TSMC to Begin Mass Production of 1.6nm Chips in 2026, Outpacing Intel and Samsung

Taiwan Semiconductor Manufacturing Company (TSMC), the world’s leading chipmaker, is set to begin mass production of chips using its A16 Angstrom process node, equivalent to 1.6 nanometers, in 2026. This development will position TSMC ahead of its competitors, Intel and Samsung, who plan to start producing chips using their 1.4nm processes—Intel’s 14A and Samsung’s SF14—in 2027.

One of the most striking aspects of TSMC’s A16 Angstrom node is that it does not require the use of ASML’s cutting-edge High-NA EUV lithography machines, which are critical for the production of smaller nodes. These machines, costing over $380 million each, are vital to the operations of Intel and Samsung, but TSMC has managed to bypass this requirement, potentially giving them a significant cost advantage.

Apple and OpenAI: Early Adopters of TSMC's A16 Angstrom Node

Apple has already shown confidence in TSMC’s next-generation technology by placing orders for chips built on the A16 Angstrom node. This move is likely part of Apple’s broader strategy to integrate more advanced AI and processing capabilities into its devices, particularly iPhones, which could see a performance boost as a result.

Additionally, there are reports that OpenAI, the organization behind the popular AI model ChatGPT, is developing its first in-house AI chip using TSMC’s A16 Angstrom technology. OpenAI’s interest in this advanced node suggests the company is looking to enhance the performance of its AI models, including features like the Sora video-generation tool, which may become a critical component of Apple’s AI suite.

OpenAI has even gone as far as requesting TSMC to build a dedicated fabrication facility (fab) specifically for their needs.

TSMC’s Technological Edge and the AI Market

TSMC’s A16 Angstrom node is expected to outperform the 2nm chips preceding it, delivering 8-10% faster performance while consuming 20% less power at the same operating voltage. This improvement is crucial, especially for AI applications, where efficiency and speed are paramount. As AI usage grows, so too does the demand for high-performance chips. For instance, each query submitted to ChatGPT currently costs OpenAI approximately 4 cents.

If ChatGPT were to scale up to handle just one-tenth of Google Search’s traffic, OpenAI would require an estimated $16 billion worth of chips annually. This massive demand highlights the lucrative opportunity for TSMC as it pushes forward with its A16 Angstrom node.

ByteDance's Chip Ambitions Amid U.S. Sanctions

Meanwhile, ByteDance, the parent company of TikTok, is also making strategic moves in the AI chip market. Earlier this year, rumors circulated that ByteDance was working with Broadcom to develop an AI chip using TSMC’s 5nm process. Despite the U.S. sanctions that restrict the export of advanced chip technology to China, ByteDance appears to have found a loophole allowing them to produce customized application-specific integrated chips (ASICs) through TSMC.

ByteDance plans to utilize its in-house chips to power advanced AI algorithms on platforms like TikTok and its Chinese counterpart, Douyin. The company also operates Doubao, a chatbot in China, which could benefit from these custom chips. In the interim, ByteDance still relies on an inventory of NVIDIA chips acquired before the sanctions took effect, ensuring continued development of its AI capabilities.

Potential Collaboration Between Apple and OpenAI

There is speculation about a possible collaboration between Apple and OpenAI to develop a chip tailored for AI-based servers hosted in the cloud. Such a partnership could significantly enhance Apple’s generative AI features, potentially integrating tools like Sora into iPhones and other devices, thereby boosting performance and user experience.

Via