XCENA raises $135M to fix AI's memory bottleneck

xcena, a startup founded by samsung and sk hynix veterans, secured $135m in series b funding at a $570m valuation to tackle ai inference bottlenecks caused by data movement between cpus, gpus, and memory. its mx1 chip uses compute express link (cxl) to process data like kv cache management directly in memory, claiming a 10x reduction in server needs compared to traditional architectures. mass production is slated for late 2026 via samsung foundry, targeting hyperscalers looking to offset rising ai infrastructure costs. the company differentiates itself from rivals like astera labs and marvell by using thousands of risc-v cores integrated into memory modules.