If your hardware refresh budget was built on last year's memory prices, it's already wrong. A global DRAM and NAND supply shortage — trade press has started calling it "RAMageddon" — has been building since 2025 and intensified sharply through the middle of 2026, and it has moved past the point where IT leaders can treat it as noise to wait out. Consumer DRAM contract prices climbed as much as 89% in a single quarter this year, and by July 2026 spot-market DRAM pricing had surged as much as 700% year-over-year. Those aren't numbers that self-correct on the timeline most refresh cycles assume. For IT directors, CFOs, and procurement teams responsible for laptop fleets, server memory, and desktop replacements, the DRAM price crisis is a budget-planning problem that needs to be addressed this quarter, not a market curiosity to monitor from the sidelines.
The shortage is already reshaping the PC market
The clearest evidence that this isn't a minor blip came from IDC, which reported that worldwide PC shipments fell to 68.2 million units in the second quarter of 2026 — a decline of 4.9%, and the first contraction after nine consecutive quarters of growth. The primary cause IDC cited was the memory shortage itself, not softening demand. That's a meaningful distinction for anyone building a budget forecast: this isn't buyers pulling back because they don't want new hardware, it's the supply chain physically constraining what OEMs can build and ship at prices anyone wants to pay.
The forward-looking estimates vary by analyst firm, and it's worth being upfront that none of them should be treated as a single confirmed number — they're a range of informed projections, not a consensus figure. Some forecasts suggest that by the end of 2026, DRAM and SSD prices could surge as much as 130% compared to 2025 levels, which could push PC prices up by roughly 17%. IDC separately expects average PC selling prices to rise up to 8% in 2026. Counterpoint Research has warned that PCs and phones could climb 10% to 20% by year-end, with the broader PC market contracting more than 11% for the full year. Different methodologies, different numbers, same direction — every credible estimate currently in circulation points toward sustained, meaningful cost inflation on memory-dependent hardware through at least the rest of 2026.
Why this shortage is structural, not the usual boom-bust cycle
Memory pricing has always been cyclical — IT buyers who've been through a few hardware refresh cycles have seen DRAM prices spike and fall before, usually tracking fab overcapacity or a demand dip. What's different this time is the root cause. This isn't primarily a supply-and-demand timing mismatch that self-corrects when a fab comes back online. It's a structural, strategic reallocation of the world's silicon wafer capacity, driven by the hyperscalers.
Microsoft, Google, Meta, and Amazon have driven voracious demand for High Bandwidth Memory — the specialized, high-margin memory that sits inside AI accelerators and data center infrastructure — to build out AI infrastructure at a pace that shows no sign of slowing. That demand has forced the largest memory manufacturers to pivot their limited cleanroom capacity and capital expenditure toward higher-margin enterprise-grade HBM components, which squeezes the supply of conventional consumer and commodity DRAM that goes into the laptops, desktops, and servers most IT departments actually buy. In plain terms: your laptop refresh is now competing directly with hyperscaler AI infrastructure buildouts for the same finite wafer capacity, and the hyperscalers are winning that competition because HBM commands a better margin than commodity DRAM.
That's the detail that should change how IT leaders think about the timeline for relief. Micron has reportedly sold most of its DRAM output through long-term contracts running through 2030, which locks up a meaningful chunk of global supply well past any near-term recovery scenario. Some industry sources — including commentary attributed to SK Hynix — suggest the shortage could persist well past 2030. More broadly, chipmakers have indicated that meaningful relief is unlikely before 2028. If you're building a three-year hardware roadmap right now, the planning assumption should not be "prices normalize next year." It should be "elevated memory pricing is the new baseline for the foreseeable future, and the earliest realistic relief is still roughly two years out."
The AI PC collision nobody planned for
There's a specific irony sitting inside this shortage that IT leaders evaluating next-generation hardware need to reckon with directly. AI PCs — the category OEMs have spent the last two years pushing as the next mandatory refresh cycle — typically need more RAM than a standard business laptop to run local AI features effectively, whether that's on-device inference, local copilots, or AI-accelerated productivity tools. That RAM requirement is colliding directly with the very shortage that makes RAM expensive and scarce. An IT department that planned to standardize on higher-memory AI PC configurations as part of a 2026 or 2027 refresh is now facing a double cost hit: base hardware prices rising alongside PC ASPs generally, plus a memory configuration requirement that's more expensive precisely because of the shortage driving those base price increases. If your AI PC rollout plan assumed 32GB or more as a baseline configuration, it's worth re-costing that assumption against current DRAM contract pricing before finalizing any purchase order, because the gap between a 16GB and 32GB configuration is a much bigger line item than it was a year ago.
It's not just laptops — server memory is exposed too
Most of the public commentary on this shortage understandably centers on consumer PC pricing, because that's where IDC's shipment numbers and the sharpest headlines live. But the same commodity DRAM squeeze runs straight through enterprise server memory, and that's arguably the bigger blind spot for IT budgeting right now. Server refresh and capacity-expansion projects that call for additional RAM — scaling out virtualization hosts, adding memory to database servers, provisioning new capacity for on-premises AI inference workloads — are drawing from the exact same constrained pool of DRAM output that's driving consumer prices up. A data center team that budgeted a server memory expansion based on 2025 unit pricing is likely to find the same category of surprise waiting for them that consumer PC buyers are already experiencing, just with larger absolute dollar amounts given how much memory enterprise servers typically carry per unit.
This is also where the HBM story loops back around in an underappreciated way. The same hyperscaler demand pulling manufacturer capacity toward AI accelerator memory is also pulling attention and priority away from the standard server-grade DRAM modules that most on-premises and colocation infrastructure still runs on. If your organization is planning any kind of server memory upgrade — for virtualization density, database performance, or supporting local AI inference — that project belongs on the same replanning list as laptop and desktop refreshes, not treated as a separate, unaffected budget line.
What to actually do with your 2026-2027 budget
The practical response here isn't panic-buying, and it isn't waiting either — it's a set of specific, deliberate moves that treat this as a multi-year infrastructure decision rather than a single bad quarter to absorb.
Lock in longer-term hardware supply agreements now, while pricing is still comparatively better than it's likely to be in future quarters. If Micron's contracted output already runs through 2030 and analysts broadly agree relief isn't coming before 2028, waiting for a better negotiating window is a bet against the entire industry's own stated capacity plans. Vendors and OEMs that can offer multi-quarter or multi-year pricing locks are worth prioritizing over spot-market purchasing even if the headline unit price looks less attractive today, because the volatility risk on the other side of that decision is real and already documented in this year's spot-price swings.
Reassess whether planned refresh cycles can be extended. A standard three-year laptop refresh stretched to four doesn't just save the direct hardware cost — it avoids buying into the worst part of the price curve. This is worth modeling explicitly rather than assuming the existing refresh cadence is untouchable policy.
Prioritize which systems genuinely need memory upgrades versus which can wait. Not every workstation needs the same memory configuration, and this shortage is exactly the moment to get disciplined about matching configuration to actual workload rather than defaulting to a single standardized spec across the fleet. Developer workstations, data-heavy analytics roles, and AI PC pilot users may justify premium configurations; general knowledge-worker laptops may not, at least until pricing settles.
It's also worth widening the negotiation beyond straight purchase pricing. Leasing arrangements, extended-life warranties on existing fleets, and refurbished or certified pre-owned equipment programs all become more attractive levers when new-unit memory pricing is inflated and unpredictable, and vendors under pressure to move volume may be more flexible on these terms right now than they will be once the shortage becomes fully priced into every quote. It's also worth pushing suppliers directly on whether they can decouple memory configuration from the base unit purchase — buying a lower-memory base configuration now and adding memory later, closer to when it's actually needed, at least keeps the option open rather than locking in a worst-case configuration cost today.
Finally, build the cost inflation explicitly into 2026 and 2027 capital and operating budgets rather than treating it as a rounding error. Somewhere in the 8% to 20% range — and potentially higher on specific memory-heavy configurations — needs to show up as a named line item in hardware budget forecasts, with the honest caveat attached that this is an estimate range drawn from multiple analyst firms, not a single certain number. A CFO who gets blindsided by a 15% hardware cost overrun in Q1 2027 is going to ask why IT didn't see this coming, and by that point, the honest answer — that IDC, Counterpoint, and multiple other analysts had all flagged it months in advance — won't be a satisfying one to give.
The organizations that come out of this cycle in the best shape won't be the ones who found a clever way around the shortage. They'll be the ones who accepted early that this is a multi-year structural shift in silicon capacity allocation, not a temporary supply hiccup, and adjusted their refresh cadence, vendor contracts, and budget forecasts accordingly before the worst of the pricing hit their invoices.