XCENA raised $135M to bring compute closer to memory and cut AI costs

Friends, a quick AI update: startup XCENA closed a $135M Series B at a $570M valuation.

The company is developing the MX1 chip, which brings computation closer to DRAM via CXL to execute operations inside memory modules, reducing costly CPU–GPU data movement.

Founders are veterans of Samsung and SK Hynix; the round was led by Altinum and IMM, bringing total funding to $185M.

MX1 is currently a prototype; mass production at Samsung’s fabs is planned for late 2026, with revenue expected from 2027.

Why it matters: a shift to memory‑centric architectures could materially lower AI deployment costs and energy consumption.

Could this approach reshape the economics of training and serving large models?

#AI #hardware #memory #startups