According to a recent report from the Economic Daily News, OpenAI has already booked production capacity on TSMC's Angstrom-level A16 chips. TSMC is scheduled to begin volume production of its A16 process in 2026, which would be a major leap in transistor density and performance compared to today's chips.
While TSMC did not comment on customer-specific relationships, the report says OpenAI becomes the second early customer for A16, joining Apple. That's a strong commitment by OpenAI to high-performance, custom AI chips.
Securing capacity on the A16 shows just how seriously OpenAI is taking AI hardware. If rumors are to be believed, this chip was tailor-made for improving video generation within OpenAI's Sora model. It wasn't that long ago that there were reports that OpenAI considered building its own fabrication facility before pulling back to stick to chip design with outsourced production.
OpenAI is working with chipmakers Broadcom and Marvell to develop application-specific integrated circuits (ASICs) for AI work. These ASICs are expected to be manufactured on TSMC's 3nm process family in 2024 or 2025 before moving to the A16 node. As one of Broadcom and Marvell's largest customers, OpenAI will be able to leverage their experience.
On the other hand, Apple has long experience developing high-performance silicon for its devices. A partnership with OpenAI would position Apple to be an early mover in AI technologies, enhancing core experiences such as photo and video and augmented reality. OpenAI, on its part, would gain chip design savvy and manufacturing clout as a strategic partner of TSMC.
This also comes after Apple integrated ChatGPT into its new Apple Intelligence platform, another sign that OpenAI's influence is beginning to extend well beyond its own products. By reserving capacity on the industry's most advanced process technologies, OpenAI is poised to remain at the forefront of AI development.
Source: Economic Daily News via Wccftech
1 Comment - Add comment