TSMC, the world’s largest foundry company, has decided to invest up to $1 billion (about 1.34 trillion won) in OpenAI and Anthropic, which are developing their own AI semiconductors. As the two AI companies pursue developing their own chips to reduce their dependence on external semiconductor makers such as Nvidia, TSMC’s move is also seen as an effort to strengthen ties with them by securing equity stakes, going beyond simply supplying semiconductors.
On the 10th, TSMC’s board of directors approved up to $500 million (about 668 billion won) for the purchase of Anthropic shares and up to $500 million for an investment in OpenAI preferred stock. The Anthropic investment involves acquiring common shares based on the IPO offering price, while the OpenAI investment is structured as an acquisition of preferred stock.
The decision came as Anthropic proceeded with preparations for a stock listing, proposing an indicative IPO price of $70–$78 per share and targeting a valuation of up to $1.4 trillion. Anthropic is also separately raising $26 billion in private investment through a group of investors that includes Google, Nvidia, AMD and Microsoft.
TSMC is a key company that manufactures cutting-edge chips for major AI semiconductor makers, including Nvidia. Recently, as AI model developers have begun designing semiconductors themselves, the need to cooperate with TSMC has been growing.
OpenAI unveiled its own AI inference chip, “Hallucino,” co-developed with Broadcom last June. The strategy is to build its own AI infrastructure, from semiconductor design through manufacturing, to reduce its dependence on Nvidia GPUs.
Anthropic also announced last August that it was forming a design team to develop its own semiconductor optimized for Claude. It currently uses Nvidia, AMD, Google and Amazon semiconductors together, and plans to improve computational efficiency through its own chip.
Accordingly, TSMC has secured an opportunity to build direct business relationships not only with existing semiconductor designers but also with AI model developers. In particular, in a situation where cutting-edge semiconductor production capacity is limited, TSMC is an important partner even for AI companies developing their own chips.
This investment can be seen as a strategy to leverage its position in the supply chain to strengthen relationships with major AI companies and secure not only semiconductor manufacturing profits but also investment returns as those companies grow.
Results are also confirming the strength of demand for AI semiconductors. TSMC announced last month that it had posted record profits for the sixth consecutive quarter, saying demand for AI-related semiconductors was “extremely strong.” It forecast annual revenue for 2026 to grow by approximately 44%, explaining that demand was outstripping production capacity.
TSMC plans to invest a total of $265 billion in semiconductor production facilities in Arizona and is also considering additional business in Texas. As large-scale investment in AI infrastructure continues, TSMC is expanding its business scope while increasing production capacity and securing stakes in major AI developers.
Reporter Park Chan cpark@aitimes.com
