XCENA raises $135M at $570M valuation for memory-centric AI chips
In brief
- XCENA raised $135M Series B at $570M valuation, bringing total funding to $185M
- MX1 chip processes compute closer to DRAM, reducing data transfers between hardware
- Founders are former Samsung and SK Hynix executives; production chips planned for late 2026
- XCENA claims architecture could reduce infrastructure from 10 servers to one
- Company expects revenue to begin in 2027
The memory bottleneck problem
Every AI query requires data to move repeatedly between memory, CPUs, and GPUs. This constant shuttling creates latency and consumes power. XCENA's approach is to move computation closer to where data lives. The MX1 connects to CPUs using CXL, a high-speed interface linking processors and memory, and performs compute tasks directly inside the memory module itself.
This design allows routine operations—preprocessing, KV cache management, and data caching—to bypass CPUs entirely. The result is less data movement, lower latency, and fewer servers needed to handle the same workload.
The founding team and competitive landscape
XCENA was founded in 2022 by CEO Jin Kim, CTO Dohun Kim, and CPO Harry Juhyun Kim, all former executives at Samsung and SK Hynix, two of the world's largest memory chip makers. That pedigree shows in their technical approach. XCENA says its competitive edge comes from its proprietary architecture, including thousands of small RISC-V cores and in-house designs for memory hierarchy, interconnects, and DRAM controllers.
The startup is positioning itself against established players like Astera Labs and Marvell in the emerging memory-centric AI infrastructure market. Unlike those competitors, XCENA built its stack from the ground up around memory-first design rather than adapting existing CPU or GPU architectures.
Timeline and scale claims
The MX1 remains in prototype stage. XCENA plans to manufacture production chips through Samsung's foundry operations by the end of 2026, with revenue expected to begin in 2027. The company has already begun discussions with global memory vendors about integrating its technology.
On the efficiency front, XCENA makes an ambitious claim. The startup says its architecture could sharply reduce infrastructure requirements, potentially cutting workloads that currently require 10 servers down to a single server. If validated at scale, that kind of consolidation would reshape how AI providers build and cost their inference infrastructure.
The Series B also included participation from Corstone Asia, SBI Investment, and Mirae Asset Capital.
