightily.
• Walmart's same-store sales were the weakest in six years and management said the consumer is struggling. "Joe Sixpack, 80% of the country is not doing well" — and 10–20% of consumers can't carry the rest of the economy.

10:07 Dot-com valuations → 0% projected for a decade

• A million dollars into SPY today is "projected to earn 0% over 10 years, which implies a huge drawdown." The question is only when.
• With an index 40–50% AI and AI-adjacent, "that concentration risk always ends badly. Always."

11:17 It's a second-derivative question — watch Korea

• The build-out doesn't have to stop, only slow. "Semiconductors are the classic canary in the coal mine… when the second derivative slows, the valuations of those stocks goes down quite a bit."
• South Korea's index — ~50% two stocks, SK Hynix and Samsung — peaked in June and is down 30–35%. A new low there says the second derivative has shifted. "I was a growth stock investor at BlackRock… second derivative does not bode well for valuations."

13:42 Housing — a frozen market with the record for-sale/sold gap

• Price declines started in the Southeast and near the southern border; blue cities still hold up but their second derivative is shifting. "We have the largest amount of homes for sale versus home sold gap on record, meaning the market is frozen."
• It's a buyer's strike against homes ~30% overvalued: 60% of listings are older boomers selling to millennials trying to start families. "The only way you clear that is through price."
• The rollover began in 2022 when new permits peaked. By the time a housing bubble makes headlines, most of the damage is already done — and housing is 20% of the economy, with big back-end effects (construction, finance, furniture).

15:38 Demographics — stop treating a house as an investment

• Once prices adjust there's a recovery, but no take-off: the demographic tailwind is gone. "You shouldn't look at a home as an investment for a long long time. It's a place to live."
• "The boomers were the python in the snake" — they supported housing for 30–40 years and are now aging out. Similar to Japan, but not as bad; Japan's own demographic decline is re-accelerating.

17:46 China — the overbuild math nobody did, and −9.5% fixed investment

• Post-GFC China built out infrastructure far beyond what its population could absorb, even though the demographic plateau in 2015 was arithmetic, not "rocket science." Decline began in 2020; ~150 million people are aging out of prime working age.
• The property crisis started in 2021 (Evergrande); new permits are down ~70% and 20–30 years of housing inventory sits unsold. Construction only rolled over recently because long-life projects were completing.
• Net fixed investment went negative year-over-year at the end of last year, bounced on stimulus in Q1, and is at new lows — the most recent month at −9.5% YoY. Internal consumption is "falling off a cliff."

20:05 Export dumping, trade wars, and the CCP's real fear

• China's answer to the consumption and property bust was to export more — which is what the trade wars are about. In a global recession those exports collapse, making it worse: a feedback loop.
• "There's only about 7 million card-carrying members of the CCP… 7 million people controlling billions. They fear their own people" — hence the permanent need to keep people employed. Robots are partly military, partly labour replacement, but "the technology is not ready for prime time": a robot's energy and maintenance cost far exceeds a human's. "Humans are still the cheapest way to get something done."

22:00 No safety net → forced savings → no Plaza Accord 2.0

• Chinese households save heavily even in prime earning years because there is no social safety net. Building one means the top sharing power and money — "there doesn't seem to be an appetite to do that."
• Japan escaped its demographic wall by exporting more, but Japan is a tenth of China's size — "the world cannot accommodate China trying to do the Japan model." Expect more trade war and a hunt for a scapegoat: "that's why there's all this chatter to Taiwan." Buffett exited Taiwan Semiconductor partly on that war risk — early, but "when that conflict does start, it'll be a problem."

24:57 Yields — "the solution to high yields is high yields"

• Exactly like 2007–08: yields rose on temporary oil-shock inflation (oil doubled to ~$149 on the China-demand story) until demand destruction hit and everything rolled over as growth slowed.
• Today sovereign issuance and the AI bubble compete for the same capital, so yields rise to attract it — then choke the economy, growth slows, yields fall and the Fed cuts. "The solution to high commodity prices is always high commodity prices because more supply comes online."
• The bond market trades on two things, growth and inflation expectations: "if I'm right on the recession and the AI bubble bursting, yields will come down in the long run."

26:47 The sovereign-debt release valve is the currency — starting in Japan

• His long-held thesis: the global sovereign debt problem manifests in currency markets, and "it's beginning in Japan." The BoJ is case study number one — a demographic disaster with debt/GDP around 270%.
• Japan faces a "Hobbesian choice" between funding the yen carry trade and caring about its own economy; the US wants it not to hike too fast and unwind the carry, which would cause a deleveraging event. China's collapsing bond yields (deflation/disinflation) eventually catch up with the rest of the world.

27:58 Stock yields are below the risk-free rate — the allocation switch

• Normally the S&P 500's dividend yield is higher than bond yields; that relationship is inverted, "and that usually doesn't last long in history." It is exactly why he can say valuations imply 0% over 10 years.
• As the growth scare arrives, it becomes "cheaper on a cash flow basis to own bonds," and portfolios over-allocated to stocks and under-allocated to bonds reverse. "Once the flows begin, it happens quick."

29:21 The long end — "one of the greatest asymmetric trades"

• Commodities (copper, gold, silver, energy) have run, but in a recession demand-led names go down — and "most people think bonds are dead," which is what makes the long end "a very contrarian call."
• Two levels of expression: "if you're Joe saver, you just want to raise cash… so you can buy cheap assets on the other side of this"; institutions willing to take risk can own the long end. "It's gone against us recently, but once this begins, it'll happen fast."

30:06 The takeaway — raise cash, don't short

• "Valuations are stretched. The bubble is in the process of coming undone, and cash allocations in your portfolio, taking advantage of lower prices, is warranted."
• Do what Buffett is doing — 40% cash, waiting for a fat pitch. David Tepper announced 40% cash in June; Paul Tudor Jones said he wouldn't recommend investing in the S&P 500 at these valuations.
• "Don't try to short the market, that's a waste of time, because timing is hard on these things." His new Substack is Adapt on the Narrative; the paid research is at phinancetechnologies.com.