Islamabad, Feb 11: OpenAI is moving forward with plans to develop its own custom AI processor to reduce reliance on Nvidia’s hardware. This decision is part of a broader trend in the tech industry, where companies like Microsoft, Amazon, Google, and Meta have developed or are developing their own AI acceleration chips. The goal is to lower costs associated with high-performance GPUs like Nvidia’s and avoid supply shortages.
The AI chip project, which began gaining attention in October 2023, is being led by Richard Ho, a former Google chip designer, with a team of 40 engineers. The chip is expected to be fabricated by Taiwan Semiconductor Manufacturing Co. (TSMC) using its 3-nanometer process. OpenAI plans to focus the first version of the chip on AI inference, which involves running AI models rather than training them.
READ MORE: Meta’s AI Can Enable Users to Type Words Using Only Their Minds
The development of the chip faces significant challenges. Custom chip design can be very costly industry estimates suggest a single design could cost around $500 million, with additional costs for software and hardware development. Moreover, OpenAI plans to deploy these chips gradually, starting with limited production and aiming for mass production by 2026, though technical issues could cause delays.
The company’s interest in creating its own AI hardware also coincides with a broader push for increased investment in AI infrastructure. For example, Microsoft plans to invest $80 billion in AI infrastructure in 2025. OpenAI, meanwhile, has announced its $500 billion “Stargate” project with partners like SoftBank, Oracle, and MGX, aiming to expand AI data centers in the U.S.
In the long run, OpenAI’s venture into custom AI chips is a strategic move to gain more control over its hardware and reduce dependence on external suppliers like Nvidia. It is also expected to help the company improve cost efficiency and scalability as demand for AI computing power continues to rise.
