Serving technology enthusiasts for more than 25 years. TechSpot is the place to go for tech advice and analysis.
Context: The most advanced AI models are still largely operated and trained on Nvidia chips. This could change quickly. OpenAI is leading a massive industry effort to bring custom AI accelerators at a lower price to the market. If successful, the push could weaken Nvidia’s dominance in AI hardware, pushing the company to a more competitive market.
OpenAI’s first custom-designed AI chips is about to be launched. Reuters expects that OpenAI will send its chip design to TSMC for validation in the coming months before mass production starts in 2026. OpenAI will need to hire more people in order to be fully self-sufficient on the AI accelerator market.
A “small” team in-house led by Richard Ho designed the custom chip. Richard Ho left Google over a year ago to join OpenAI. The 40-person team worked with Broadcom, an controversial company that is well-known for its custom ASIC solutions. The two companies began to negotiate a chip-focused collaboration in 2024 with the ultimate goal of creating new AI chips.
According to industry sources, OpenAI’s design is capable of both training and running AI models. However, the company will initially only use it for AI inferencing. OpenAI expects the final chip to have a certain amount high-bandwidth RAM, just like any other major AI or GPU silicon design. OpenAI’s chip, despite playing a minor part in the company infrastructure for the next couple of months, could be a disruptive force in the future. Ho’s team must fix any hardware bugs found during the initial manufacturing testing.
While many tech companies are actively working on replacing Nvidia products with custom solutions for AI accelerators, the GPU maker still controls around 80 percent of market share. Microsoft, Google Meta, and other Big Tech companies are employing hundreds engineers to solve the problem of silicon, while OpenAI is last in both timing and workforce size.
OpenAI, led by Richard Ho, will need more than the small team of in-house engineers currently working on its AI prototype. The chip project is viewed as a vital tool for future strategic moves within the growing AI industry. OpenAI engineers have already begun planning for more advanced iterations of the chip while they wait for design validation by TSMC.
