Meta's stock-bond double kill ignites a pivot day: spending commitments balloon to $825B, CDS at record, and the 10-year Treasury logs its biggest drop since the SVB crisis

The AI trade's "night of reckoning" arrived Thursday: Meta, up 6.5% after hours just one night earlier on a double beat, turned into a stock-and-bond rout within a day — a regulatory filing revealed total committed future spending of roughly $825 billion, five-year CDS hit an all-time high, and the stock fell about 7%, erasing over $100 billion of market cap; the Nasdaq dropped more than 2% on the "last straw," while the 10-year Treasury yield collapsed to 4.953% — a single-day 19.3bp move, the biggest since the 2023 Silicon Valley Bank crisis^[1][2][4].

Meta's $825B: from $700B to $825B in one quarter

The direct fuse was Meta's Thursday filing: as of end-September, total committed future spending stood at roughly $825 billion across leases, cloud contracts and datacenter agreements, much of it AI-compute related — over $100 billion more than the ~$700 billion disclosed in July^[1]. Unpacked, the $825 billion has three parts: $355.7 billion of off-balance-sheet operating and finance leases (datacenter and energy infrastructure phasing in through 2036); $454.5 billion of irrevocable short- and long-term contracts (third-party cloud compute purchases, datacenter energy agreements and Reality Labs hardware); plus up to $14.7 billion of cloud capacity committed over five years^[1]. The cash flow ledger is just as tight: Q3 free cash flow was -$2.12 billion, the first deficit in some 14 years as a listed company; Barclays models Q4 capex at $49–59 billion (midpoint $55 billion, "the largest single quarter ever"), implying a cash shortfall that could exceed $20 billion — with C-share equity funding unavailable until early 2027, a bond deal this year is nearly inevitable, baseline around $25 billion, with "larger transactions no longer a low-probability event"^[1].

Market structure: tech breaks, Treasuries rally — a "take-back trade"

Meta's filing became the last straw. Thursday's Wind US Magnificent-7 index fell 2.80% in a day: Meta down nearly 7%, Amazon and Nvidia down over 4%, Tesla and Alphabet down over 3%; the Nasdaq fell more than 2% and gold slumped in tandem — while Treasuries surged, the 10-year yield dropping 19.3 basis points to 4.953%, below the 5% mark for the biggest single-day move since the 2023 banking crisis^[2][4]. PGIM's Tipp calls it "a take-back trade" — "not an all-out fear-of-recession type of situation," but a concentrated repricing of the AI-capex narrative^[3]. Note the shift in haven choice: money flowed into Treasuries rather than gold — analysis points out that gold's haven demand is being diverted by Treasuries, which in a still-high-rate world offer "yield-bearing safety"^[4].

The financing chill cascades

Meta's double kill landed at a critical window for AI fundraising. OpenAI just cut its round's valuation from $1.4 trillion to $1.25 trillion — Morgan Stanley's private-wealth channel saw "no scrambling at all, when every previous round was a feeding frenzy"; some investors explicitly prefer to wait until Anthropic's IPO before committing to its rival. The tech selloff has turned "can the debt-fueled AI infrastructure boom continue" from a fringe worry into a mainstream question — Meta's record CDS is effectively the credit market's new risk premium on the whole industry's capex model^[2].

What to watch

First, next week's Fed meeting — the inflation language under oil shock decides whether the rate side can steady risk assets. Second, the size and pricing of Meta's in-year bond deal — the repricing anchor for AI credit risk. Third, capex-guidance wording in this week's remaining earnings. Fourth, where Treasury yields stabilize — if the 4.9% area holds, the "take-back" is just a pullback; a fast break lower would flip the regime into recession trading^[1][3][4].

Also worth recording is the speed: from Wednesday's after-hours earnings euphoria to Thursday's stock-bond rout, only a regulatory filing and a dozen-odd hours stood in between. A reminder to everyone trading the AI narrative — in a cycle where committed spending compounds by hundreds of billions, the pricing regime can switch faster than any model's reasoning.