Korean startup DeepX will manufacture its next-generation DX-M2 AI processor using Samsung Foundry’s 2-nanometer process. Production is scheduled to begin next year. The DX-M2 processor delivers 80 trillion operations per second. The chip consumes less than 5 watts of power.
The partnership grants DeepX access to advanced manufacturing nodes to compete in the low-power, on-device AI and large language model inference markets. This move directly challenges TSMC’s global foundry leadership.
DeepX is targeting the physical AI sector, including robotics, smart mobility, and intelligent cameras. The company projects this market will eventually surpass the data center industry in size. For investors, the deal signals a diversifying competitive landscape beyond dominant semiconductor firms.