South Korean chip startup XCENA has raised $135 million in a Series B at a $570 million valuation, bringing total funding to $185 million. The round was co-led by Seoul-based VC firms Atinum and IMM Investment, alongside Corstone Asia and existing backers SBI Investment and Mirae Asset Capital.

Founded in 2022 by veterans of Samsung and SK Hynix, XCENA is tackling a structural inefficiency at the heart of AI inference: every word a model generates requires data to travel repeatedly between memory, CPUs, and GPUs—a costly, power-intensive relay race. The company's solution is its MX1 chip, which places compute capabilities directly inside the memory module, handling routine data operations near DRAM without expensive round trips between processors.

Connected to CPUs via CXL—a dedicated high-speed lane between processor and memory—the MX1 manages preprocessing, KV cache management, and data caching entirely within the memory module itself. XCENA claims the technology could consolidate what currently requires 10 servers into a single machine.

"Inference isn't just a compute problem; it's increasingly a memory scaling problem," said CEO Jin Kim, noting that memory prices and related stocks reflect a broader industry shift toward memory-centric AI architectures.

The MX1 remains a prototype, with mass production through Samsung's foundry scheduled for late 2026 and revenue expected from 2027. XCENA's target customers are hyperscalers where even marginal memory efficiency gains translate into hundreds of millions in savings.