Anthropic, the prominent U.S.-based artificial intelligence company, has reportedly finalized supply agreements with South Korean semiconductor giants Samsung Electronics and SK Hynix. The announcement was made by Anthropic CEO Dario Amodei at an AI summit in South Korea, marking a significant development in the ongoing competition for advanced AI infrastructure.
While the deals signal a major step in Anthropic's computing strategy, specific details regarding the components or services to be supplied, the overall deal value, or the duration of the agreements have not been publicly disclosed. Both Samsung and SK Hynix are key players in the semiconductor industry, offering distinct strengths in memory and logic chip technologies.
This move comes after Anthropic's significant funding round in May 2026, which valued the company at $965 billion. At that time, Anthropic had already named Samsung, SK Hynix, and Micron as strategic technology partners supporting its AI infrastructure. It remains unconfirmed whether the latest agreements specifically include High Bandwidth Memory (HBM) chips, other memory components, storage products, or logic chips.
SK Group Chairman Chey Tae-won also hinted that Anthropic is developing its own "computing power," a statement that fuels speculation about Anthropic's potential foray into designing proprietary AI chips. While Anthropic has not confirmed dedicated chip-designing efforts, the company has stated its continuous exploration of various computing options, maintaining a diversified approach by utilizing chips from multiple suppliers rather than relying on a single provider.
The pursuit of custom AI chips is a growing trend among leading AI companies. Competitors such as OpenAI, Google, and Amazon AWS are also aggressively developing in-house silicon to decrease their dependence on dominant suppliers like Nvidia. OpenAI, for instance, has partnered with semiconductor designers like Broadcom, while Google continues to advance its proprietary Tensor Processing Units (TPUs), and Amazon scales its custom chip division with offerings like Trainium for training and Inferentia for inference tasks.