OpenAI moves to develop its own AI chips by 2026

OpenAI is putting the finishing touches on its first custom AI chip design, according to exclusive reporting from Reuters. The company plans to partner with Taiwan-based TSMC for manufacturing, with initial production trials expected to take several months.

The chip will use TSMC's cutting-edge 3-nanometer manufacturing process, keeping OpenAI on track to begin mass production in 2026.

A team of 40 people, led by former Google chip expert Richard Ho, is spearheading the project. The group is working closely with Broadcom to create a chip capable of both training and running AI models, though initial deployment will likely be limited.

The team's size raises some questions - industry experts say a full-scale chip program typically needs hundreds of engineers, making OpenAI's group look modest compared to similar efforts at Google or Amazon.

Sources familiar with the project say OpenAI sees the chip primarily as a way to gain leverage in negotiations with other suppliers. While Nvidia currently dominates the AI chip market, tech giants like Amazon, Microsoft, and Meta have been trying to develop their own hardware for years.

The push comes as demand for AI chips continues to surge. Amazon, Meta, and Microsoft are planning to pour tens of billions into AI infrastructure, while OpenAI itself is involved in the $500 billion Stargate infrastructure program.