AI for Business

Google's Multi-Vendor Chip Strategy Takes Shape with Marvell Talks

Google is in discussions with Marvell Technology to design two new AI processors, according to a report from The Information. The potential deal would see Marvell create a memory processing unit...

Share:

Google is in discussions with Marvell Technology to design two new AI processors, according to a report from The Information. The potential deal would see Marvell create a memory processing unit to complement Google's Tensor Processing Units (TPUs) and a new TPU variant specifically for AI inference. While no contract is finalized, the talks point to a clear strategic shift.

This follows Google's recent extension of its primary chip design partnership with Broadcom through 2031. The parallel efforts reveal a calculated move to distribute its custom silicon work across several suppliers. For Google, inference—the act of running trained AI models for user queries—represents a colossal and continuous expense, powering everything from search to its Gemini assistant. Efficiency gains here translate directly to significant financial savings.

Bringing Marvell into the fold would add a specialized designer to a roster that already includes Broadcom for high-end TPUs and MediaTek for cost-optimized versions, all manufactured by TSMC. This structure is designed to mitigate supply and pricing risks. A previous attempt to replace Broadcom, dubbed 'Granite Redux,' was abandoned, with Google later affirming Broadcom as an "excellent partner." The current approach is about addition, not substitution.

For Marvell, a Google win would cement its position in the booming custom AI chip sector. The company's data center revenue recently surged to $6.1 billion, fueled by design wins with other cloud giants. Broadcom, meanwhile, maintains a dominant market share and shows no signs of slowing, with its AI-related sales more than doubling last quarter.

The underlying dynamic is one of necessity. As inference costs balloon, hyperscalers like Google are building intricate, multi-supplier ecosystems to ensure flexibility and control. By engaging Marvell, Google isn't just sourcing chips; it's buying strategic optionality for the future of AI at scale.

Source: Webpronews

Ready to Modernize Your Business?

Get your AI automation roadmap in minutes, not months.

Analyze Your Workflows →