Beta The Briev beta is out. Free on iPhone via TestFlight — install it in under a minute.

Join the beta ↗
Briev
Live
Business

Google taps Marvell for custom AI silicon, widening its supplier base

Google disclosed a partnership with Marvell to develop custom silicon for its TPU ecosystem, while still working with Broadcom.

Google revealed through an SEC filing that it has contracted Marvell to design custom silicon that will augment its Tensor Processing Unit (TPU) platform, covering AI inference accelerators, storage and network controllers, among other parts. The arrangement coexists with Broadcom’s ongoing role as a primary TPU supplier, showing the cloud giant’s strategy of sourcing from multiple IP vendors. Marvell’s filing also included a warrant allowing Google to acquire nearly 59 million shares, estimated at $12.2 billion, though it remains uncertain whether the warrant will be exercised.

The chipmaker’s extensive IP suite—high-speed SerDes, multi-die XPU designs, CXL memory controllers and photonic fabric technology acquired from Celestial AI—offers Google alternatives for memory-rich, optical interconnects across data-center racks. Market reaction was swift, with Broadcom shares dropping roughly 4 % after the news. The partnership underscores Google’s effort to avoid over-reliance on a single supplier and to potentially drive down costs while enhancing performance.

Why it matters

Google’s new chip partnership could reshape AI hardware sourcing and affect market dynamics for major silicon vendors.

In this story

GoogleMarvellBroadcomTPUcustom siliconAI inference acceleratorphotonic fabricSEC filingstock price drop
Get the beta ↗