AMD formally notified add-in board partners on July 31 that Radeon GPU and GDDR6 memory kit prices will rise by at least 10% effective August 2026, following NVIDIA's earlier price increases. The driver is a threefold surge in GDDR6 spot prices—from approximately $2.50 per gigabyte in late 2025 to around $7.50 per gigabyte—driven by data-center AI accelerator demand (H100, MI300X) consuming 70% of global memory production. On high-VRAM Radeon models, memory now accounts for over 80% of total bill-of-materials cost, forcing both AMD and NVIDIA to pass margin pressure to AIB partners and retail.
The August increase follows AMD's July 10% raise and is the continuation of phased hikes that began in January 2026. NVIDIA cut RTX 50-series production by 30–40% in H1 2026 and raised prices earlier in the cycle; analysts now expect additional GPU price rounds in Q3–Q4 2026 if HBM demand remains constrained and TSMC's wafer capacity continues reallocating from GDDR7 to HBM3e/HBM4 for data-center accelerators. For Radeon RX 9070 cards currently near $600, the increase could push street prices to $660–$700 within weeks as new inventory filters through channel. Both console markets (Xbox Series X now $799.99, Nintendo Switch 2 launching $499.99 September 1) have absorbed concurrent memory-driven price increases, eliminating affordable gaming alternatives.
Architects sizing compute budgets should front-load GPU purchases before August inventory flips to higher-priced SKUs if current-generation performance meets requirements. The structural message is clear: memory scarcity is now the binding constraint on AI infrastructure scaling, not silicon design or supply, and pricing is a pass-through mechanism—expect pass-through to continue until HBM production capacity expands in late 2027.