Meta is reported to be considering adopting Alphabet Inc. Class C TPU. Bank of America: the commercial GPU advantage of NVIDIA Corporation (NVDA. US) is still difficult to shake, reaffirming a "buy" rating.

In the market rumors that Meta (META.US) is considering using Alphabet Inc. Class C (GOOGL.US) TPU (Tensor Processing Unit) to supplement its existing NVIDIA Corporation GPU supply, Bank of America Corp reiterated its "buy" rating on NVIDIA Corporation (NVDA.US), AMD (AMD.US), and Broadcom Inc. (AVGO.US). Bank of America Corp analyst Vivek Arya wrote in a report to clients, "Yesterday evening, market news suggested that Alphabet Inc. Class C may rent out its TPU to Meta next year, followed by possible local deployment in 2027. Neither company has commented officially on such a deal, but if true, this could intensify the competitive landscape that Meta's current GPU suppliers NVIDIA Corporation and AMD face." Broadcom Inc. is the joint manufacturer of Alphabet Inc. Class C TPU, responsible for design work and assisting in the supply chain. The physical manufacturing of the chips is done by Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR. NVIDIA Corporation released a statement saying, "We are pleased with the success of Alphabet Inc. Class C" and will continue to supply to them, but also emphasized that NVIDIA Corporation remains the market leader. The company posted on social media, "NVIDIA Corporation leads the industry generation - it is the only platform that can run all AI models in all computing scenarios. Compared to ASICs designed for specific AI frameworks or functions, NVIDIA Corporation offers superior performance, versatility, and replaceability." However, Arya believes that competition in the AI accelerator field is intensifying. He expects that by the end of this decade, the total addressable market for AI data centers will grow approximately fivefold to around $1.2 trillion, from $242 billion this year. Arya explains that by then, NVIDIA Corporation will still dominate the market, but its market share may drop from the current estimate of 85% to 75%. Arya adds, "Commercial GPU chips have many advantages: readily available, can be ported to multiple clouds, NVIDIA Corporation's full-stack software and developer ecosystem, and a larger addressable market for government and enterprise local customers without specialized chip knowledge. In addition, tight supply chains and NVIDIA Corporation's scale advantage make it more difficult to capture more market share, as component shortages in the near to mid-term make changes in market share essentially gradual. For specific ranges of internal workloads, custom chips may be lower cost, which may suit customers like Alphabet Inc. Class C and Meta with a large amount of internal workloads. However, they have a smaller usage in public clouds like Microsoft Corporation Azure or Amazon.com, Inc. AWS which require high flexibility, which is why even Alphabet Inc. Class C uses GPUs in their public cloud."