Company is finalizing the design of the 1st internal chip, aiming to reduce dependence on NVIDIA
OpenAi is in the final phase of development of its 1st Own Artificial Intelligence Processor (AI), scheduled for release at the end of 2025. The project, led by the former Google Executive Richard Ho, will be manufactured by Taiwan Semiconductor Manufacturing Co. (TSMC) and is part of the company’s strategy to reduce Nvidia’s technological dependence. The information was released by.
The chip design will be sent for manufacturing in the coming months, in a process known as the 1st version of the project. Initial production has an estimated cost of millions of dollars and a period of completion of 6 months.
The development team, which has grown to 40 engineers in recent months, works in partnership with Broadcom. The processor will use TSMC 3 nanome technology and will feature systolic matrix architecture, high bandwidth memory and advanced network features – similar to Nvidia chips.
OpenAi plans to start production on a large scale by 2026. The new processor will be initially used on a limited scale, focusing on the execution of AI models. The company intends to develop more advanced versions after the first tests.
The project is inserted in a scenario of increasing investments in AI infrastructure. The goal announced an investment from $ 60 billion to 2025, while Microsoft foresees to allocate US $ 80 billion for the area in the same period.
The initiative is also part of the Stargate program, a $ 500 billion project announced in January that brings together OpenAi, Oracle and Softbank for AI infrastructure development in the United States.
