Samsung Electronics is developing a dedicated AI accelerator chip for AI PCs, codenamed GAIA, with mass production targeted for 2027. The project, led by the System LSI division under the DS unit, has already shipped prototype samples to major PC makers including Lenovo and HP for performance validation.
Built on a 4nm process, GAIA is architected around an NPU (neural processing unit) , focusing on on-device generative AI workloads rather than general-purpose CPU or GPU tasks. Samsung is also co-optimizing the chip with PIM (processing-in-memory) technology to reduce data movement overhead and improve edge AI efficiency.
Beyond AI PCs, GAIA targets humanoid robots and other physical AI endpoints, with a strong emphasis on performance per watt. Official specifications—including TOPS ratings, interface standards, and platform integration details—have not yet been disclosed.
Samsung previously ventured into PC processors with its Exynos lineup for Chromebooks in 2012, but exited after two years. Now, with AI PCs emerging as the next major battleground after data centers, and competitors like Qualcomm, NVIDIA, and Huawei already in the space, GAIA represents Samsung’s return to the client-side accelerator market – this time with a clear AI-first focus.
From ICgoodFind: Samsung is betting on PIM + NPU to break into AI PCs. If GAIA delivers on efficiency, it could challenge the incumbents where it matters most – battery life and local LLM performance.
