[Entropy Core Intelligence Lei Kun: MRAM integrated storage and computing chips can reduce the cost of AI inference tokens by about 10 times] On September 12, Lei Kun, founder and CEO of Entropy Core Intelligence, said at the 2026 Computing Power China AI and Computing Power Industry Exhibition Science and Technology Roadshow that the MRAM-based integrated storage and computing architecture consumes nearly 10 times less energy than SRAM and can reduce the cost of AI reasoning tokens by about 10 times. Samsung's 8nm MRAM has matured for mass production and will advance to 5nm next year. At present, the company has completed two rounds of financing (Yingnuo, Haibei, Junshan), and the third round of financing is nearing completion. (Reporter Guo Songqiao)