tech
This chip startup just raised $135M on a bet that AI's biggest bottleneck isn't compute -- it's memory
South Korean chip startup XCENA is betting that AI's real bottleneck is not compute, but memory.
TL;DR
- XCENA is a startup aiming to solve the inefficiency and high cost of AI data processing by placing compute capabilities closer to memory (DRAM).
- The company's new chip, the MX1, processes data directly within the memory module, reducing the need for data to travel between CPUs, GPUs, and memory.
- XCENA raised $135 million in a Series B funding round, bringing its total funding to $185 million.
- The startup's founders are veterans of Samsung and SK Hynix, memory giants.
- XCENA's thesis is that AI inference is increasingly a memory scaling problem, not just a compute problem.
- The MX1 chip connects to the CPU via CXL and handles tasks like preprocessing and KV cache management directly within the memory module.
- The company expects mass production of the MX1 chip by the end of 2026, with revenue generation starting in 2027.
- XCENA's chip design uses thousands of small, efficient RISC-V cores and a high degree of vertical integration.