Google plans nearly two million new AI chips as it turns to Marvell for custom designs

2026-04-20 · Source: The Decoder · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure · Depth: Intermediate, quick

Summary

Google is reportedly collaborating with Marvell Technology to develop two new specialized AI chips for its data centers, aiming to produce nearly two million units. One chip is a memory processing unit designed to complement Google's existing Tensor Processing Units (TPUs) by handling AI tasks based on memory and compute requirements. The second chip is a new TPU specifically optimized for AI inference, which involves running finished AI models. This strategic move, with designs expected to be finalized by next year, also seeks to reduce Google's reliance on its current chip design partner, Broadcom, despite Broadcom having a contract with Google through 2031.

Key takeaway

For CTOs and VPs of Engineering evaluating AI infrastructure, Google's shift towards custom Marvell-designed chips for inference and memory processing signals a critical trend. You should assess your organization's reliance on single-vendor AI hardware and explore custom silicon partnerships to optimize performance and mitigate supply chain risks. Consider how specialized inference hardware could impact your operational costs and model deployment strategies.

Key insights

Google is developing custom AI chips with Marvell to enhance inference capabilities and diversify its supply chain.

Principles

• Custom silicon optimizes AI workload performance.
• Diversifying chip partners reduces vendor dependency.

In practice

• Integrate memory processing units with TPUs.
• Develop TPUs specialized for inference tasks.

Topics

• Google AI Chips
• Marvell Technology
• Tensor Processing Units
• AI Inference
• Memory Processing Units

Best for: Investor, CTO, VP of Engineering/Data, AI Hardware Engineer, AI Architect, Director of AI/ML

Related on AIssential

• Marvell Rallies 5% as Google Plans to Diversify for AI Chips — Diversifying custom AI chip suppliers reduces costs and supply chain risks for major tech companies.
• A PhD pivot, a stint in VC and a rumoured Anthropic deal: why Europe is watching Walter Goodwin — AI inference demand is growing exponentially, creating a strategic opportunity for specialized chip development to challenge incumbents.
• Google to Release New AI Chips, Challenging Nvidia — Google is specializing its TPU program for AI inference to meet surging demand and maintain its competitive edge.
• Meta Is Developing 4 New Chips to Power Its AI and Recommendation Systems — Meta is developing custom MTIA chips to power its internal generative AI and content ranking systems.
• Google and Intel deepen AI infrastructure partnership — Expanded Google-Intel partnership focuses on Xeon CPUs and custom IPUs to meet growing AI infrastructure demands.
• Rebellions Bets on Memory-Centric Architecture as It Weighs IPO Options — Memory-centric AI architectures and supply chain security are critical for large-scale LLM inference and market traction.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Decoder.