If you’re planning a PC build or upgrade this year, you’ve likely noticed the uncomfortable trend: prices are climbing, and availability is shrinking. What began as isolated pressure on DRAM modules has now metastasized across three critical component categories—GPUs, RAM, and SSDs. Understanding why this is happening, and how long it might last, is essential for making informed purchasing decisions in 2026.
GPUs: The AI Hangover Meets Gaming Demand
Graphics cards sit at the epicentre of 2026’s supply crunch. The root cause extends far beyond the gaming market.
Following the explosive AI infrastructure build-out of 2023–2025, foundry capacity at TSMC and Samsung remains overwhelmingly committed to AI accelerator chips. NVIDIA’s datacentre revenue continues to dwarf its gaming division, and AMD has similarly prioritised Instinct MI series production. The result? Gaming GPU die allocation has contracted substantially.
Compounding this structural constraint:
- GDDR7 yield challenges: The transition to faster memory has produced higher defect rates than anticipated, reducing usable supply for mid-range and high-end cards
- Advanced packaging bottlenecks: CoWoS and similar technologies remain capacity-constrained, affecting both AI and consumer GPUs that share packaging lines
- Tariff uncertainty: Ongoing trade policy volatility has disrupted inventory planning, with distributors holding leaner stock than usual
Current street prices for RTX 50-series cards sit 25–40% above MSRP at the mid-range, with flagship models commanding even steeper premiums. The used market has tightened correspondingly, as upgraders delay hand-me-down sales.
RAM: From Oversupply to Strategic Constraint
DRAM pricing has executed a dramatic reversal from the oversupply depression of 2023–2024. The mechanisms here differ subtly from GPUs but interconnect critically.
Major manufacturers—Samsung, SK Hynix, and Micron—executed aggressive production cuts in late 2024 as DDR4 inventory swelled and spot prices collapsed. These cuts, combined with a faster-than-expected DDR5 adoption curve driven by Intel and AMD platform transitions, eliminated the buffer stock that had suppressed prices.
The HBM3E factor deserves particular attention. High Bandwidth Memory for AI accelerators now commands premium wafer allocation at all three major DRAM houses. Every wafer dedicated to HBM3E is one diverted from commodity DDR5. With AI demand showing no deceleration, this cannibalisation effect will persist through at least Q3 2026.
Practical impact for builders:
- 32GB DDR5-6000 kits that traded at £85–£95 in mid-2024 now command £130–£160
- DDR4, paradoxically, has seen selective firming as legacy system maintenance demand meets shrinking production
- ECC and workstation-grade memory carries extended lead times, particularly for registered DIMMs
SSDs: NAND’s Quiet Squeeze
Solid-state storage has attracted less headline attention than GPUs or RAM, yet the pricing trajectory is equally significant for system builds.
The NAND flash industry underwent substantial consolidation and capacity rationalisation following the brutal price war of 2023. Kioxia-Western Digital merger completion, Samsung’s deliberate fab underutilisation, and SK Hynix’s disciplined capex have all constrained bit supply growth.
Simultaneously, demand vectors have strengthened:
- QLC enterprise adoption: Hyperscalers have accelerated QLC deployment for warm storage, absorbing capacity that previously flowed to consumer TLC drives
- AI training datasets: Large-scale AI operations require substantial high-speed local storage, creating new enterprise demand at price-insensitive levels
- PCIe 5.0 transition costs: Controller complexity and thermal management requirements have raised BOM costs for next-generation drives
Consumer TLC drives have seen 20–30% price increases from 2024 troughs, with high-performance PCIe 4.0 and 5.0 models most affected. Budget QLC options remain more accessible but represent a compromise on endurance and sustained write performance that many builders understandably resist.
When to Buy: A Tiered Strategy
Universal guidance proves impossible in this environment, but component-specific timing logic can improve outcomes.
Buy now if:
- You require a system for professional or income-generating purposes—opportunity cost typically exceeds further price speculation
- Your target GPU is available at a price you can tolerate; waiting for “normalisation” carries genuine risk of further increases through H1 2026
- DDR5 RAM is needed urgently; current pricing, while elevated, may not represent a peak given HBM allocation pressures
Consider waiting if:
- Your GPU requirement is flexible or modest; entry-level discrete cards and integrated solutions may see relative stabilisation as Intel and AMD refresh budget offerings
- SSD capacity is the primary need rather than speed; QLC and older-generation TLC may soften as new fab capacity ramps in late 2026
- You can defer until Q4 2026, when foundry allocation discussions for 2027 typically create temporary inventory positioning and potential promotional activity
Mitigation Tactics for Budget-Conscious Builders
Strategic flexibility can offset market headwinds:
- Consider last-generation platforms: DDR4-compatible motherboards with Ryzen 5000 or Intel 12th/13th-generation processors remain capable, with significantly cheaper memory ecosystems
- Evaluate used GPUs carefully: Mining-era cards carry risks, but ex-corporate workstation GPUs (Quadro/RTX A-series) or warranty-backed refurbished gaming cards offer viable alternatives
- Right-size your SSD: A modest boot drive with bulk cold storage on HDD remains economically rational for non-latency-sensitive applications
- Monitor bundle opportunities: Retailers facing margin pressure occasionally package RAM, SSDs, or motherboards with GPUs at effective discounts
The Outlook: 2026 and Beyond
Structural supply constraints will not resolve quickly. Foundry capacity expansion—TSMC’s Arizona and Japan facilities, Samsung’s Taylor, Texas fab—will not meaningfully augment gaming-relevant supply before 2027. HBM demand growth appears structural rather than cyclical. NAND rationalisation has proven more durable than prior industry cycles.
However, demand destruction at current price levels may create localised relief. The PC upgrade cycle has extended considerably; enthusiasts and professionals will pay premiums, but mainstream consumers increasingly defer. This elasticity boundary—where volume collapse forces pricing recalibration—represents the most plausible mechanism for improvement in late 2026 or early 2027.
For now, builders face a market that rewards decisiveness over perfectionism. The “deal” of 2024 is not returning imminently. The relevant question is whether current prices, however elevated, justify the utility extracted from a completed build—or whether patience and tactical compromise can bridge to a more favourable environment.
