Photo Credit: Getty Images
Meta Platforms is reportedly considering a multibillion-dollar deal to adopt Google's in-house AI chips, a move that could shift momentum in the global race for artificial intelligence hardware. According to The Information, Meta is evaluating Google's tensor processing units (TPUs) for its data centers beginning in 2027, while potentially renting additional chip capacity from Google Cloud as early as next year.
The talks come as Google aggressively positions its custom chips as a viable alternative to Nvidia, the long-standing market leader. Earlier this year, Google struck a major supply agreement with Anthropic for as many as one million TPUs, a deal analysts described as a "powerful validation" of the technology's readiness to compete.
The timing is notable: Alphabet is poised to reach a historic $4 trillion valuation, while Nvidia is facing market pressure and investor concerns over an emerging AI bubble. High-profile critics like Michael Burry have questioned the sustainability of Nvidia's growth, pointing to circular AI spending patterns and rapid hardware depreciation. Nvidia shares dipped in premarket trading following the Meta, Google report.
Google says demand is accelerating for both its TPUs and Nvidia's GPUs, noting that customers increasingly want diversified AI infrastructure. Analysts at Bloomberg Intelligence estimate Meta will spend as much as $40 billion to $50 billion on inferencing-chip capacity in 2026 alone, giving Google a rare opportunity to chip away at Nvidia's dominance.
Asian suppliers rallied on the news, with South Korea's IsuPetasys, a maker of multilayer boards for Alphabet, surging to a record high. Taiwan's MediaTek also posted strong gains.
Google's newest TPU, Ironwood, reflects more than a decade of internal development, strengthened by collaboration between the company's DeepMind researchers and chip designers. Unlike GPUs, originally built for gaming but later adopted for AI, TPUs are application-specific chips engineered from the ground up for machine learning workloads.
If finalized, the Meta deal would be a major boost for Google and its partner Broadcom. But the long-term impact will depend on whether TPUs can deliver the performance and efficiency required to seriously challenge Nvidia's hold on the AI ecosystem.
