DDR5 RDIMM contract prices up ~90–95% QoQ in Q1 2026; no near-term relief.
FOMO
The IT supply chain moves fast — and so should your purchasing. FOMO is Sivility’s running watch on the supply chain news affecting the manufacturers we represent: memory and storage shortages, GPU and HBM constraints, networking and server lead times, and the pricing changes that follow. Each signal is mapped to the partners it touches, with a clear read on what it means for buyers.
Conditions at a glance
2026 HBM3e capacity fully allocated; Blackwell lead times stretching past 8–16 weeks.
Capacity reallocated toward AI demand; pricing and lead times climbing.
Memory-driven cost pressure forcing OEM price changes and shorter quote validity.
Commercial PC pricing rising as memory costs flow into client device BOMs.
Latest signals
AI-driven DRAM supercycle drives extreme price volatility and long lead times
Reporting describes a structural memory supercycle: AI demand, hyperscaler stockpiling, and wafer capacity reallocated toward high-bandwidth memory (HBM) are squeezing conventional DDR5. Conventional DRAM contract prices rose a record 90–95% quarter-over-quarter in Q1 2026, with further increases projected and no credible near-term relief — new fab capacity isn't expected to reach meaningful volume before late 2027 or 2028.
Anything with memory in the BOM — servers, storage arrays, switches, and PCs — is exposed. Extend purchase orders to 90–120 days, lock long-term pricing where possible, and place orders earlier in the budget cycle than you used to.
Cisco revises partner contract terms and repricing rights amid memory shortage
Cisco told channel partners it can now cancel compute orders up to 45 days before shipment and reserves the right to adjust pricing on compute orders if component costs rise between order and ship dates. Cisco also canceled certain compute promotions and deal-registration discounts and signaled list price changes on product lines containing memory, citing 'unprecedented market volatility.'
Cisco compute quotes are honored for shorter windows and may be repriced before shipment. Approve and place Cisco compute orders quickly once quoted, and budget for the possibility of price movement on memory-heavy configurations.
NVIDIA warns AI demand could keep GPU, memory, optics, and networking supply tight through 2027
NVIDIA leadership said demand may keep outrunning capacity across memory, optics, networking, and finished AI systems. Reporting points to the real bottleneck being CoWoS advanced packaging and HBM rather than the GPU die itself, with the entire 2026 HBM3e supply already sold out and Blackwell lead times reported well beyond a quarter.
AI infrastructure projects need long runways. Begin sourcing GPUs, DGX/HGX systems, and AI networking a quarter or more ahead of need, and consider phased deployments so a single constrained component doesn't stall the whole build.
SK Hynix and Micron sell out 2026 HBM3e capacity; older accelerators appreciating
Coverage of the 2026 GPU market reports that SK Hynix and Micron have sold out their entire 2026 HBM3e production, the memory shortage that drives accelerator availability. H100 and H200 lead times are reported in the 36–52 week range at some brokers, and even aging A100/H100 hardware is reportedly holding or appreciating in value.
If your roadmap needs accelerated compute in 2026, treat availability — not list price — as the gating factor. Lock allocation early, and evaluate certified refurbished or prior-generation GPUs as bridge capacity.
HPE and Cisco shorten how long quoted prices are honored
Analysts told The Register that spiraling memory costs are pushing hardware vendors to shrink the window a quoted price is valid, with pricing for many memory products having nearly doubled in a matter of months. Both HPE and Cisco adjusted terms governing how long prices hold and when orders can be repriced or canceled.
Quotes for servers, storage, and compute are perishable. Move from quote to PO faster, and confirm price-protection windows in writing before approving budget.
2026 server hardware crunch stretches OEM lead times and budgets
Coverage of the 2026 server market describes long waits, soaring prices, and quotes that look nothing like a year ago — driven by AI infrastructure consuming memory production upstream. The reporting positions certified refurbished hardware as a faster, lower-cost path for workloads that don't require the newest silicon.
For refresh and capacity projects that don't need bleeding-edge parts, certified refurbished and prior-generation gear can cut both lead time and cost. Sivility can quote new and certified-refurbished options side by side.
NAND flash capacity reallocated toward AI pushes SSD pricing and lead times up
The same AI-driven capacity reallocation hitting DRAM is tightening NAND flash, with reporting describing the broader memory market shifting from a structural constraint into an acute shortage that affects enterprise storage, consumer devices, and cloud infrastructure alike.
Storage refreshes and capacity expansions are exposed to the same volatility as memory. Size storage projects early, and consider blanket orders to hold pricing on high-capacity SSD tiers.
Cisco moves silicon-first to insulate AI supply — a signal worth watching
Cisco is reportedly investing billions to move upstream into silicon and optics, using its balance sheet to secure long-term semiconductor and memory capacity. Roughly half of its AI hyperscale business is described as systems built on in-house Silicon One, a structural shift in how upstream supply is managed.
A vendor securing its own supply is generally good news for delivery reliability over time, but the near-term cost of that strategy can show up in pricing and commercial terms. Watch for continued term changes on Cisco compute and memory-heavy lines.
Commercial PC and laptop pricing rises as memory costs flow into client BOMs
IDC analysts note memory price increases of nearly double in a few months will affect both consumer and commercial electronics. As DRAM and NAND feed directly into notebooks, desktops, and workstations, client device pricing is expected to follow the broader memory market upward.
Fleet refreshes and large laptop/desktop rollouts may cost more than last year's pricing implied. Plan device standardization and volume orders early to hold pricing and availability for the workforce.
Lock pricing and availability before the next move.
Conditions are changing weekly. Send us your part list or project and we’ll help you secure pricing, compare new vs. certified-refurbished options, and time your orders around real lead times.
FOMO summarizes publicly reported supply chain conditions for general guidance and links to original sources. It is not a guarantee of pricing or availability — your Sivility quote reflects live, current terms.