TSMC Announces Apple and OpenAI as Initial Clients for A16 Chip Production

Today, TSMC announced that its A16 manufacturing process, set for mass production in the second half of 2026, has successfully attracted its first customers, including long-time partner Apple and AI frontrunner OpenAI. This marks a significant breakthrough for TSMC in advanced process technology and signals new development opportunities in the AI chip industry.

The A16 process utilizes innovative Nanosheet Gate-All-Around (GAA) transistor technology, which enhances chip performance and energy efficiency while improving integration and reliability. Notably, this is TSMC's first process to implement the Super Power Rail backend power solution, optimizing power delivery while retaining the benefits of traditional frontend power. This makes it especially suitable for high-performance computing (HPC) products that demand exceptional performance.

As a longstanding partner, Apple has reserved the initial capacity of the A16 process to bolster performance in its upcoming products. OpenAI's involvement is particularly noteworthy; as a leader in the AI sector, its pre-orders for AI chips demonstrate strong confidence in the A16 technology. OpenAI is also collaborating with companies like Broadcom and Marvell, positioning itself to become one of Broadcom's top four clients, further solidifying its place in the AI chip market.

Reports from Taiwanese media suggest that OpenAI’s custom AI chips will be produced using TSMC's 3nm series and the upcoming A16 process. This decision highlights OpenAI's urgent need for high-performance AI chips and underlines TSMC's leadership in advanced process technology.

With the rapid advancement of AI technology, AI chips are becoming increasingly critical as foundational components for AI applications. The launch of TSMC's A16 process, along with its endorsements from Apple and OpenAI, will undoubtedly inject new vitality into the AI chip industry, driving the entire sector towards greater efficiency and intelligence.