Samsung and AMD announce a strategic partnership where Samsung will license custom graphics IP based on AMD's Radeon GPU tech for use in its mobile GPUs
Andrei Frumusanu / AnandTech :
Context & Ripple Effects
This 2019 announcement is the origin point of one of the longest-running silicon alliances in mobile: Samsung licensing custom graphics IP built on AMD's Radeon architecture rather than continuing to develop its mobile GPU in-house. What followed validates the bet — the pair moved to RDNA 2-based Exynos chips with ray tracing support in 2021, then signed a multi-year extension in 2023 covering multiple generations of Radeon graphics across more Exynos SoCs.
The relationship has since outgrown phones entirely: by 2026 the two companies were negotiating a preliminary deal for Samsung to supply next-gen HBM4 memory for AMD's MI455X data-center accelerators (per Bloomberg), making this mobile graphics license the first strand of what became a broad component-and-IP partnership.
First-order effects
- Samsung's Exynos SoC team gets a competitive mobile GPU roadmap built on proven Radeon architecture, addressing the graphics gap against rivals instead of spending years on internal development.
- AMD gains a recurring licensing revenue stream and puts Radeon-derived silicon inside hundreds of millions of Samsung handsets — a mobile footprint it never had with its own chips.
Second-order effects
- Qualcomm's Adreno GPUs and Arm's Mali licensees face an Exynos lineup whose graphics pedigree is now externally validated, raising the differentiation bar for every Android flagship chipset bid.
- AMD's IP-licensing playbook proves out at Samsung scale, strengthening its position to strike similar custom-GPU deals with other SoC vendors rather than ceding mobile entirely.
Third-order effects
- If the pattern holds, deep IP licensing becomes the entry wedge for broader supplier relationships — here evolving from mobile graphics into Samsung supplying HBM4 and DDR5 for AMD's data-center accelerators, tying the two companies' fates across consumer and AI infrastructure markets.
- Custom licensed GPU IP could consolidate as the default path for merchant chipmakers, shrinking the set of companies that still design their own mobile graphics from scratch.
The trend: GPU IP licensing is becoming the connective tissue of cross-company silicon alliances, starting in mobile and expanding into data-center supply chains.