344
Productivity & Workflow355
Automation & Workflow224
Software Development251
Marketing & Growth192
AI Infrastructure & MLOps174
Writing & Content Creation203
Data & Analytics142
Photography & Imaging156
Design & Creative170
Customer Support132
Sales & Outreach125
Voice & Speech135
Education & Learning131
Operations & Admin87
Google will work with Marvell on custom AI chips and has a warrant to buy up to $12.2bn of Marvell stock, according to a filing.
In short: Google expanded a deal with chipmaker Marvell to build custom AI chips and got the option to buy up to $12.2bn of Marvell shares.
Google has struck an expanded partnership with Marvell Technology to help develop custom hardware for Google’s AI chips, which Google calls tensor processing units, or TPUs (special chips made to run AI tasks faster).
Marvell said it issued a warrant that lets Google buy up to 58.9 million Marvell shares at $206.58 each, according to a regulatory filing. If Google bought the full amount, it would total about $12.2bn.
Marvell works with large cloud companies, sometimes called “hyperscalers” (think of them as giant landlords for computing power), to design specialized AI “accelerator” chips. These chips can be an alternative to Nvidia’s more general chips, which are widely used across the industry.
After the announcement, Marvell’s shares rose 8% on Wednesday. Broadcom, which the Financial Times describes as Google’s main supplier for TPUs, fell 5%.
The expanded partnership comes as Google starts selling TPUs to outside customers, not just using them inside its own data centers. Marvell said the work with Google could include parts that help AI systems communicate and move data, like network and memory controllers (similar to traffic signals and on-ramps that keep data flowing).
AI tools depend on huge amounts of computing power, and the chips inside data centers help set the speed and cost. Deals like this show big tech companies are trying to secure their own supply and reduce reliance on any single chip supplier. Over time, that can affect how quickly AI features roll out and how expensive they are to run.
Source: Financial Times