Aiming for GPU independence, Samsung plans to debut a fully in-house graphics architecture in the Galaxy S28’s Exynos 2800—joining an elite circle of chipmakers.
Samsung is reportedly preparing to cut ties with AMD and launch its own in-house GPU architecture—a seismic shift in its silicon strategy. The first product to feature the new graphics tech? The upcoming Exynos 2800, likely powering the Galaxy S28 series in 2028.
From AMD to Self-Reliance
Since 2021, Samsung’s mobile GPUs have leveraged AMD’s RDNA architecture, starting with the Exynos 2200.
- The Exynos 2400 and Exynos 2500 continued the AMD collaboration.
- The upcoming Exynos 2600 reportedly uses AMD’s architecture but is internally developed by Samsung’s System LSI.
But according to a South Korean report, the Exynos 2800 will drop AMD completely, featuring a GPU designed from scratch by Samsung.
What does that mean for Samsung’s future?
Joining the Elite Few in GPU Design
Designing a GPU in-house is no small feat. Only four companies—AMD, Nvidia, Intel, and Qualcomm—have successfully built proprietary GPU architectures at scale.
- If Samsung pulls this off, it joins a rarefied group and gains full-stack control over its mobile silicon.
- This allows tighter optimization between CPU, GPU, and AI accelerators—crucial for performance-per-watt gains in phones and beyond.
Could this be Samsung’s path to compete with Apple’s vertical chip dominance?
Heavy Investment, Big Talent
Samsung’s intent has been clear: over the past three years, the company has aggressively hired GPU engineers, offering annual salaries between KRW 300–400 million (~$207K–$277K).
- One notable hire: John Rayfield, a GPU veteran with stints at AMD, Intel, and Broadcom.
- These hires are fueling Samsung’s semiconductor R&D push as it doubles down on custom IP development.
Where else might these GPUs go?
Beyond Phones: Vehicles, Robots, and AI Chips
Once proven in mobile, Samsung’s in-house GPU could power:
- Autonomous vehicle platforms
- Humanoid robots and AR glasses
- AI-focused ASICs—custom silicon akin to what Broadcom and Marvell make for hyperscalers like Amazon, Google, and OpenAI
This shift would allow Samsung to address everything from consumer electronics to enterprise AI infrastructure—with homegrown compute power.
TL;DR:
Samsung is reportedly set to debut a fully in-house GPU inside the Exynos 2800, ending its reliance on AMD’s RDNA architecture. If true, this move would place Samsung among a tiny elite of GPU architecture designers and open the door to next-gen chips for phones, cars, and AI.
