AI news ·
Marvell partners with Google on custom AI chips as ASIC revenue hits $1.5 billion
Marvell is building a memory processing unit and next-gen TPU with Google as cloud giants work to cut dependence on Nvidia, which holds up to 92% of the AI accelerator market. Marvell's custom chip revenue hit $1.5B this fiscal year.
Marvell Teams With Google on Custom AI Chips as Cloud Giants Escape Nvidia Dependency
Marvell Technology is developing a memory processing unit and next-generation TPU alongside Google, marking the chipmaker's emergence as a core partner in custom AI silicon development.
The partnership reflects a broader industry shift. Nvidia controls 85% to 92% of the AI accelerator market, prompting Microsoft, Amazon, Google, and other hyperscalers to build their own chips and reduce reliance on a single supplier.
Marvell's Rapid Growth in Custom Silicon
Marvell's custom ASIC business grew from nearly zero to $1.5 billion in annual revenue in a single fiscal year. The company secured 18 design wins from major cloud providers, establishing itself as a viable alternative for AI infrastructure.
The Google collaboration increases Marvell's visibility among tech decision-makers evaluating chip suppliers. It also signals confidence in the company's ability to co-design solutions rather than simply manufacture existing designs.
Financial Projections and Technology Bets
Marvell projects $15 billion in total sales by fiscal 2028, with annual growth exceeding 30%. The company acquired Celestial AI to strengthen its photonic interconnect capabilities, addressing a bottleneck in data center performance.
For IT and development teams, this matters because custom chips from major cloud providers will increasingly influence architecture decisions. Marvell's role in designing these systems could affect availability, pricing, and performance characteristics of infrastructure you build on.