The next generation of graphics memory, GDDR7, is still several quarters away from broad availability. Limited output in that pipeline has pushed system builders and add-in-board partners to compete for existing GDDR6 inventories, reinforcing upward pressure on costs. Because most retail pricing follows component contracts with a delay of a few months, higher memory quotes booked now will translate into higher shelf prices for consumers next year.
AI demand diverts production capacity
Manufacturers of memory chips and advanced packaging are prioritizing high-bandwidth components used in data-center accelerators. Hyperscale operators are racing to train larger AI models, driving demand for HBM stacks and DDR5 RDIMMs. Wafer starts, assembly lines and back-end capacity are therefore shifting toward parts that carry the highest margins.
This reallocation has tightened availability of consumer-grade DRAM and graphics memory. Desktop GPU makers feel the impact immediately because memory accounts for a significant share of the total board cost. AMD’s internal mitigation efforts—such as negotiating longer contracts and paying premiums to suppliers—have delayed a list-price adjustment for most of 2025, but executives are now preparing retailers and system integrators for changes in the first calendar quarter of 2026.
Competitive landscape
Nvidia, which also relies on GDDR6 for its mainstream GeForce products, is experiencing similar supply pressure. Instead of lowering prices or accelerating a new consumer lineup, the company is focusing on data-center accelerators that use high-bandwidth memory, a segment with higher gross margins. Market watchers do not expect Nvidia to counter AMD’s forthcoming hike with discounts, as both suppliers confront identical cost structures in the memory chain.
While a 10 percent increase may appear modest compared with the spike in component costs, board partners note that retail mark-ups can amplify the effect. A Radeon card currently selling for $599 could approach $660 after the adjustment, depending on channel inventory and regional taxes.
Demand outlook
During AMD’s third-quarter earnings call held in October 2025, Chief Executive Officer Lisa Su said “the demand for compute has never been greater,” referencing strong orders for data-center processors and AI accelerators. Although the company’s gaming unit remains profitable, management signaled that input costs need to be balanced with long-term product investment.
Market research firm IDC projects continued growth in AI-related semiconductor spending through 2027, suggesting that competition for high-performance memory could persist. If capacity expansions at major DRAM fabs lag behind demand, consumer components may remain in short supply well into 2026.
Retailers and system builders are preparing for constrained inventory during the key spring upgrade cycle. Some integrators have begun locking in purchase orders for first-quarter shipments, anticipating that successive batches will carry higher invoices. Resellers with existing stock purchased under 2025 contracts may briefly offer lower prices, but channel participants expect parity once those units sell through.
What changes for consumers
For PC enthusiasts, the price of memory now plays an outsized role in the total cost of a graphics upgrade. Decisions that once centered on core count, clock speed or ray-tracing performance increasingly start with component availability. AMD’s planned adjustment underscores the link between AI infrastructure spending and retail GPU pricing—an indirect but tangible effect for gamers and hobbyist system builders.
Company representatives declined to comment on specific pricing strategy but reiterated that product roadmaps remain on schedule. The next Radeon refresh is still expected in late 2026, according to supply-chain contacts.
Until additional DRAM capacity comes online or AI demand normalizes, component costs are likely to keep board prices elevated. Analysts do not believe the situation will reverse quickly, noting that new memory fabrication facilities require multi-billion-dollar investments and at least 24 months to reach volume production.
Crédito da imagem: Getty Images