en.Wedoany.com Reported - Samsung is developing a PC-specific AI accelerator codenamed GAIA, with HP in the US and Lenovo in China already testing prototypes to verify its performance. Mass production could begin as early as 2027, with devices featuring the chip expected to launch by late 2027 or early 2028. GAIA is being developed by Samsung's LSI division, which also handles Exynos mobile chips. Samsung has not officially confirmed the project, and no performance data, power specifications, or architectural details have been publicly disclosed for the chip.
GAIA is described as a coprocessor rather than a general-purpose CPU. Built on a 4nm-class process node, it is positioned as a memory-centric AI accelerator that brings computing units closer to memory. GAIA is designed to work alongside processors from Intel, AMD, or Qualcomm, rather than replacing them. The chip is specifically tailored for generative AI workloads on PCs, such as on-device language models, real-time translation, and image generation, with the goal of offloading these tasks from the CPU or GPU to a dedicated NPU. This marks a clear distinction from GPU-based AI accelerators used for large-scale training and inference.
According to reports, Samsung is pushing for further integration with Processing-in-Memory (PIM) technology, which performs computations within memory itself, reducing data transfers between memory and processors. PIM has been a research project at Samsung for years but has yet to achieve a major commercial breakthrough. As GPU speeds improve and software ecosystems mature, the bottlenecks PIM once aimed to solve have become less pressing. A dedicated NPU validated by OEMs may be better suited for PIM technology than GPUs. Samsung is one of the few companies capable of combining custom AI logic with its own DRAM manufacturing, giving it a vertical integration advantage that competitors find hard to match.
Samsung last provided PC chips in 2012, when Exynos chips briefly powered early Samsung Chromebooks, a business that was shelved two years later. Since then, Samsung's Galaxy Book laptops have used Intel or Qualcomm chips, including the Snapdragon X2 Elite in the latest Galaxy Book models. If GAIA enters mass production, Samsung will use its own chips in its laptops for the first time in over a decade. Third-party OEM adoption through HP and Lenovo could expand its influence beyond Samsung's own product line. Samsung's LSI division has been in a structural loss for years, and success in the AI business, along with revenue from Exynos and automotive chips, could provide growth opportunities for the division.
This strategy may also create potential conflicts with existing customers. Both Nvidia and Qualcomm rely on Samsung's foundries to produce some of their chips. Samsung competing with its own customers in the AI PC space while still manufacturing for them could complicate partnerships. How Samsung balances this will impact its foundry business and its ability to sell GAIA to PC makers.
GAIA is currently in prototype testing with HP and Lenovo, with mass production expected to begin in 2027 and consumer devices potentially launching later that year or in early 2028. The timeline may change based on validation results, OEM commitments, and Samsung's ability to provide supporting software for the hardware. Many details about GAIA remain unclear, including its performance compared to competing NPUs (such as the integrated NPU in Intel Core Ultra, AMD Ryzen AI with XDNA NPU, Qualcomm Snapdragon X2 with Hexagon NPU, and Nvidia RTX Spark for AI workloads), as well as power and thermal characteristics, and the maturity of software support and drivers at launch.
