Oracle turns to new AI chip player

Oracle has added Cerebras AI chips to its cloud infrastructure, alongside Nvidia and AMD hardware. Oracle, CEO Clay Magouyrk, said that Oracle is utilising these chips to deliver AI workloads and large-scale computing workloads.

Cerebras CEO Andrew Feldman said that the company specialises in large-scale chips that are efficient in handling large-scale AI workloads. It had filed its Initial Public Offering (IPO) in 2024 but pulled it in October last year.

Getting Oracle as a client is a big win for them, especially since it reduces their dependency on a client in a Middle Eastern country that had been their main source of business.

This was after Cerebras disclosed its commitment of $10 billion from OpenAI. OpenAI’s Chief Technology Officer Mira Murati noted that the startup’s chips were used in its Codex-Spark, which is a generative AI model used for software development in ChatGPT Pro.

During a conference call after Oracle’s quarterly earnings announcement, Magouyrk noted that the company’s approach was to use a variety of accelerators. “We continually offer the latest in accelerators, from Nvidia and AMD to emerging designs from companies like Cerebras and Positron.”

This is seen as a partnership that shows how much the AI chip is in demand in cloud computing. Cerebras’ partnership with Oracle shows that even as chipmakers compete with industry leaders, it is easier for them to offer faster and more efficient AI operations.

Similarly, Nvidia acquired assets of AI chipmaker Groq to add to its portfolio. Magouyrk noted that investing in hardware is crucial to support AI workloads in all cloud environments.