OpenAI is finalizing the design for its first in-house chip in the next few months and is looking to send it to Taiwan Semiconductor Manufacturing Co (TSMC) for fabrication, reported Reuters. The first generation AI chip manufacturing effort by ChatGPT maker is aimed at reducing the reliance on Nvidia.
Reportedly, the process of sending first designs to a chip factory for fabrication is called ‘taping out’. A tape out process can take around six months to complete and incur tens on millions in costs. Additionally, there is no guarantee that the chip will function as expected in the first tape out and if there is a problem then OpenAI will have to diagnose it and repeat the whole process.
What will OpenAI do with its first-generation AI chip?
If the initial tape-out goes smoothly, OpenAI will reportedly be able to mass-produce its AI chips by 2026. Reportedly, OpenAI’s AI chip is being built on TSMC’s 3 nanometer process technology and also uses a commonly used systolic array architecture with high-bandwidth memory which is also employed by Nvidia for its chips.
The AI chip will be capable of both training and running AI models but it will only be deployed on a limited scale and majorly for running AI models. Intially, the chip is likely to have play a limited role withing OpenAI’s infrasctructure.
Why is OpenAI building its own AI chip?
Generative AI chatbots like ChatGPT, Gemini and Meta AI have traditionally required large number of chips to train their foundational models. The powerful chips required for these operations are currently mostly supplied by Nvidia whcih holds around 80% of the market share.
The effort by OpenAI to build its own chip will likely allow the company to strengthen its negotiating leverage with chip suppliers including Nvidia. Moreover, after the success of its first chip, OpenAI is reportedly planning on develping more advanced processors with capabilties increasing each generation.
Is OpenAI ahead of other big tech?
OpenAI’s plans to send its first chip design to TSMC later this year would make it stand ahead of the curve, with other chip designers taking years longer. Notably, other big tech companies like Satya Nadella led Microsoft and Mark Zuckerberg led Meta have failed to produce satisfactory chipset even after years of trying.
Meanwhile, the recent rise of DeepSeek, which is built at a fraction of the cost and computing power of ChatGPT and other western AI chatbots, suggests that fewer chips could be required in the future for developing powerful large language models.
