[Edaily Reporter Kim So-yeon] Samsung Electronics is set to begin production of next-generation artificial intelligence (AI) central processing unit (CPU) chips designed by Akeana, a U.S.-based high-performance semiconductor intellectual property (IP) startup. Industry observers note that this move positions Samsung to establish a track record of reliably manufacturing high-value-added high-performance computing (HPC) and AI chips.

According to industry sources on the 10th, Akeana’s top-tier CPU core, the “Akeana 5000” series, and its proprietary interconnect IP have been realized as physical test chips using Samsung Foundry’s (semiconductor contract manufacturing) cutting-edge process. The test chip, named "Alpine," was manufactured using Samsung’s 4-nanometer (nm) FinFET process. With this achievement, Akiana has successfully progressed beyond the stage of simply providing design blueprints to the stage of validating high-performance server chips capable of actual operation.

Akiana is a fabless company founded in 2021 by key engineers who designed Marvell’s server processor, the “ThunderX2.” The industry is watching it as a strong contender to Arm, the global IP company that has established a dominant position.

Arkiana supplies high-performance CPU design IP for data centers and AI servers using the open-source architecture "RISC-V." While Arm architecture has been virtually the only option for global big tech companies designing their own chips, the RISC-V ecosystem is growing rapidly as RISC-V emerges as an alternative, offering the benefits of no licensing costs and the ability to create custom designs.

There is also speculation that global hyperscalers such as Google and Meta, which are currently developing their own AI chips, may adopt Akiana’s high-performance RISC-V IP. This is because there is growing demand to design custom chips optimized for proprietary AI algorithms and specific workloads while reducing the high licensing costs associated with existing architectures. Akiana has previously stated that it is collaborating with major global companies to verify the performance of its IP and optimize its architecture.

Samsung Electronics and Akiana plan to accelerate the development of next-generation servers and the Agentic AI ecosystem based on the latest 'RISC-V RVA23' standard and server-grade System-on-Chip (SoC) specifications. This is expected to demonstrate that Samsung possesses the technical capabilities to produce custom server CPUs for global big tech companies. At the same time, the successful implementation of these highly complex server-grade chips is seen as evidence that the technical maturity and yield of Samsung’s 4-nanometer process have reached a stable level.

In fact, Samsung Electronics has applied its 4-nanometer process to the manufacturing of the base die (logic die)—the bottom layer of "HBM4," the sixth-generation High Bandwidth Memory (HBM)—significantly boosting the utilization rate and process maturity of that node. Samsung Foundry is currently producing NVIDIA’s autonomous driving chips and Grok chips using 4-nanometer and 8-nanometer processes.

In particular, the successful validation of this server-grade chip is expected to serve as a strong reference for attracting HPC customers to Samsung’s cutting-edge 2-nanometer (2nm) Gate-All-Around (GAA) process in the future. Analysts suggest that fabless and big tech companies that have verified chip performance and quality on the 4-nanometer process are highly likely to consider Samsung Foundry as their preferred partner during the transition to the next-generation 2-nanometer process.

An industry insider stated, “The AI chip production capabilities demonstrated in the 4-nanometer process will play a positive role in retaining and expanding our customer base during the transition to the 2-nanometer process,” adding, “This also signals a green light for Samsung Foundry’s future order growth.”