OpenAI has made its choice: AMD processors will be used!

 OpenAI has partnered with Broadcom and TSMC for its chip development plans, and the company will be using AMD MI300 chips for now.

OpenAI continues its path in the field of artificial intelligence without slowing down. Its next big goal is to develop its own special chips. In this field where technology giants are in high competition, OpenAI has taken action to produce special chips that will accelerate artificial intelligence processes. However, until that day comes, AMD's MI300 chips will be used. Here are the details...

Important step from OpenAI: It will use AMD processors until it produces its own AI chips

OpenAI is reportedly collaborating with Broadcom on chip development and has reached an agreement with TSMC (Taiwan Semiconductor Manufacturing Company) to secure production capacity. Plans are expected to be fully operational by 2026, but OpenAI is determined not to sit idle until then. The company will use AMD’s MI300 chips for its infrastructure.

AMD MI300 chips are particularly effective at meeting AI and data center workloads. The chips, introduced in 2023, have contributed greatly to AMD nearly doubling its data center revenues. OpenAI will continue to develop its own custom AI chips while supporting its infrastructure with these powerful chips.

It is said that the chips, which will be located in Microsoft Azure infrastructure in particular, will make OpenAI's ChatGPT and other artificial intelligence models much faster and more efficient. One of OpenAI's goals in the chip development process is to achieve the success we have seen in the custom chips of giants such as Google, Microsoft and Amazon.

However, there are challenges to this ambitious goal, such as cost and time, that cannot be ignored. According to previous Bloomberg reports, OpenAI also planned to build its own chip manufacturing facilities, but this plan was temporarily put on hold due to high costs and time constraints.

Still, OpenAI has taken steps toward chip development, building an engineering team of about 20 people, including veteran engineers who worked on Google’s Tensor, an artificial intelligence chip.