December 25, 2025, Suwon, South Korea — Samsung Electronics is accelerating plans to bring mobile graphics processing fully in-house, with multiple reports pointing to a proprietary GPU architecture arriving in an Exynos application processor as early as 2027.
A report citing Cailian Press, says Samsung’s System LSI Division is pushing toward a “100% proprietary technology” mobile GPU and could apply a self-developed architecture in the Exynos 2800 timeframe. The same report frames the move as part of a broader device-side AI push—extending beyond smartphones to categories such as smart glasses, vehicle “smart cockpits,” and humanoid robots—before eventually leveraging that capability to pursue custom chips (ASICs) for external customers, likening the ambition to becoming a “second Broadcom.”
Some sources report that Samsung plans to launch an in-house GPU by 2027 and integrate it into the Exynos 2800 mobile application processor, characterizing the shift as a step toward an end-to-end on-device AI ecosystem.
Taken together, the reports suggest Samsung sees custom GPU IP as increasingly central to differentiating on-device AI performance and power efficiency—while also reducing dependency on third-party suppliers and building a platform it can reuse across future edge-AI products.
