- Big Tech is tired of trusting Nvidia, so they are doing their own chips
- Openai’s first GPU is almost ready, heading to TSMC to record
- If everything goes well, mass production could start in 2026
The increase in costs and a worrying dependence on the NVIDIA giant have led technological giants such as Microsoft, Google and goal to consider the construction of their own artificial intelligence chips.
Openai, who participates in the Stargate initiative of $ 500 billion recently announced, is also developing its own AI hardware to give it some freedom of Team Green.
A report from PakGazette He states that the company is in the final stages of producing this first chip, which could have cost Openai more than $ 500 million to design, and hopes to send it for the manufacture in TSMC in the coming months, with a mass production that probably It will begin in 2026.
Deployed on a limited scale
Openai chips development has been in progress for a while. We reported for the first time in July 2024 that Sam Altman’s company was in conversations with Broadcom to design and build his own silicon and, more recently, that the AI signature was approaching this reality.
The report states: “If the initial cap goes well, it would allow the Chatgpt manufacturer to produce its first internal and potentially chip chip to try an alternative to the Nvidia chips at the end of this year. Openai’s plan to send its design to TSMC this year demonstrates that the startup has made rapid progress in its first design, a process that can lead to other chips designers years longer. ”
Directed by Richard Ho, who joined Openai more than a year ago and previously played a key role in the development of Google custom processors, the team that develops the chip is relatively small, and consists of only 40 engineers.
While the internal ai chip will be able to training and inference tasks, PakGazette“The sources say that it will initially be” implemented on a limited scale, and mainly to execute AI models. ” It also will have a limited role within the company’s infrastructure. According to the media, Operai sees his personalized AI chip as a way to improve his negotiation position with existing suppliers, including Nvidia.
The chip is taking place using the Advanced TSMC process and will have a commonly used systolic matrix architecture, HBM and advanced networks.
Openai and TSMC declined to comment on the report.
