AI chip stocks are unwinding, leveraged ETFs are hammered, and SpaceX shed 30% post-IPO. The market is asking a question the bulls haven’t answered yet: who actually captures the value from the AI capex buildout?
Bloomberg reported this week that a retail-fueled unwind in AI chip stocks has wiped billions from leveraged ETFs — the instruments that amplified both the run-up and the correction. Semiconductor shares fell across Asia and the US as investors began stress-testing a core assumption: that AI capex translates into equity value at the pace the market priced in.
The SpaceX IPO was supposed to be a signal moment. SPCX launched at $225 and has since settled around $153 — a 30% haircut that dented the “everything AI-adjacent is a buy” thesis that carried markets through 2025. Funds with SpaceX exposure got bruised in the process.
Meanwhile, OpenAI delayed its IPO to 2027 amid a $3.7 billion quarterly burn rate. The revenuestory is real — AI revenue hit $25 billion in Q1 2026 (ex-China), exceeding depreciation for the first time in the sector’s history. But the marginstory is not. Thin margins at scale are not the same as the compounding returns the market was pricing in at peak multiples.
HOW THE ROUT UNFOLDED
2025 Peak
AI chip stocks reach peak multiples. Leveraged ETFs (NVDL, SOXL) balloon to $270B in retail exposure. SpaceX pre-IPO buzz fuels speculative inflows.
SPCX IPO Launch
SpaceX goes public at $225. Immediate retail enthusiasm — then gravity. Shares slide to ~$153. SpaceX-linked funds report losses.
June 26–27, 2026
Bloomberg reports the $270B leverage unwind. Semiconductor shares fall across Asia and US. Retail-fueled positions flush out. OpenAI IPO formally pushed to 2027.
The Undercurrent
RAMageddon: memory prices up 98%. Apple devices +$500. Xbox +$150. AI demand is structurally real — but equity markets are repricing who captures it.
The Structural Read
Two forces are colliding, and the market is confused about which one wins.
On one side: RAMageddon is proving that AI demand is not theoretical. Memory prices up 98% year-over-year. Consumer device prices rising across every major OEM. The substrate of AI — compute and memory — is getting paid. That thesis holds.
On the other side: equity markets are asking a different question. Not “is AI demand real?” but “who captures the economic surplus from AI demand?” And the answer is murky. SpaceX’s post-IPO slide signals that even the most hyped AI-adjacent assets are not immune to valuation gravity. OpenAI’s $3.7B quarterly burn suggests the frontier modelbusiness is expensive infrastructure, not a margin machine — at least not yet.
The $270B leverage unwind is the market’s way of de-rating the “AI capex always wins” trade. It does not mean AI is failing. It means the easy money from undifferentiated AI exposure — levered ETFs, speculative pre-IPO vehicles, chip baskets — has run its course.
The key insight: RAMageddon proves AI demand is real. The stock rout proves the market can no longer assume every player in the stack captures value equally. The substrate gets paid. The application layer is still fighting for its margin.
Three Implications
LEVERAGE UNWINDS FAST
Retail-fueled leveraged ETFs amplified both the AI bull run and the correction. The $270B number is not a loss — it is an exposure figure. But when retail exits leveraged vehicles simultaneously, the unwind is mechanical and fast. AI chip stocks fall not because the fundamentals changed overnight, but because the margined positions must be unwound.
THE SUBSTRATE THESIS SURVIVES
Memory and compute get paid regardless of which AI application wins. RAMageddon — 98% memory price inflation — is the clearest evidence that demand is structural. The companies selling picks and shovels (memory fabs, advanced packaging, custom silicon) have pricing power the application layer does not yet have.
IPO TIMING IS NOW A STRATEGIC SIGNAL
OpenAI pushing its IPO to 2027 is not a retreat — it is a read on the market. At $3.7B quarterly burn with thin margins, going public now would price in the coststructure without the margin expansion story. Waiting for the marginstory to emerge before going public is the rational move. But it also tells you where management thinks current valuations are: not yet justified.
FDE Framework
Founders, Distributors, Enablers — Who Captures the Surplus?
In every technology wave, value concentrates at one layer. In the current AI buildout: Enablers (chips, memory, infrastructure) are clearly winning on revenue. Founders (OpenAI, Anthropic) are burning fast. Distributors (Microsoft, Google) are absorbing the cost and hoping to monetize through existing channels. The rout is a market vote that Enabler economics are already priced in — and Founder economics remain unproven at scale.
Business Engineer Framework
The AI Supercycle: Who Captures Value When Every Layer Is Under Pressure
The AI Supercycle framework maps which players in the stack hold pricing power, which ones are margin-squeezed, and how to position before the next rerating. The rout changes the urgency of getting this right.
The $270 billion leverage unwind is not a verdict on AI — it is a verdict on the assumption that all AI exposure is created equal. The substrate gets paid: memory is up 98%, compute demand is structural, and RAMageddon is real. But the stock market is now asking which layer captures the surplus — and that question does not have a clean answer yet. Until it does, undifferentiated AI bets are going to keep finding gravity.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
