Google Talks with Marvell on New AI Memory Chip, Marvell Stock Up 5.3%
Google is discussing with Marvell Technology to co-develop a memory processing unit to augment its tensor processing units and a TPU for AI inference, targeting memory chip design finalization by next year. Marvell shares jumped 5.3% premarket as a win with Alphabet would diversify its AI chip sourcing beyond Broadcom.
1. Deal Overview
Alphabet's Google is in negotiations with Marvell Technology to jointly develop two new chips aimed at improving AI workload efficiency. The proposed collaboration focuses on a memory processing unit (MPU) and a tensor processing unit (TPU) dedicated to AI inference, marking Google's move to broaden its chip ecosystem.
2. Technical Roadmap
Under the proposed timeline, the companies aim to finalize the MPU design by next year and proceed to test production shortly thereafter. The MPU is intended to complement Google's existing TPUs, while the new TPU will be optimized specifically for inference tasks, enhancing performance for large-scale AI models.
3. Strategic Implications
Google currently partners with Broadcom for TPU design but is seeking to diversify supply as demand for custom AI processors surges. A partnership with Marvell would add a marquee hyperscaler to Marvell's roster, alongside its recent $2 billion funding from Nvidia and revenue projections approaching $15 billion by fiscal 2028.
4. Market Reaction
News of the talks drove Marvell stock 5.3% higher in premarket trading, reflecting investor enthusiasm for a potential hyperscaler deal. The move underscores strong market appetite for AI chip developers that can secure long-term partnerships with leading cloud providers.