This chip startup raised $135 million on its bet that the biggest bottleneck in AI isn’t compute — it’s memory

Every time you ask a question on ChatGPT, your request starts a data migration race. The information leaves memory, passes through the CPU for preprocessing, moves to the GPU for heavy calculations, and then comes back – and this entire journey is repeated for every word the AI ​​generates.

The bottleneck is structural — it means routing through some of the most expensive and most power-hungry chips in the industry on every request. This inefficiency is exactly what XCENA, a startup with offices in South Korea and the United States, is trying to solve. The four-year-old startup has designed a chip that puts computing capabilities much closer to DRAM — fast, short-term memory chips that store data that the processor is actively using — allowing routine data operations to be handled closer to memory, without costly round-trips between CPUs, GPUs and memory.

If this idea succeeds on a large scale, the consequences for AI infrastructure costs could be significant, which largely explains the enthusiasm of investors across the country. In fact, XCENA just raised $135 million in a Series B at a valuation of $570 million, bringing the total amount raised to $185 million.

XCENA CEO Jin Kim co-founded the startup in 2022 alongside CTO Dohun Kim and CPO Harry Juhyun Kim, all veterans of Samsung and SK Hynix, the memory giants that supply the chips that power Nvidia’s GPUs. “CPUs and GPUs have gotten smarter over the decades. Memory has never done that.

XCENA is betting its business on the premise that “inference is not just a computational problem; it is increasingly a memory measurement problem,” Kim said.

The XCENA chip, the MX1, connects to the CPU through the CXL (Compute Express Link) — essentially a dedicated fast lane between the processor and memory — to process data before it needs to leave the memory module. It brings computation to data, not the other way around. The company claims that what used to require 10 servers can run on just one.

“While GPUs excel at matrix multiplication — the heavy math behind training an AI model — much of the surrounding data formatting, including preprocessing and KV cache management [the system that stores prior conversation context so a model doesn’t have to reprocess it]Data caching still runs on central processing units (CPUs). “Our chip handles these tasks directly within the memory module itself,” Kim said.

Demand for memory solutions has been rising since the second half of last year, and the company believes the timing is working in its favour.

Talks with several global memory vendors are still in the early stages, although Kim declined to name them. The company’s ideal customers are hyperscalers who spend tens of billions annually on AI infrastructure, where small gains in memory efficiency could mean hundreds of millions in savings.

The MX1 is still a prototype. Mass production chips are scheduled to roll out of Samsung’s foundry lines by the end of 2026, with the company expecting to generate revenue starting in 2027.

While NPU makers compete to challenge Nvidia in training workloads, XCENA targets the memory-intensive layer underneath it all.

XCENA’s closest competitors include Astera Labs and Marvell, both Nasdaq-listed companies working on next-generation memory communications. Kim said Marvel is a large, well-established company already working in the same field, adding that the difference comes down to intellectual property. “We have thousands of cores,” Kim said. Based on the general specifications, Marvell’s approach relies on a small number of general-purpose kernels in comparison.

These cores are built on RISC-V — an open source chip design scheme — and are specifically optimized for data processing, While intentionally keeping each core small and efficient. Beyond the cores themselves, XCENA designs its own internal memory hierarchy, interconnect bus, and dynamic random access memory (DRAM) controller—a level of vertical integration that most chip companies, including larger competitors, typically outsource.

Seoul-based venture capital firms Altinum and IMM Investment co-led the Series B round, along with Corstone Asia and existing investors SBI Investment and Mirae Asset Capital. The company, which has more than 90 employees across its offices in Pangyo, a technology hub outside Seoul, and in Sunnyvale, is also in talks with international investors about additional financing.