OpenAI's Self-Developed Chips Expected to Enter Production Next Year

According to Reuters, OpenAI is advancing with its plans to produce self-developed AI chips, with production expected to begin next year. Insiders have revealed that OpenAI will finalize the chip design in the coming months and then submit it to Taiwan Semiconductor Manufacturing Company (TSMC) for manufacturing.

By developing its own chips, OpenAI aims to reduce reliance on NVIDIA chips for training and running AI models. TSMC is reportedly set to manufacture these chips using advanced 3-nanometer technology, integrating "high-bandwidth memory" and "powerful networking capabilities."

Reuters noted that during the initial phase of the chip's release, OpenAI will deploy the self-developed chips within a "limited scope," primarily for operating AI models. Additionally, OpenAI is planning to develop future versions of the chip, featuring more advanced processors and functionalities.

This news follows last year's Reuters report, which indicated that OpenAI was collaborating with Broadcom to develop custom chips. The chip design team at OpenAI is led by former Google TPU engineer Richard Ho, and its size has expanded from 20 members to 40 over recent months, indicating increased R&D efforts.

Tech giants like OpenAI are heavily investing in AI infrastructure and chip procurement to meet the demands of their data-intensive AI models. Although AI startup DeepSeek has questioned the necessity of purchasing thousands of chips to support AI systems, spending in this field shows no signs of slowing down.