Skip to main content

Get Ready For A.I.-Mageddon | Fred Hickey

Adam Taggart | Thoughtful Money® Jul 14, 2026

Summary

Fred Hickey expects the AI equity bubble to break, and argues the money leaving tech will rotate into undervalued commodity producers.

Gold holds near $4,000 because central banks keep buying; China has cut its Treasury holdings roughly in half to fund gold purchases.

Hickey reads gold capitulation as largely complete: futures open interest sits near a 13-year low and ETF outflows are stabilizing.

Gold miner stocks now average about 10 times earnings, their cheapest valuation in more than four decades, per Scotiabank.

Hedge funds hold almost no energy positions, and thin new supply leaves commodity producers undemanding against richly valued tech.

Adam Taggart's takeaways

  • Top takeaway Taggart frames the concentration risk he puts to Hickey: AI-linked stocks are a record 45% of S&P 500 market capitalization and nearly 70% of the NASDAQ 100, and have driven nearly all of the index's gains in 2026 — a single-sector dependency he calls a major vulnerability.
  • He argues AI profits face a squeeze from both ends at once — falling revenue as buyers switch to cheaper models ("just like when steak becomes too expensive, you switch to chicken") and rising depreciation costs that have not yet fully hit reported profits.
  • On the real-economy read-through, he notes the rest of the economy is weak too: David Sacks's split of roughly 2% GDP growth has AI contributing about 3.1 percentage points and everything else 0.1%.
  • He concedes commodity producers carry far less froth than the AI complex on adjusted valuations, but counters the commodity bull case by pointing out that the physics of the buildout — land, permits, water, copper, capacitors and memory — may cap the pace well below analyst forecasts.
  • On gold he argues the froth has largely cleared since early 2026, but that US-Iran kinetic episodes from late February pushed oil up and had some sovereign gold buyers selling reserves to pay for it, which he thinks is what has pinned gold near $4,000.

Kevin Warsh's takeaways

  • Top takeaway Hickey argues the Fed under Kevin Warsh is unlikely to hike rates despite a Volcker-style posture, because federal interest expense is at a record $1.35 trillion on a roughly $2 trillion annual deficit that keeps growing, so hikes would widen the deficit further.
  • He says Warsh likely made some kind of understanding with Trump to avoid rate moves and would instead cut back on the $6.7 trillion balance sheet.

Fred Hickey's takeaways

  • Top takeaway Hickey argues the economic model behind AI is "disintegrating": hyperscalers have committed about $750 billion of data-center capex this year, up roughly 75% on last year, on FOMO and hoped-for future productivity rather than measured returns, leaving massive overcapacity and malinvestment while only about 5% of planned capacity is even switched on.
  • He calls this an earnings bubble on top of the stock bubble: the Buffett indicator is 241% of GDP versus 160% in 2000 and price-to-sales is 3.7x versus 2.3x, and once he strips one-time mark-to-market gains (40% of Q1 hyperscaler earnings growth) and the supplier gains at Nvidia, Micron, Seagate and Western Digital, S&P 500 earnings growth falls from roughly 28% to single digits; an adjusted CAPE near 67 would be about 50% above the 2000 record.
  • Chinese and open-source models are breaking the pricing model: Chinese models now take about 46% of token usage on OpenRouter versus an 11% trailing-twelve-month average, Zhipu AI's latest model exceeds Gemini 3.5 and 3.1 on his chart, and DeepSeek reportedly does about 90% of tasks at 1.5% of the cost; Meta just cut prices 75% and Microsoft, Amazon and Coinbase are trimming spend or moving to cheaper models.
  • He says hyperscaler cash flows have collapsed — only Alphabet and Microsoft are still cash-flow positive while consensus models them soaring through 2028-2030 — and Google's depreciation expense is set to roughly double from 17% to 35% by 2028, so the 23x-27x PEs on the "Lag 7" are far higher on failure-case earnings and about half of hyperscaler backlog sits with OpenAI and Anthropic, which he thinks carry real bankruptcy risk.
  • He will not date the correction but points to exhaustion signals as of July 2026: margin debt above $1.4 trillion (up 55% year over year, 4.5% of GDP versus 3% in 2000), buybacks replaced by heavy equity and debt issuance, Amazon's $25 billion bond adding 8 basis points to the 10-year, $300 billion of data-center projects cancelled, a data-center moratorium in his own city of Reno, and a record $1.35 trillion in federal interest expense; he expects a 2000-style collapse that drags the hyperscalers down with it, then a rotation into energy and commodities, and is sitting in more cash than he can remember, largely in Treasury bills.

Jeremy Grantham's takeaways

  • Top takeaway Hickey says he is in complete agreement with Jeremy Grantham that this is the greatest stock bubble in US history, and cites Grantham's expectation of a crash with stocks declining 70% or more.
  • He stops short of endorsing the 70% figure — "I'm not sure that will happen" — while still expecting at least a severe correction that pulls the hyperscalers down with the rest of the market.