XCENA Nets $135M to Solve AI's Memory Crisis with Smart Chips

SANTA CLARA, CA – May 29, 2026 – As artificial intelligence workloads push traditional computing to its breaking point, startup XCENA has secured a major financial endorsement for its mission to reinvent data center architecture. The company announced it has closed a $135 million Series B funding round, bringing its total capital raised to $185 million and catapulting its valuation to $570 million.

The investment, co-led by Atinum Investment and IMM Investment, will be used to accelerate the global rollout of XCENA's memory-centric computing solutions, which aim to solve the critical data bottleneck problem plaguing the AI industry. The funds are earmarked for scaling customer deployments, expanding go-to-market capabilities, and advancing the company's next-generation product roadmap.

The 'Memory Wall' Holding AI Back

Modern AI, particularly the large language models and generative applications capturing global attention, is incredibly data-hungry. This has exposed a fundamental weakness in computer design that has existed for decades: the so-called "von Neumann bottleneck" or "memory wall." Traditional systems waste immense time and energy shuttling data back and forth between the central processing unit (CPU) or graphics processing unit (GPU) and the memory where data is stored. As AI models grow larger and their data appetites become more voracious, this bottleneck becomes a critical performance-limiting factor.

“AI workloads are exposing the fundamental limitations of traditional computing architectures as larger models, expanding context windows, and increasingly data-intensive inference workloads drive unprecedented memory demands,” said Jin Kim, CEO and cofounder of XCENA, in the company's official announcement.

This challenge has created a fertile ground for innovation, with a new class of companies seeking to fundamentally redesign how computers handle data. XCENA, founded in 2022 by semiconductor veterans from industry giants Samsung and SK Hynix, is positioning itself at the forefront of this shift.

XCENA's Answer: Computation Inside Memory

Instead of making the path between processor and memory faster, XCENA’s solution is to eliminate much of the journey altogether. The company is a pioneer in computational memory, a technology that merges high-capacity memory with processing cores, enabling computation to happen directly where the data resides. This approach is often called near-data processing (NDP).

XCENA's flagship product, the MX1, is built on the open Compute Express Link (CXL 3.x) standard, an interconnect protocol that allows for more flexible and efficient communication between processors and memory devices. The MX1 integrates thousands of the company's own custom-designed RISC-V processing cores directly with high-capacity pooled DDR5 memory. This allows it to expand a system's memory far beyond the traditional limits of a CPU, with some configurations supporting up to 2TB of memory. An advanced feature, dubbed InfiniteMemory™, even allows the system to use SSD storage as a petabyte-scale extension of memory with minimal latency.

By performing tasks like vector database operations and data analytics directly on the memory module, the MX1 drastically reduces data movement, which in turn promises to lower latency, reduce energy consumption, and decrease the total cost of ownership for data center operators. To ease adoption, XCENA provides a full-stack software development kit (SDK) designed to help hyperscalers, telecom companies, and research institutions integrate the technology with minimal friction.

The $570M Bet on a New Market

Investors are taking notice of the massive market opportunity. The $135 million round, which also saw participation from a wide syndicate of new and existing investors, signals strong confidence in XCENA's approach. Lead investor Atinum Investment has a history of backing successful DeepTech and semiconductor firms, adding significant weight to the endorsement.

“XCENA is redefining how computational memory is applied in real-world systems,” noted Sangmin Lim, Investment Director from Atinum Investment. “Their MX1 product is already showing how customers can simplify complex infrastructure, accelerate deployments, and eliminate inefficiencies that have traditionally slowed down advanced computing workflows.”

The investment is a bet on a market that analysts predict will explode in the coming years. Projections for the CXL component market forecast growth from just over $700 million in 2025 to as high as $16 billion by 2028. The broader compute-in-memory chip market is expected to surge from under $2 billion in 2024 to over $16 billion by 2033, demonstrating the immense financial upside for companies that can successfully commercialize this new technology.

A New Blueprint for the AI Data Center

XCENA's vision extends beyond a single product to a new blueprint for the data centers that power the digital world. The company, which recently rebranded from MetisX to XCENA to reflect its mission, aims to shift the industry from a CPU-centric to a data-centric paradigm.

This places it in a dynamic and competitive landscape. Memory giants Samsung and SK Hynix are developing their own CXL-based memory products, and other major players like Intel, AMD, and NVIDIA are all incorporating the CXL standard into their platforms. XCENA's key differentiators are its deep integration of thousands of processing cores and its full-stack software solution, which it believes will give it an edge.

The company's leadership, steeped in the expertise of South Korea's dominant memory industry, provides a strong foundation. CEO Jin Kim is a former Corporate VP from SK Hynix, while CTO Dohun Kim and CPO Harry Juhyun Kim also bring decades of high-level SoC and architecture experience from SK Hynix and Samsung.

As part of its growth strategy, XCENA is expanding its presence in Northern California to work more closely with the hyperscalers and tech partners that represent its primary customer base. The company plans to provide working samples of its MX1P model to select partners starting in October 2025, with a more advanced MX1S model slated for 2026. While widespread revenue is not expected until production scales in 2027, the successful Series B round provides a long runway to prove that its intelligent memory can become the new standard for the AI era.