OpenAI will start using AMD chips and will be able to build its own AI hardware in 2026

OpenAI will start using AMD chips and will be able to build its own AI hardware in 2026

According to reports ReutersOpenAI is working with Broadcom to develop new specialized chips designed to handle large-scale AI tasks. The company has already secured production capacity at TSMC plants. According to sources, OpenAI has formed a chip development team of about 20 people, including top engineers who previously worked on Google’s Tensor processors for AI.

However, according to preliminary estimates, production of these specialized chips may not begin until 2026.

At the same time, sources report that OpenAI embeds AMD chips into its Microsoft Azure infrastructure. AMD unveiled its MI300 chips last year, which was big news as its data center business doubled in a year in an attempt to catch up with market leader Nvidia.

The Information reported back in July that OpenAI was in talks with Broadcom and other semiconductor developers to create its own AI chip. Earlier this year, Bloomberg reported plans for OpenAI to build its own network of manufacturing facilities, but those plans have been put on hold due to high costs and long implementation timelines, according to Reuters.

The strategy puts OpenAI on par with other tech companies trying to manage the cost and access to AI server hardware with their own chip designs. However, Google, Microsoft and Amazon already have significant experience in their development, and OpenAI may need significantly more funding to become a real competitor.

Source

Related posts