SK Hynix has told investors that memory shortages will persist until at least 2030, a forecast that forces a rethink of UK server budgets, refresh timelines and vendor contracts for the rest of the decade. This is no longer a short-term price spike — it's a structural market shift.
View the data behind this chart
| Phase | Starts (week) | Duration (weeks) |
|---|---|---|
| Shortage deepens, HBM4 ramps | 0 | 52 |
| 14-buyer contracts lock supply | 52 | 104 |
| Capacity expansion continues | 155 | 1 |
| Demand still exceeds supply | 155 | 1 |
What SK Hynix actually told investors
Releasing quarterly results, SK Hynix said it expects a structure in which demand exceeds supply to persist through 2030, citing the technical difficulty of leading-edge memory processes and the long lead times needed to build new fabs. The company reported quarterly revenue of ₩104.6 trillion ($78.1 billion), up 328 percent year-over-year, and operating profit of ₩87.68 trillion ($62.8 billion), up 486 percent year-over-year.
SK Hynix frames this as a virtuous cycle: AI-driven productivity gains fuel more AI adoption, which drives more infrastructure investment, which in turn pulls in new categories of buyer — general enterprises and nations building sovereign AI infrastructure — on top of the hyperscalers already competing for capacity.
Why UK procurement teams should pay attention now
For UK buyers, the headline is simple: this isn't a blip that eases once a new fab comes online. SK Hynix is explicit that it can't build capacity fast enough to catch up with demand, and it expects the imbalance to run for roughly four more years. That changes the maths on every server refresh, colocation expansion or on-prem AI build currently sitting in a capex plan.
Organisations that assumed memory prices would normalise by 2027 or 2028 now need to reassess. Teams tracking component costs should track DDR5 server memory prices as a standing input to procurement decisions rather than a one-off check before a purchase order.
The 14-buyer contract squeeze and what it means for everyone else
SK Hynix has 14 long-term supply deals with 14 major buyers that will together consume more than half of its output during 2027 and 2028, and some of those customers are already pushing to extend their agreements. The deals include pricing mechanisms tied to market conditions, which the company says secures its own profitability while still letting it flex with demand.
The practical consequence for UK enterprises outside that top tier is reduced bargaining power and less visibility on allocation. Buyers who aren't locked into hyperscaler-scale contracts will be competing for the remaining supply, which argues for earlier ordering, longer lead-time assumptions, and a willingness to decide whether to buy servers now or wait based on realistic allocation risk rather than hope that prices will dip.
- •14 long-term deals cover over 50% of SK Hynix's 2027-2028 output
- •Some customers are seeking to extend those agreements further
- •Non-contracted buyers face tighter allocation and less price leverage
Procurement strategies for a multi-year squeeze
With shortages now framed as a 2026-2030 phenomenon rather than a passing spike, UK IT leaders have three broad levers: buy early and absorb current pricing, phase refreshes to spread the cost impact, or extend the life of existing estate. Each carries trade-offs that need weighing against budget cycles and workload growth.
Extending hardware lifespan is increasingly attractive where refresh budgets haven't kept pace with component inflation. Teams should extend server lifespan with third-party maintenance on ageing fleets, and compare the full economics by choosing to compare new vs. refurbished servers amidst DRAM shortages before committing to a like-for-like replacement. Procurement teams more broadly should revisit how they optimise IT procurement strategies given that memory, not compute, is now the dominant cost and availability constraint in most server configurations.
HBM4, DDR5 and what it means for spec decisions
SK Hynix's roadmap adds another layer for buyers configuring new systems. The company expects HBM4 to overtake HBM3E in both shipment volume and revenue in the fourth quarter and become its mainstream product, while it is also accelerating its 10nm-class 1c process and expanding DDR5 and LPDDR5 SOCAMM sales specifically for AI servers.
Chief financial officer Kim Woo-hyun told the earnings call that PC and mobile device shipment volumes are expected to decrease due to cost burdens from rising memory prices and supply shortages — a signal that manufacturers are prioritising higher-margin AI server memory over consumer products. UK buyers specifying new builds should factor this into vendor discussions when they server configuration plan, since component availability will increasingly favour AI-optimised configurations over standard enterprise builds.
The bottom line for 2026-2030 budgeting
Investors reacted positively to SK Hynix's outlook, sending its share price up roughly three percent in after-hours trading — a sign the market reads this shortage as durable rather than temporary. For UK buyers, the sensible response is to stop treating memory costs as a line item that will correct itself and start treating it as a structural planning input.
That means revisiting multi-year capex models now, building flexibility into refresh schedules, and using tools that plan your IT hardware refresh around realistic supply assumptions rather than historical pricing. Those who adapt procurement cycles to a 2030 horizon will be far better positioned than those still budgeting for a 2027 return to normal.
- 01The Register — SK Hynix says memory shortages to persist until 2030 · 28 October 2026