OpenAI Develops Its First Custom AI Chip | OpenAI italiano | ChatGPT login | OpenAI API | Turtles AI

OpenAI is set to finalize the design of its first custom AI chip, aiming to reduce dependence on outside chip suppliers such as Nvidia. The process, which will culminate in production at TSMC, includes an initial testing phase and limited deployment. The ambitious plan calls for mass production by 2026.

Key points:

• OpenAI develops its custom AI chip to reduce dependence on Nvidia.
• The design will be completed within a few months, with the production phase scheduled for 2026.
• The chip will have an initially limited role in running AI models.
• The project is being led by Richard Ho, with input from Broadcom and manufacturing by TSMC.

OpenAI is accelerating its plans to design a custom chip for AI, with the goal of decreasing its dependence on Nvidia, a dominant AI chip supplier. By the end of the year, the company is expected to complete the silicon design, which will then be sent to Taiwan Semiconductor Manufacturing Company (TSMC) for the production phase, known as “tapping out.” This is an important step in the chip development process, but one with high costs and the uncertainty of possible failure, which would require new iterations of the design. Despite the complexity and associated risks, OpenAI seems on track to complete the project on schedule, with mass production of the chip planned for 2026. The chip, developed by the in-house team led by Richard Ho, will initially have a limited role, dedicated primarily to running AI models, although it is designed to perform training tasks as well. Collaboration with Broadcom helped bring it to fruition, with TSMC’s manufacturing infrastructure relying on three-nanometer technology and a systolic array architecture with high-bandwidth memory, similar to that used by Nvidia chips. Although OpenAI is pursuing an ambitious agenda, competition with other big tech companies such as Google and Meta, which are also investing substantial resources in their own AI chips, makes the challenge particularly interesting. However, OpenAI’s innovation and progress to date could prove crucial to the future evolution of the AI chip market.

With production likely to begin at full capacity in 2026, OpenAI is marking a significant step in the world of advanced technology, with a desire to reduce dependence on external players and pursue self-sufficiency in AI infrastructure.