Samsung returns to PC silicon after 13 years — "GAIA" AI accelerator targets in-memory compute, HP and Lenovo already testing
TL;DR
Samsung is developing "GAIA," a 4nm PC-specific AI chip, with HP and Lenovo already testing samples. First PC-processor comeback since 2012 Exynos Chromebook.
Samsung's LSI division is developing a PC-specific AI chip codenamed GAIA, built on a 4nm process, with HP and Lenovo already receiving samples for testing. This is Samsung's first return to the PC processor market since the 2012 Exynos Chromebook — a 13-year gap.
GAIA is not a CPU or GPU replacement. It's positioned as a "memory-intensive" AI accelerator for local generative AI tasks: language model inference, real-time translation, image generation. Samsung is chasing the Qualcomm Copilot+ PC NPU lane, but with an extra card — the plan is to integrate GAIA with the in-development Processing-in-Memory (PIM) DRAM, pushing compute into the memory itself and bypassing the Von Neumann bottleneck.
Mass production targets 2027, with devices arriving late 2027 through early 2028. Samsung has not confirmed GAIA publicly and has released no performance or power figures.
The backdrop is Windows-on-ARM crossing 15% market share this year. Qualcomm X Elite has held the chair for two years, MediaTek Kompanio Ultra just entered, NVIDIA + MediaTek's GB10 ships early next year. Samsung is late to the lane, but PIM integration is a path no rival has walked.
Win the bet, and PIM + AI silicon becomes Samsung's second PC leverage lever alongside HBM. Lose it, and GAIA becomes the second Exynos Chromebook — a Samsung PC chip that stays niche.
via TechSpot
GAIA is not a CPU or GPU replacement. It's positioned as a "memory-intensive" AI accelerator for local generative AI tasks: language model inference, real-time translation, image generation. Samsung is chasing the Qualcomm Copilot+ PC NPU lane, but with an extra card — the plan is to integrate GAIA with the in-development Processing-in-Memory (PIM) DRAM, pushing compute into the memory itself and bypassing the Von Neumann bottleneck.
Mass production targets 2027, with devices arriving late 2027 through early 2028. Samsung has not confirmed GAIA publicly and has released no performance or power figures.
The backdrop is Windows-on-ARM crossing 15% market share this year. Qualcomm X Elite has held the chair for two years, MediaTek Kompanio Ultra just entered, NVIDIA + MediaTek's GB10 ships early next year. Samsung is late to the lane, but PIM integration is a path no rival has walked.
Win the bet, and PIM + AI silicon becomes Samsung's second PC leverage lever alongside HBM. Lose it, and GAIA becomes the second Exynos Chromebook — a Samsung PC chip that stays niche.
via TechSpot
