On September 12, during the 2026 Computing Power China AI and Computing Power Industry Exhibition, Lei Kun, the founder and CEO of Entropy Xuanxin Zhilei, announced a groundbreaking development. He revealed that the in-memory computing architecture leveraging MRAM (Magnetoresistive Random-Access Memory) exhibits an energy consumption rate nearly 10 times lower compared to traditional SRAM (Static Random-Access Memory). This significant reduction in energy usage translates directly into a substantial decrease—approximately 10 times—in the cost of AI inference tokens.

Furthermore, Lei Kun highlighted that Samsung's 8-nanometer MRAM technology has already reached a mature stage in mass production, with plans to advance to an even more refined 5-nanometer process in the coming year. This progression underscores the rapid evolution and increasing viability of MRAM in high-performance computing applications.

Entropy Xuanxin Zhilei, riding on the wave of these technological advancements, has successfully secured two rounds of financing. The company is now on the brink of finalizing its third round, signaling strong investor confidence and a promising future ahead.