Meta's plan to sell excess AI computing power triggered a 9% collapse in momentum-driven chip stocks and upended the core thesis that has fueled a $750 billion infrastructure buildout.
Meta's plan to sell excess AI computing power triggered a 9% collapse in momentum-driven chip stocks and upended the core thesis that has fueled a $750 billion infrastructure buildout.

Meta's plan to sell excess AI computing power triggered a 9% collapse in momentum-driven chip stocks and upended the core thesis that has fueled a $750 billion infrastructure buildout.
Meta Platforms Inc.'s plan to sell excess AI computing capacity through a new cloud initiative sent semiconductor and infrastructure stocks plunging, as investors reassessed the assumption that compute scarcity would sustain endless capital spending.
"The market's core premise has been compute scarcity — if supply increases and rental prices decline, that narrative gets disrupted," Rich Privorotsky, head of Goldman Sachs' 1-Delta trading desk, said in a note.
Meta shares surged 8.8% to $612.91, its best single-day gain of the year, as the market rewarded the prospect of capital discipline. The Bloomberg Magnificent 7 Index outperformed the Philadelphia Semiconductor Index by 8 percentage points, the widest spread since 2015. Goldman's high-beta momentum basket, dominated by chip and storage stocks, fell 9% in a single session — the worst since 2020, according to BTIG strategist Jonathan Krinsky.
The divergence signals a potential inflection point for the AI trade. After two years of hyperscalers spending without constraint — Meta alone guided $115 billion to $135 billion in 2026 capital expenditures — the first hint of monetizing idle capacity has shifted investor focus from infrastructure buildout to free cash flow generation.
The selloff was broad. Nvidia Corp. fell 1.3%, Advanced Micro Devices Inc. lost 6.9%, Intel Corp. dropped about 4%, and Broadcom Inc. slipped roughly 2%. Storage names fared worse: Micron Technology Inc. declined 7%, SanDisk Corp. fell 8%, and Seagate Technology Holdings PLC lost 4%. Chip-equipment makers including KLA Corp., Lam Research Corp., Applied Materials Inc. and ASML Holding NV also posted losses.
The hardest-hit names were the neocloud providers that built businesses around renting GPU capacity to hyperscalers. Nebius Group, which holds a $27 billion contract with Meta, tumbled 17%. CoreWeave, with a $21 billion Meta agreement, fell 14%. IREN Ltd. declined about 7%.
Meta Compute and the Competitive Shift
Meta's initiative, internally called Meta Compute, is led by infrastructure chief Santosh Janardhan, Meta Superintelligence Labs executive Daniel Gross and President Dina Powell McCormick. The company is weighing two approaches: selling access to AI models hosted on its own infrastructure — similar to Amazon Web Services' Bedrock service — or offering raw GPU computing capacity, the model that CoreWeave and Nebius have built their businesses on.
The irony is not lost on the market. Meta has been one of the largest customers of neocloud providers, signing tens of billions of dollars in contracts. Now it is considering becoming their competitor. The company has a 2,250-acre hyperscale campus in Louisiana and a gigawatt-scale data center under construction in the Midwest, alongside external deals including a 1.6-gigawatt agreement with data center firm Crusoe.
Liquidity Risk Compounds the Selloff
The rotation is unfolding in increasingly fragile market conditions. Goldman Sachs warned that S&P 500 E-mini futures top-of-book liquidity fell 33% month-over-month in June, from $12 million to $8 million — the shallowest depth of the year. That means large orders are moving prices more than usual, raising the risk of cascading moves in crowded trades.
UBS trader Christina Dwyer said the Meta report shifted the narrative toward stricter financial discipline, easing concerns about ever-rising capital expenditure. "Capex expectations are no longer tilted one-way upward," she said. "The focus is turning to free cash flow stability."
For investors, the question is whether this is a one-day repricing or the start of a structural rotation. Nvidia shares trade at roughly 35 times forward earnings, and any sustained compression in AI compute pricing could pressure that multiple. The upcoming second-quarter earnings reports from major hyperscalers will be the next test: if more companies signal capital expenditure restraint, the selloff in hardware names could deepen.
This article is for informational purposes only and does not constitute investment advice.