Nvidia Shares Slip on Report of Meta-Google AI Chip Talks
A potential deal for Meta to use Google's TPU accelerators marks one of the most significant challenges to Nvidia's dominance in the AI hardware market.
Nvidia Corp. shares fell in late trading Monday following a report that Meta Platforms is in discussions to use Google's custom-designed artificial intelligence chips, a move that would represent a formidable new challenge to Nvidia’s near-monopoly in the AI hardware sector.
Shares of the chipmaking giant, which boasts a market capitalization of over $4.3 trillion, slid as much as 1.8% in after-hours trading. The dip came after a report from The Information detailed advanced talks for Meta to incorporate Google's Tensor Processing Units, or TPUs, into its data centers. Conversely, shares of Google's parent company, Alphabet Inc., rose 2.1% on the news, signaling investor optimism about the potential validation of its long-running AI silicon efforts.
The potential partnership is significant not just for its scale—Meta is one of the world's largest buyers of AI accelerators—but for its strategic implications. For years, Nvidia has maintained an iron grip on the market for the powerful GPUs that train advanced AI models, holding an estimated 86% market share. While major cloud providers like Google, Amazon, and Microsoft have all developed in-house chips, they have primarily been for internal use. A deal to deploy Google's TPUs directly within Meta's infrastructure would mark a major strategic shift and the first credible threat from a rival cloud provider to Nvidia's commercial dominance.
According to a Benzinga report covering the news, the discussions underscore a more aggressive phase in the battle for AI infrastructure. As demand for generative AI continues to soar, large technology companies are desperate to secure computational power and are increasingly wary of being beholden to a single supplier. For Meta, diversifying its hardware supply chain could provide crucial leverage against Nvidia's high prices and tight supply, while also aligning with its long-standing focus on open-source and open-hardware initiatives.
Nvidia’s powerful H100 and next-generation Blackwell-series GPUs are the gold standard for training large language models, largely due to their raw performance and the strength of Nvidia's proprietary CUDA software platform, which has created a deep competitive moat. However, the costs associated with building out massive AI factories have pushed major players to seek alternatives.
Google has invested billions in developing its TPUs over the past decade, tuning them specifically for the AI workloads that power its Search and Cloud businesses. A high-profile customer like Meta would serve as a powerful endorsement of its technology, potentially opening the door for Google to sell its custom silicon to other large enterprises. Reports from Investing.com suggest the deal could involve Meta spending billions and renting additional capacity from Google Cloud, representing a significant new revenue stream for the company.
Despite the competitive overture, Wall Street remains broadly bullish on Nvidia's prospects. The company has 59 analysts covering the stock with a 'Buy' or 'Strong Buy' rating, compared to just five with a 'Hold' or 'Sell' rating, according to market data. The company's recent earnings have consistently shattered expectations, with its quarterly revenue growth year-over-year standing at over 62%.
Still, the market's swift reaction to the report highlights a growing sensitivity to any cracks in Nvidia’s armor. The long-term question for investors is whether the formidable software ecosystem around CUDA will be enough to fend off increasingly powerful and cost-effective alternatives from deep-pocketed rivals. While a definitive agreement between Meta and Google has not been announced, the reported discussions alone signal that the landscape for AI hardware is becoming intensely competitive.