TAIPEI (Taiwan News) —Taiwan Semiconductor Manufacturing Co. (TSMC) has been benefiting from the intensified focus on artificial intelligence (AI) and high-performance computing (HPC), according to DigiTimes.

TSMC is responsible for manufacturing some of the most complex processors for AI and HPC machines, including those from Nvidia, AMD, Intel, Tenstorrent, Cerebras, and Graphcore, DigiTimes noted. The Taiwan chipmaker produces Nvidia’s A100 and H100, the most popular compute GPUs employed for AI and HPC workloads, in addition to AMD’s EPYC CPUs and Instinct GPUs, according to Tom’s Hardware.

The world’s largest contract chipmaker is especially dominant in the shipment of AI GPUs, because it manufactures them both for Nvidia, which controls over 90% of shipments, and AMD, which is responsible for less than 10%, per Tom’s Hardware.

While TSMC clearly distinguishes between automotive, smartphone, IoT, and HPC revenues, it does not provide enough information to tell the difference between semiconductors for AI, HPC, client PCs, servers, and game consoles, Tom’s Hardware said. TSMC considers all of these chips related to the HPC segment.

In 2019, HPC products made up 30% (US$10.39 billion/NT$322 billion) of TSMC’s revenue. Since then, it has been steadily increasing to 33% (NT$464 billion) in 2020, 37% (NT$650 billion) in 2021, and 41% (NT$963 billion) in 2022, according to Tom’s Hardware.

Despite producing so many AI and HPC related chips for Nvidia and AMD, Apple still remains TSMC’s biggest client. Apple made up 23% of the Taiwan chipmaker’s total sales in 2022, DigiTimes said.

Looking ahead, Apple, Nvidia, and AMD have committed to using TSMC’s 3nm process and future 2nm process silicon for future chips, per DigiTimes.