AMD gets another vote of confidence in the AI chip wars - this time from Oracle
Briefly

AMD gets another vote of confidence in the AI chip wars - this time from Oracle
"The AI chip deals keep on coming, and this time it's between AMD and Oracle. The companies said Tuesday that AMD plans to deploy 50,000 of its most advanced AI chips inside data centers run by Oracle, starting in the second half of 2026. Oracle and AMD did not disclose the financial terms of the deal in Tuesday's announcement. Neither immediately responded to a request for comment from Business Insider."
"Oracle is set to use AMD's Instinct MI450 graphics processing units, or GPUs, the same chips set to be deployed by OpenAI. They can be used for training AI models and for inference: when models apply their knowledge to tasks. "We feel like customers are going to take up AMD very, very well - especially in the inferencing space," Karan Batta, senior vice president of Oracle Cloud Infrastructure, told CNBC on Tuesday. AMD shares opened about 2.5% on Monday's close - far lower than the 27% surge after last week's deal with OpenAI. Oracle's shares opened up 2.2% but quickly lost their gains and were down as much as 5% in morning trading."
"On Monday, Broadcom and OpenAI confirmed a deal for 10 gigawatts of custom AI accelerators. Last month, Nvidia announced that OpenAI will gain access to 10 gigawatts of GPUs, alongside a $100 billion investment from the chipmaker."
AMD plans to install 50,000 of its most advanced AI chips in Oracle-run data centers beginning in the second half of 2026. Oracle will use AMD's Instinct MI450 GPUs, which support both training of AI models and inference workloads. The deployment follows a multi-billion-dollar supply deal between AMD and OpenAI and aims to shore up AMD's competitiveness against market leader Nvidia. Oracle's cloud executive projected strong customer adoption in inference use cases. The announcement occurred amid other large AI infrastructure deals, including Broadcom and Nvidia arrangements with OpenAI.
Read at Business Insider
Unable to calculate read time
[
|
]