OpenAI Reportedly Seeks Google TPU Chips, Oracle to Add TPU to Cloud
SingTao · 2 SOURCES1 day ago2 MIN
Summary
OpenAI is reportedly planning to purchase Google's tensor processing units (TPUs) to expand its AI computing capacity, marking a significant shift in the AI chip supply chain. According to reports from The Information and Bloomberg, the deal would deepen cooperation between OpenAI and Google despite their competing positions in AI products and search. Oracle simultaneously announced plans to become the first hyperscaler beyond Google to offer TPU services, with launch expected in 2027. This development reflects the AI industry's push to reduce reliance on Nvidia GPUs and develop diverse computing infrastructure.
Key Points
- OpenAI plans to purchase Google TPUs to diversify its AI computing sources beyond its traditional supplier Nvidia
- OpenAI is developing its own AI chip called Jalapeno in partnership with Broadcom, with Google alumni on the development team including Richard Ho, a former Google TPU researcher
- Oracle announced it will offer Google TPU services on its cloud platform in 2027, becoming the first hyperscaler besides Google to provide this option
- Google began developing TPUs over a decade ago, initially to accelerate search engine operations before expanding into AI computing
- These moves represent a broader industry trend of tech companies developing custom processors to challenge Nvidia's dominance in the AI chip market
Why It Matters
The strategic partnerships and chip developments signal a fundamental shift in AI infrastructure development, with major tech companies seeking alternatives to Nvidia's dominant position in GPU supply. For Hong Kong, this evolution in AI computing could affect cloud service costs and availability, as regional data centers may integrate these new chip options into their offerings .
The strategic partnerships and chip developments signal a fundamental shift in AI infrastructure development, with major tech companies seeking alternatives to Nvidia's dominant position in GPU supply. For Hong Kong, this evolution in AI computing could affect cloud service costs and availability, as regional data centers may integrate these new chip options into their offerings .