Oracle said on October 27 that it would expand its AI cloud partnership with Nvidia, offer Google Tensor Processing Unit chips for the first time, and add systems from Cerebras and SambaNova.

Oracle and Nvidia said Oracle would deploy Vera Rubin servers at “multi-gigawatt scale” while replacing some Nvidia networking chips with Acceleron, which Oracle describes as its in-house SmartNIC platform for cloud and AI workloads.

Cerebras, a US AI chip startup, supports AI training and inference, while SambaNova, a US AI systems company, focuses on large-scale inference.

Google’s TPUs have typically been available as a service within Google Cloud.

Oracle did not disclose how it would fund the purchases.

On its September earnings call, it maintained its fiscal 2027 capital spending target of US$90 billion to US$95 billion and said customers were increasingly paying for chips.

🔗 Source: Reuters

Recent Oracle developments