Broadcom will build Google’s next-generation TPU chips and scale Anthropic computing to 3.5GW, challenging Nvidia’s GPU dominance
Broadcom will produce future versions of Google’s next-generation artificial intelligence processor, the Tensor Processing Unit (TPU). (Reuters)
Broadcom has agreed to produce Google’s next-generation AI chips and will significantly expand its partnership with AI startup Anthropic.
According to major foreign media outlets such as CNBC on April 6 (local time), Broadcom disclosed through a securities filing that it has agreed to produce future versions of Google’s next-generation artificial intelligence processor, the Tensor Processing Unit (TPU). Alongside this, the company also signed an expanded agreement to provide Anthropic with approximately 3.5 gigawatts (GW) of computing capacity using Google’s AI processors.
According to industry sources, this collaboration comes amid a surge in demand for infrastructure to operate generative AI models. In particular, as Anthropic’s Claude app has seen a rapid rise in popularity, reaching the top spot for free apps on the U.S. App Store in February, the need for infrastructure expansion has become urgent. Anthropic stated that its current Annual Recurring Revenue (ARR) recently surpassed $30 billion, up from $9 billion at the end of last year, and that the number of corporate clients spending more than $1 million annually has doubled in two months to over 1,000 companies.
Broadcom’s CEO, Hock Tan, had mentioned during last month’s earnings call, “We are off to a good start, providing 1GW of Google TPU-based computing capacity for Anthropic in 2026.” According to industry sources, this demand is expected to exceed 3GW in 2027 through the expanded agreement. Analysts at Mizuho Securities have projected that Broadcom will generate $21 billion in AI−related revenue in 2026 and $42 billion in 2027 from its partnership with Anthropic.
Currently, major AI model developers like Anthropic and OpenAI heavily rely on Nvidia’s GPUs through the cloud services of Amazon, Google, and Microsoft. According to industry sources, Broadcom is collaborating not only with Google but also with OpenAI to develop custom AI silicon, which is interpreted as a significant variable that could alter Nvidia’s dominant market position.
According to industry sources, Broadcom’s latest move is seen as an intention to solidify its position as a key design partner across the entire AI computing infrastructure and supply chain, moving beyond simple chip manufacturing. As the performance of AI models becomes more advanced, the importance of Application-Specific Integrated Circuits (ASICs) grows. The analysis suggests that Broadcom’s custom semiconductor strategy will reshape the profit structure of the future AI market.
Kim Eun-jin
댓글
0