OpenAI’s First In-House Chip Revealed, Manufactured by TSMC, with Apple Already Placing Orders

New information has surfaced about OpenAI’s first internally developed AI chip. According to reports, TSMC will develop a custom A16 Ångström-level process chip specifically designed for OpenAI's Sora video model, aiming to enhance Sora's video generation capabilities.

This A16 process node represents the most advanced technology publicly disclosed by TSMC and marks the company's debut into the Ångström process era, with mass production expected to begin in the second half of 2026.

The A16 will reportedly utilize next-generation nanosheet transistor technology and employ Super Power Rail (SPR) technology. SPR is an innovative, industry-leading backside power delivery solution and the first of its kind. By moving the power delivery network to the backside of the wafer, SPR frees up more space on the front side, significantly increasing logic density and performance. This makes the A16 suitable for high-performance computing (HPC) products that require complex signal routing and dense power networks.

Compared to the N2P process, the A16 offers a 1.10x improvement in chip density, an 8% to 10% increase in speed at the same operating voltage, and a 15% to 20% reduction in power consumption at the same speed.

Although TSMC’s A16 process has not yet entered mass production, the first batch of customers has already been revealed. Reports indicate that Apple has pre-booked the initial production capacity, and OpenAI has also reserved A16 capacity to meet its own AI chip manufacturing needs.