Marvell Technology entered an expanded commercial agreement with Google on July 29 for custom semiconductors attached to Google's Tensor Processing Unit (TPU) ecosystem, securing a stock warrant worth up to $12.2 billion. The deal covers AI inference accelerators, storage controllers, network interface controllers, memory interface controllers, and near-memory compute solutions — a broader attachment surface than Marvell's previous GPU-only relationship with hyperscalers.
The warrant vests based on cumulative custom product revenue, with one tranche vesting for every $500 million in Google purchases across 240 tranches from Marvell's Q3 FY2027 through the end of FY2033. Marvell stock rallied ~10% on announcement; analysts at UBS raised their price target to $310 and Barclays estimated the deal could add $18.5 billion in annual revenue. Management's prior guidance projects custom AI chip revenue to double to over $4 billion in the coming year and exceed $10 billion by 2028.
For chip procurement strategists, this signals Google's deliberate pivot toward a multi-supplier TPU strategy, reducing concentration on Broadcom while diversifying into near-memory acceleration and infrastructure silicon — the bottlenecks that matter more than raw accelerator throughput as inference dominates model economics. Marvell's success here sets a template: the most valuable silicon partners are those solving the unglamorous plumbing around the headline accelerators.