XCENA Raises $135M Series B for Memory-Centric Computing
Semiconductor startup XCENA closed a $135 million Series B funding round, bringing its total to $185 million and valuing the company at $570 million post-money, with backing from South Korean firms Atinum Investment and IMM Investment. The company’s MX1 architecture, which integrates memory arrays directly with computational resources to address the von Neumann bottleneck, is in prototype stage and targets AI workloads by reducing data transfer inefficiencies.
Semiconductor startup XCENA has secured $135 million in Series B funding, co-led by South Korean venture capital firms Atinum Investment and IMM Investment, alongside a syndicate of institutional investors including SBI Investment and the Korea Development Bank. The round brings total funding to $185 million and a post-money valuation of $570 million, positioning XCENA as a leader in memory-centric computing—a field focused on integrating large memory arrays directly with computational resources. The company’s flagship MX1 architecture uses the open CXL 3.x standard to link directly with host processors, embedding thousands of proprietary RISC-V cores and vector engines within pooled DDR5 memory. This design eliminates the need for constant data transfer between memory and processors, a bottleneck known as the von Neumann bottleneck, which limits AI performance. Preliminary tests show MX1 can deliver results comparable to multiple high-end servers while reducing infrastructure demands. XCENA’s approach shifts computational tasks like preprocessing and key-value cache management into memory modules, offloading pressure from primary server chips. To ease adoption, the startup offers XFLARE, a full-stack SDK enabling enterprises to migrate workflows to memory-centric hardware with minimal software changes. Founded in 2022 by semiconductor veterans from Samsung Electronics and SK Hynix, XCENA operates from R&D centers in Sunnyvale, California, and Banqiao, South Korea, employing over 90 engineers. The Series B funding will accelerate commercialization, with plans to expand in Northern California to collaborate with hyperscale cloud providers. The MX1 prototype is currently under evaluation with early partners, targeting long-term deployment in AI-driven industries. The funding reflects growing demand for hardware solutions that address AI’s memory constraints, as generative AI workloads increasingly outpace traditional processor-based systems. XCENA’s focus on memory-centric design aims to redefine efficiency in data-intensive applications, aligning with broader trends in AI infrastructure optimization.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.