What happened
XCENA, a four-year-old startup with offices in South Korea and the U.S., secured $135 million in Series B funding, valuing the company at $570 million and bringing its total raised to $185 million. The company developed the MX1 chip, designed to place compute capabilities closer to DRAM, processing data within the memory module via CXL. This approach aims to reduce costly data round trips between CPUs, GPUs, and memory, potentially allowing what required 10 servers to run on one. Mass production of the RISC-V based MX1 chips is scheduled for late 2026, with revenue expected in 2027.
Why it matters
AI infrastructure costs could significantly reduce for hyperscalers, as XCENA's MX1 chip addresses the memory bottleneck by handling data orchestration tasks directly within memory. This mechanism, which brings compute to data via CXL, promises to cut server requirements. This follows recent reports highlighting memory as a dominant AI chip component cost.
