Samsung eyes billions in AI chip orders from Meta and Anthropic amid the AI chip race

Samsung Electronics appears to be making a major move in the global AI chip race. According to South Korean business daily Seoul Economic Daily, Samsung’s foundry division is in talks to manufacture new AI chips for both Meta and Anthropic. If the deals land, Samsung could secure tens of billions of dollars in additional orders and significantly strengthen its position against market leader TSMC.

Industry sources say Meta is developing the third generation of its in-house AI accelerator, the Meta Training and Inference Accelerator (MTIA). While the first two generations were built by TSMC, Samsung Foundry is now poised to become the primary manufacturing partner.

The new chip is expected to use Samsung’s advanced 2-nanometer process and be produced at scale. South Korean media value the project at over 10 trillion won—more than $6.5 billion. Samsung stresses no final agreement has been reached. Meta is investing heavily in its own AI infrastructure, targeting data centers totaling around 5 gigawatts by 2030. By developing its own AI processors, Meta aims to reduce reliance on Nvidia while cutting the cost of AI training and inference.

It’s not just Meta looking at Samsung. AI company Anthropic is reportedly exploring the development of a custom AI chip that Samsung could manufacture.

Talks are still early. Anthropic would use the chips to lessen its dependence on Nvidia GPUs and run its AI systems more efficiently. The company has also recently hired chip designers who previously worked on similar efforts at OpenAI.

The interest comes at a pivotal moment for Samsung. Its foundry unit has been trying to claw back share from TSMC, currently the dominant producer of cutting-edge AI processors.

Samsung has already landed a sizable AI chip deal with Tesla. Reuters also reported last month that Google is in talks with Samsung over parts of a future AI chip. The trend suggests more tech giants want to diversify production across multiple suppliers instead of relying solely on TSMC.

More AI firms are investing in custom silicon instead of relying exclusively on Nvidia hardware.

Key reasons include:

Google, Amazon, Microsoft, and Meta all develop their own AI processors. Anthropic appears to be following suit, though its plans aren’t final.

If Samsung secures both Meta and Anthropic, it would mark a significant breakthrough for the South Korean chipmaker. Beyond memory, Samsung could more forcefully position itself as a producer of advanced AI accelerators.

That would intensify competition with TSMC while making the global AI infrastructure less reliant on a single manufacturer. As demand for AI chips continues to surge, spreading production could become increasingly strategic for both tech companies and governments.