Google Broadcom Chip Move

Google has expanded its custom-chip work with Broadcom, tightening a partnership that now runs through 2031 and complicating Intel’s push to win more hyperscaler chip business. (finance.yahoo.com 1) (finance.yahoo.com 2) Broadcom said on April 6 that it will produce future versions of Google’s artificial intelligence chips, and CNBC reported the announcement came alongside an expanded Anthropic arrangement tied to Google’s tensor processing units, or TPUs. Broadcom shares rose 3% in extended trading after the disclosure. (cnbc.com) A custom chip is a processor built for one customer’s workload instead of the wider market, and Google has spent years using TPUs as its in-house alternative to more general-purpose Nvidia graphics processors. Bloomberg’s reporting, distributed by Yahoo Finance, said the Broadcom-Google supply agreement runs for five years, through 2031. (finance.yahoo.com) The move lands just days after Intel announced a separate multiyear collaboration with Google on Xeon central processing units and custom infrastructure processing units, or IPUs, which handle networking, storage, and security tasks inside data centers. Intel said on April 9 that Google Cloud will keep deploying Xeon chips, including Xeon 6 processors in C4 and N4 instances. (intc.com) That split shows how one cloud customer can buy different layers of its artificial-intelligence stack from different suppliers at the same time. Google can lean on Broadcom for TPU design and supply while still using Intel for server processors and infrastructure offload chips. (intc.com) (finance.yahoo.com) The stakes are large because Broadcom’s artificial-intelligence chip revenue reached $8.4 billion in fiscal first-quarter 2026, up 106% from a year earlier, and the company told investors it expects AI semiconductor revenue of $10.7 billion in the next quarter. Yahoo Finance’s Bloomberg item said that revenue visibility is one reason investors treated the Google deal as more than a routine renewal. (finance.yahoo.com) Broadcom’s April 6 filing also tied Google’s chip program to Anthropic’s expansion, with Anthropic set to access about 3.5 gigawatts of computing capacity built on Google TPUs. CNBC reported that most of that new infrastructure will be located in the United States. (cnbc.com) Intel has not been pushed out of Google altogether, and Intel’s own statement framed the April 9 agreement as a multigeneration effort to improve performance and energy efficiency across Google’s global infrastructure. But the Broadcom expansion undercuts any simple reading that one Google announcement settles the wider contest for hyperscaler spending. (intc.com) (finance.yahoo.com) What happens next is likely to show up less in one headline contract than in how Google divides future orders across CPUs, networking silicon, custom accelerators, and manufacturing partners. For Broadcom, the latest win extends a long runway; for Intel, it leaves the burden on proving that separate Google ties can still turn into durable foundry and custom-chip volume. (finance.yahoo.com) (intc.com)