The company also added AMD chips along with Nvidia to meet soaring infrastructure demand, according to Reuters.
Specifically, OpenAI, the fast-growing company behind ChatGPT, considered a range of options to diversify its chip supply and reduce costs. OpenAI considered building everything in-house and raising funds for an expensive plan to build a network of factories called “foundries” to manufacture the chips.
OpenAI builds first chip with Broadcom and TSMC. (Illustration photo)
The company has now scrapped ambitious foundry construction plans due to the cost and time required to build the network, instead focusing on in-house chip design efforts.
The company’s strategy leverages industry partnerships and approaches to secure chip supplies and manage costs, like major rivals Amazon, Meta, Google, and Microsoft. As one of the largest chip buyers, OpenAI’s decision to source from multiple chipmakers while developing its own custom chips could have broader implications for the tech sector.
OpenAI, a company that has emerged as a leader in helping commercialize AI that generates human-like responses to queries, relies on significant computing power to train and run its systems. As one of the largest buyers of Nvidia’s graphics processing units (GPUs), OpenAI uses AI chips to train models in which AI learns from data and to make inferences, applying AI to make predictions or decisions based on new information.
Reuters previously reported on OpenAI’s chip design efforts. The Information reported on the company’s talks with Broadcom and others. OpenAI has been working with Broadcom for months to build its first AI chip focused on inference, according to the sources. There is currently greater demand for training chips, but analysts predict that demand for inference chips could surpass them as more AI applications are deployed.
Broadcom has helped companies including Alphabet refine chip designs for manufacturing, and it also provides design components that help move information in and out of chips quickly, which is crucial in AI systems where tens of thousands of chips are strung together to work in parallel.
OpenAI is still determining whether to develop or acquire other components for its chip design and may bring in additional partners, two sources said. The company has assembled a chip team of about 20 people, led by former top engineers who built the Tensor Processing Unit (TPU) at Google, including Thomas Norrie and Richard Ho.
Through Broadcom, OpenAI secured manufacturing capacity with Taiwan Semiconductor Manufacturing Company (TSMC) to produce its first custom chips by 2026, the sources said. The timeline is subject to change, they said.
Nvidia’s GPUs currently account for more than 80% of the market. But shortages and rising costs have forced major customers like Microsoft, Meta, and OpenAI to look for in-house or external alternatives.
Source
Comment (0)