Samsung Electronics has begun sampling GAIA, its first NPU-based AI PC accelerator, to Lenovo and HP - a direct shot at Qualcomm and NVIDIA in the fast-expanding Windows AI PC market. Built on Samsung's 4nm process with a memory-centric architecture, the chip signals the company's return to the PC silicon segment it exited more than a decade ago.
Inside the GAIA Design
Samsung's System LSI Business is designing GAIA specifically as a generative-AI accelerator rather than a full CPU replacement, according to reports first surfaced by South Korea's Chosun Ilbo and Seoul Economic Daily. The chip pairs a large NPU with a memory-centric layout tuned for on-device generative workloads: real-time language translation, image generation, productivity assistants and background AI tasks now expected on Copilot+ class PCs. Future revisions are slated to integrate Samsung's Processing-in-Memory (PIM) DRAM stack to lift performance per watt further.
Sampling to Lenovo and HP
Samsung has already begun supplying GAIA samples to Lenovo and HP for performance testing, and mass production could begin as early as 2027. That timing lines up with Windows 11 refresh cycles and with a growing OEM push to differentiate on-device AI experiences from those that require constant cloud round-trips. It also puts Samsung in direct competition with Qualcomm's Snapdragon X-family, NVIDIA's RTX Spark and Intel and AMD roadmaps for AI-first laptops.
Part of a Bigger Silicon Push
GAIA slots into a wave of Samsung silicon moves. Its foundry arm just won Meta's $6.5B 2nm MTIA training-chip contract and entered talks with Anthropic on a custom 2nm inference accelerator. Meanwhile the group is watching memory rival SK Hynix launch a $28B Nasdaq uplisting and Micron ramp its own $250B U.S. investment. GAIA is Samsung's answer to a market where every hyperscaler and OEM is racing to control its own AI silicon stack.
Why This Matters
If Samsung ships GAIA on schedule, PC OEMs will finally have a fourth credible AI accelerator supplier alongside Qualcomm, NVIDIA and the x86 incumbents - and Samsung will have a foothold in laptops it can pair with its DRAM, NAND and PIM stack. It also gives the company an internal design win to run through its 4nm capacity as foundry rivals TSMC and Intel push their own AI PC customers.
Reporting based on coverage from Sammy Fans, Seoul Economic Daily, KED Global and Wccftech.