Key Takeaways
- • XCENA, a South Korean chip startup, secured $135 million in Series B funding co-led by Atinum Investment and IMM Investment, valuing the company at $570 million and bringing total capital raised to $185 million.
- • The company's MX1 chip addresses a fundamental architectural bottleneck in AI infrastructure by placing compute capabilities closer to DRAM, eliminating costly round-trip data transfers between CPUs, GPUs, and memory during AI model inference and generation.
- • XCENA's memory-centric computing approach targets the memory inefficiency problem that has persisted while compute power has advanced, with CEO Jin Kim noting that "CPUs and GPUs have both gotten smarter over the decades. Memory never did."
- • The funding will scale customer deployments, expand go-to-market efforts, and accelerate innovation in CXL-based computational memory solutions designed for enterprise AI infrastructure, positioning XCENA in an increasingly competitive South Korean AI semiconductor market.
- • The valuation and investor appetite reflect broader market recognition that memory architecture—not raw compute—represents the next critical optimization layer for reducing AI infrastructure costs, particularly as major memory suppliers (Samsung, SK Hynix, Micron) have each reached trillion-dollar valuations.
Summary
XCENA, a four-year-old startup with operations in South Korea and the United States, has raised $135 million in Series B funding at a $570 million valuation, according to multiple sources including TechCrunch. The round was co-led by Atinum Investment and IMM Investment. This Series B brings the company's total capital raised to $185 million since its 2022 founding.
The startup was co-founded by Jin Kim (CEO), Dohun Kim (CTO), and Harry Juhyun Kim (CPO), all veteran engineers from Samsung and SK Hynix, the memory chip manufacturers that supply components for Nvidia's GPU ecosystem. This pedigree positions XCENA at the intersection of memory innovation and AI infrastructure, a space gaining significant institutional investor attention.
XCENA's core thesis challenges prevailing assumptions about AI infrastructure bottlenecks. Rather than focusing on compute capacity—where companies like Nvidia have dominated—the startup targets what it identifies as the true structural inefficiency: memory. During AI inference and generation, data continuously cycles between DRAM (fast, short-term memory), CPUs, and GPUs in what XCENA characterizes as an expensive and power-intensive relay. The company's MX1 chip relocates compute capabilities much closer to DRAM, enabling routine data operations—including key-value cache management and data orchestration—to be processed near memory without costly round trips through external processors. According to CEO Jin Kim in a TechCrunch interview, this represents a fundamental architectural shift: "CPUs and GPUs have both gotten smarter over the decades. Memory never did. XCENA wants to change that."
The timing of this funding reflects broader market dynamics. The Series B announcement coincided with a notable milestone: Samsung, SK Hynix, and Micron—the three companies dominating the global memory chip market—each achieved trillion-dollar valuations for the first time, signaling investor recognition that memory infrastructure has become a critical competitive frontier in AI systems. Kim further noted that "the recent rise in memory prices and related stocks points to a broader shift in AI infrastructure toward memory-centric architectures," suggesting XCENA's approach aligns with emerging structural trends in how enterprises deploy AI systems.
The funding will be deployed to scale customer deployments, expand sales and marketing operations, and advance the company's CXL-based computational memory solutions for enterprise customers. As the South Korean AI chip sector accelerates, XCENA's Series B positions it to compete for infrastructure contracts at a moment when memory optimization has shifted from theoretical advantage to practical necessity for cost-conscious enterprise AI deployment.
The startup was co-founded by Jin Kim (CEO), Dohun Kim (CTO), and Harry Juhyun Kim (CPO), all veteran engineers from Samsung and SK Hynix, the memory chip manufacturers that supply components for Nvidia's GPU ecosystem. This pedigree positions XCENA at the intersection of memory innovation and AI infrastructure, a space gaining significant institutional investor attention.
XCENA's core thesis challenges prevailing assumptions about AI infrastructure bottlenecks. Rather than focusing on compute capacity—where companies like Nvidia have dominated—the startup targets what it identifies as the true structural inefficiency: memory. During AI inference and generation, data continuously cycles between DRAM (fast, short-term memory), CPUs, and GPUs in what XCENA characterizes as an expensive and power-intensive relay. The company's MX1 chip relocates compute capabilities much closer to DRAM, enabling routine data operations—including key-value cache management and data orchestration—to be processed near memory without costly round trips through external processors. According to CEO Jin Kim in a TechCrunch interview, this represents a fundamental architectural shift: "CPUs and GPUs have both gotten smarter over the decades. Memory never did. XCENA wants to change that."
The timing of this funding reflects broader market dynamics. The Series B announcement coincided with a notable milestone: Samsung, SK Hynix, and Micron—the three companies dominating the global memory chip market—each achieved trillion-dollar valuations for the first time, signaling investor recognition that memory infrastructure has become a critical competitive frontier in AI systems. Kim further noted that "the recent rise in memory prices and related stocks points to a broader shift in AI infrastructure toward memory-centric architectures," suggesting XCENA's approach aligns with emerging structural trends in how enterprises deploy AI systems.
The funding will be deployed to scale customer deployments, expand sales and marketing operations, and advance the company's CXL-based computational memory solutions for enterprise customers. As the South Korean AI chip sector accelerates, XCENA's Series B positions it to compete for infrastructure contracts at a moment when memory optimization has shifted from theoretical advantage to practical necessity for cost-conscious enterprise AI deployment.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.