Google TPU custom silicon negotiations (FundaAI/Bitget report)
Key Questions
What are the details of Google's TPU negotiations with Marvell?
Google is negotiating with Marvell for TPU customization and large model inference chip development, per FundaAI/Bitget reports. This builds on Marvell's $2B Nvidia partnership and 18 custom wins with $1.5B run-rate. It signals hyperscaler diversification post-Broadcom.
How does this impact Marvell's data center growth?
Marvell's FY26 data center revenue is $1.65B, up 42% YoY. Google TPU talks could escalate the pipeline amid 20-25% share in $118B ASIC market. Custom TPU/LLM inference chips and high-perf interconnects are key focuses.
What is the status of Google's chip supplier diversification?
Google pushes diversification, entering TPU talks with Marvell, which stands out in AI custom chips via Nvidia partnership. Pending primary confirmation for pipeline escalation. It adds to Marvell's >20 AI XPU wins.
Google TPU talks active per FundaAI/Bitget (ex-1ARZraQt/ex-062fd295) atop Nvidia $2B/18 custom wins/$1.5B run-rate; custom TPU/LLM inference chips/high-perf interconnect; FY26 $1.65B DC/42% growth; 20-25% $118B ASIC market upside amid diversification post-Broadcom; pending primary confirmation for pipeline escalation.