Forget about Galaxy S26: Samsung could pull an Apple for the Galaxy S28's GPU

Samsung has reportedly developed a proprietary GPU architecture planned for the Exynos 2800 in 2027, which could appear in some Galaxy S28 models in 2028, marking a potential shift away from its AMD RDNA-based GPU partnership established in 2019 (extended 2023). The move is framed as an effort to enable improved on-device AI and tighter software optimization, and Samsung plans to deploy the custom GPU technology beyond smartphones into smart glasses, infotainment, autonomous vehicles and humanoid robots, representing a strategic push toward vertically integrated, AI-optimized silicon.

Analysis

Market structure: Samsung moving toward a proprietary GPU (targeting Exynos 2800 in 2027 and phones in 2028) is a long-term competitive shift that directly benefits vertically integrated OEMs (Samsung, AAPL) and in-house software stacks while hurting GPU IP licensors (AMD) for mobile/embedded revenue. Expect incremental pricing power for Samsung phones on AI/UI differentiation but limited immediate share loss for AMD in PCs/consoles; mobile GPU licensing revenue impact could be mid-single-digit percent of AMD revenue by 2028 if adoption scales.

Risk assessment: Key tail risks include execution failure by Samsung (R&D/driver shortfalls), contract/legal disputes with AMD, and fragmentation costs to Android partners; low-probability regulatory intervention around vertical foreclosure is possible. Near-term market reaction should be muted (days–months); credible revenue/contract shifts emerge on 2026–2028 cadence when silicon, supply agreements, and OEM SKUs are announced.

Trade implications: Favor long positions in proprietary beneficiaries (AAPL) and AI/accelerator leaders (NVDA) while hedging or trimming AMD exposure. Use calendar/options structures to time the binary risk around Samsung/AMD announcements (buy-dated puts on AMD out to 12–24 months; buy calls on NVDA or AAPL into product/earnings catalysts). Rotate modest exposure from broad semicap suppliers into software/AI moat names over 6–24 months.

Contrarian angles: Consensus treats this as an immediate AMD negative; that likely overstates near-term revenue lost because AMD’s desktop/console and data-center TAM dominate its P&L. Historical parallel: Apple’s in-house GPU/SoC transition took years and raised Apple margins but also increased SW support costs—Samsung may similarly face multi-year integration drag, creating an opportunity to short the knee-jerk overreaction rather than AMD’s long-term franchise.

Key Decisions for Investors

• Establish a 2–3% portfolio short position in AMD (AMD) sized to risk appetite, or alternatively buy a 12–18 month AMD put spread (buy 12-month put ~20% OTM / sell 12-month put ~35% OTM) to cap premium, targeting a payoff if AMD mobile/IP revenue guidance falls >5% YoY by FY2027.
• Initiate a 2% long position in NVDA (NVDA) and a 1–2% long in AAPL (AAPL) to express upside in AI acceleration and validation of in-house GPU strategy; scale into any pullbacks >10% from current levels and trim into product-announcement rallies (Samsung Exynos launch windows in 2027–2028).
• Construct a relative-value pair: dollar-neutral long NVDA / short AMD (ratio market-cap adjusted) for a 6–24 month horizon to capture divergence if mobile GPU/IP weakness re-rates AMD while NVDA benefits from AI compute demand.
• Reduce cyclical semicap/supplier exposure (e.g., narrow foundry-exposure positions) by 1–3% and redeploy into software/AI franchises over the next 3–12 months; monitor Samsung/AMD contract updates and Samsung M&A/R&D spend — if Samsung confirms full self-sourcing by H2 2026, increase defensive hedges.
• Monitor these catalysts over the next 90–540 days before sizing up: Samsung public roadmap/Exynos 2800 technical reveal (target 2026–2027 windows), AMD quarterly guidance for mobile/IP revenue changes (next 2–4 quarters), and any formal amendments to the 2019/2023 AMD-Samsung agreement.