Reports suggest Meta Platforms, Inc. (META.US) is exploring the use of Alphabet's (GOOGL.US) TPUs (Tensor Processing Units) to supplement its existing NVIDIA (NVDA.US) GPU supply. In response, Bank of America reaffirmed its "Buy" ratings on NVIDIA, AMD (AMD.US), and Broadcom (AVGO.US).

Analyst Vivek Arya noted in a client report that recent market speculation indicates Alphabet may lease its TPUs to Meta as early as next year, with potential localized deployment by 2027. Neither company has officially commented on such a deal, but if confirmed, it could intensify competition for NVIDIA and AMD, Meta’s current GPU suppliers.

Broadcom, a co-manufacturer of Alphabet’s TPUs, handles design and supply chain support, while TSMC manages the physical chip production. NVIDIA responded by stating it is "pleased with Alphabet’s success" and will continue supplying them, while emphasizing its market leadership. The company highlighted on social media: "NVIDIA is a generation ahead—the only platform capable of running all AI models across all computing scenarios. Compared to ASICs designed for specific AI frameworks, NVIDIA delivers superior performance, versatility, and replaceability."

Arya views competition in AI accelerators as a rising tide lifting all boats. He projects the total addressable market for AI data centers to grow fivefold by the end of the decade, reaching ~$1.2 trillion (up from $242B this year). While NVIDIA is expected to retain dominance, its market share may decline from ~85% to ~75%.

"Commercial GPUs offer key advantages: off-the-shelf availability, multi-cloud portability, NVIDIA’s full-stack software and developer ecosystem, and broader accessibility for government and enterprise clients lacking custom chip expertise," Arya explained. "Tight supply chains and NVIDIA’s scale make significant share shifts unlikely in the near term. Custom chips may be cost-effective for specific internal workloads—ideal for giants like Alphabet and Meta—but they’re less practical for public clouds like Microsoft Azure or Amazon AWS, or new cloud providers requiring high flexibility. That’s why even Alphabet uses GPUs in its public cloud."