Meta Platforms (Nasdaq: META) is negotiating a multibillion-dollar agreement to purchase Google‘s custom artificial intelligence chips for its data centers starting in 2027, according to a report Monday that sent tech stocks moving in opposite directions.
The social media giant plans to integrate Google’s tensor processing units into its facilities while also exploring rental options through Google Cloud as early as next year, The Information reported, citing a person familiar with the discussions. Meta currently depends primarily on Nvidia (Nasdaq: NVDA) graphics processing units for its AI infrastructure.
Alphabet (Nasdaq: GOOG) shares climbed 2.1% in after-hours trading following the news, while Nvidia stock declined 1.8%. The market reaction reflects anticipation that Google may capture a larger share of the booming AI chip market, where Nvidia has maintained near-total dominance.
Google developed its TPUs more than a decade ago for internal AI operations. Google designed the chips as application-specific integrated circuits that target exclusively machine learning tasks, unlike Nvidia’s GPUs, which Nvidia originally created for video game rendering before developers adapted them for artificial intelligence work.
If the deal pushes through, it would mark a new era in Google’s chip strategy. The company has historically made its processors available only through cloud rental services rather than selling them directly to customers for installation in their own facilities.
The company also recently secured an agreement to supply up to 1 million tensor processing units to AI startup Anthropic, a deal that generated attention across technology markets.
Meta ranks among the world’s largest spenders on AI infrastructure, with Bloomberg Intelligence analysts estimating the company could allocate $40 billion to $50 billion for inference chip capacity in 2026 alone. Any change in Meta’s hardware purchases carries implications far beyond Meta and Google.
Google Cloud executives believe expanding TPU adoption could help the company capture approximately 10% of Nvidia’s annual revenue, which would represent billions in sales. The company has pitched prospective clients on how installing the chips on their premises can address strict security and compliance requirements for sensitive data.
The negotiations come as demand for AI computing power continues to surge across technology companies. Nvidia has struggled to meet customer orders despite expanding production capacity, creating opportunities for competitors to gain market share.
Meta is not the first to seek to diversify its chip sources. Amazon Web Services introduced its own AI chips, while Microsoft (Nasdaq: AMZN) announced plans for custom processors. These efforts also aim to potentially lower costs.
The success of any agreement between Meta and Google will depend on whether the TPUs can deliver competitive performance and power efficiency for Meta’s specific AI workloads. Meta is reportedly planning to test the chips for long-term viability before making final commitments.
Google recently announced its latest processor generation, called Ironwood, which Google is making widely available. The company describes it as more than four times faster than its predecessor and capable of handling demanding AI tasks from model training to real-time chatbot operations.
Information for this story was found via the sources and companies mentioned. The author has no securities or affiliations related to the organizations discussed. Not a recommendation to buy or sell. Always do additional research and consult a professional before purchasing a security. The author holds no licenses.
