Nasdaq futures slid over 1% as conflicting AI spending signals from Tesla and Meta reignited the "is this boom paying off" debate. Chip names led the premarket damage: Intel down roughly 5%, AMD down about 4%, and Nvidia off 2.3%, with Qualcomm, Marvell, Broadcom and Micron sliding between 2.9% and 5.3%. This follows an eight-session rout through October 16 that erased roughly $2 trillion from the Philadelphia Semiconductor Index before a partial rebound.
Meta posted revenue of $64.60 billion, up 26% year over year and ahead of the $63.3 billion consensus, with ad prices up 14% on AI-driven targeting. GAAP EPS came in at $3.83 including a $9.89 billion charge — roughly $7.40 excluding it — and the company narrowed full-year data-center capex to $135 billion to $145 billion, lifting the low end. Shares rose between 2.5% and 5% after hours; JPMorgan nudged its target to $940 from $920, Overweight.
Tesla delivered adjusted EPS of $0.48 on $28.05 billion in revenue, topping estimates, with gross margin up 160 basis points to 19.6%. GAAP net income still fell 43% to $787 million as operating expenses jumped 39% on AI compute, Optimus lines and the Terafab chip project; Q3 capex more than doubled to $4.73 billion. Shares rose 3.41% in the session to $386.47, then fell 2.22% after hours to $377.88 as the lower full-year capex number read as a slower AI buildout.
Lam Research reported adjusted EPS of $2.20 on revenue of $8.22 billion, up 54% year over year, with systems revenue up 62% to $5.73 billion. Guidance of $8.6 billion plus or minus $400 million also beat, but management said supply-chain bottlenecks capped the December outlook and flagged a smaller China contribution. Shares fell about 7.4% after hours from a $260.81 close to roughly $241.57; Lam's preliminary 2027 wafer-fab-equipment call is above $200 billion, led by DRAM.
At GTC Berlin, Nvidia lined up Ibiden, Samsung Electro-Mechanics, Unimicron, Amkor, SPIL and King Yuan to scale packaging and test capacity through 2028, including a $1.1 billion advance to King Yuan to reserve U.S. testing capacity. Separately, Maeil Business Newspaper reports Nvidia and Tesla will put 3.7 trillion won — about $2.8 billion — into Samsung Electro-Mechanics FC-BGA substrate lines in Sejong, Nvidia for roughly 2.6 trillion won and Tesla 1.1 trillion. Forbes also tallies about $164.5 billion in Nvidia guarantees and commitments behind data centers running its chips, which is fueling the circular-financing critique.
OpenAI is reportedly seeking about $30 billion at a valuation north of $1.4 trillion, with SoftBank and MGX in discussions, trying to close before Anthropic's expected mid-November listing puts comparable financials in front of investors. OpenAI has told investors it expects annualized revenue of $70 billion or more by year-end, up from roughly $50 billion in September, and is targeting a 2027 listing. Anthropic's Q3 revenue reportedly rose to $16.8 billion from $11.6 billion, though some investors have floated a roughly $1.5 trillion valuation versus the $2 trillion originally sought. Neither is publicly traded.
IBM posted $17.1 billion in revenue, up 6% at constant currency, with software at $7.8 billion up 9%, Red Hat up 11% and data up 20%. Generative AI now sits inside about 35% of the consulting backlog, and IBM raised its free-cash-flow outlook. Shares still faded after hours as management held full-year constant-currency revenue growth at 4% to 5%.
Bernstein said Thursday that Samsung overtook SK hynix in high-bandwidth memory revenue in the third quarter, with SK hynix HBM revenue up 11% sequentially to $7.6 billion. The shifting share picture matters for Micron, which says HBM4 is now in high-volume manufacturing with its 1-gamma DRAM node anchoring the next twelve months of product. Micron fell 14.6% during last week's chip rout even as sell-side targets have been climbing, including D.A. Davidson's move to $3,000 from $2,100 earlier this month.
Anthropic is reportedly negotiating with Nscale for as much as $45 billion of AI computing capacity at a second facility on Nscale's Monarch campus in West Virginia. The buildout would run on Nvidia's next-generation Vera Rubin accelerators, adding to the demand signal underpinning Nvidia's supply-chain expansion. For public-market investors, this is a read-through on 2027 and 2028 accelerator and memory order books rather than a tradable name.
Palantir rose 3.38% Tuesday and has been collecting target hikes on commercial momentum, with a consensus target near $207 and Goldman Sachs upgrading to Buy at $230 earlier this month. New sovereign-AI customers and deployments include L3Harris, McCarthy, Nvidia, Ondas and Tampa General Hospital. The pushback is price: roughly 124 times forward earnings against about 22 times for application-software peers, and the stock dropped about 5% during last week's AI-financing scare.
AI Stocks
