At a glance
Global devices market
Exhibit 1Memory prices are still rising, just more slowly
Conventional DRAM contract price change, quarter on quarter, 2026
Each quarter's rise compounds on the last. TrendForce expects PC DRAM supply to contract in 2027 as capacity moves to servers.
The category on a page
Exhibit 2Where hardware sits now
Profit impact against supply risk, technology subcategories
Exhibit 3Memory now dominates the cost of a PC
DRAM and SSD share of a PC's manufacturing cost
HP said memory alone reached 35% of PC build materials in early 2026.
Exhibit 4Who shipped fewer PCs
Q3 2026 shipments vs Q3 2025
Bars show the size of each fall. Lenovo: 14.9m units vs 19.3m (derived).
Exhibit 5Where IT budgets are going in 2027
Spending growth by segment, 2027 forecast
Data centres take about half of 2027's $961bn rise, which is why memory goes to servers first.
Exhibit 6Flash prices are rising too
NAND flash contract price change, Q4 2026, quarter on quarter
Typical savings available in a shortage
From specification and lifecycle changes, not unit price
Value levers
In order of value during the shortage.
Agree volumes and delivery dates for 2027 now. Suppliers serve committed customers first.
Match RAM and SSD to the role. Memory is now about 40% of a PC's cost, so every extra gigabyte is expensive.
Fewer models means larger, more predictable orders and better allocation. HP itself halved its configurations.
Move from 3-4 to 4-5 year refresh where performance allows; use refurbished devices for some roles.
Vendors have shortened quote validity. Ask for price holds tied to a memory index, with caps.
How to buy it
KPIs to track
For the monthly review
Where most buyers are, and what good looks like
| Most buyers today | Best practice | ||
|---|---|---|---|
| Buying | Buy when needed at list | → | Allocation agreed a year ahead |
| Specification | Over-spec by default | → | Spec matched to role and reviewed quarterly |
| Model | Capex refresh on a fixed cycle | → | Life extended; device-as-a-service where it fits |
| Supply | One OEM, spot orders | → | Two OEMs, committed volumes, buffer stock |
| End of life | Dispose | → | Refurbish, redeploy, recover value |
What AI changes
More in the AI disruption tab.
From the category playbook
IT hardware & end-user devices
| Positioning | Leverage normally - Bottleneck-leaning in 2026 (memory shortage) |
| Recommended route | Standardise to 2-3 OEMs; aggregate volume; forward-buy/lock pricing; multi-distributor for supply security |
| Spec approach | Standard SKUs/images; right-size RAM/SSD (don't over-spec amid shortage); TCO incl. lifecycle & disposal |
| Top value levers | 1) SKU standardisation 2) Volume aggregation + buy-ahead 3) Lifecycle/refresh & asset recovery |
| Key risk | Memory-driven price spikes · lead-times · Win-10 EOL refresh surge |
| Headline KPIs | Price variance vs benchmark · lead-time adherence · standard-SKU % · cost per seat · asset-recovery value |
Exhibit 7Kraljic positioning (category playbook)
Sub-categories by profit impact and supply risk
Exhibit 8Where the money goes
Typical cost split
Exhibit 9DRAM contract price trajectory (QoQ)
%
Savings potential (category playbook)
Typical range on addressable spend
Verified benchmarks - Global
- Memory is the whole story: TrendForce sees Q4'26 conventional DRAM +10-15% QoQ (+15-20% incl. HBM) and NAND +15-20% - after Q1'26's record +90-95% DRAM. AI-server/HBM demand is starving PC & consumer DRAM into 2027T2TrendForce (30 Sep 2026) · Q4 2026F
- Street-level proof: a 32GB DDR5 retail kit went ~$400 (Apr)→~$500 (Oct 2026); 64GB DDR5 server RDIMM $873→~$1,586 projected by Q4 (Citi); retail SSD prices ~+303% YoYT3TrendForce / Citi / market trackers · Oct 2026
- PC ASP +6-8% while shipments fall -10.4% (2026) - buy-timing and RAM/SSD spec control beat vendor rate negotiation; lock long-term agreements and standardise configs before 2027 tightensT2Gartner / IDC · 2026
- SK Hynix says HBM/DRAM/NAND essentially sold out for 2026; its chair warns the shortage could run to 2030 - treat memory as a secured-supply problem, not a price negotiationT3Industry (Tom's Hardware) · Oct 2026
Supplier landscape - Global
Top global players and the local or regional players that matter here. Illustrative, from public sources; not an endorsement.
Value levers (category playbook)
Hardware is a should-cost and lifecycle game - component indices move the price, and the real savings sit in refresh discipline and what you do with the asset after year one, not the unit negotiation.
Break the quote into memory, CPU/GPU, display, chassis and assembly/margin, and index the memory line to DRAM/NAND spot so you negotiate the component, not the sticker.
Extend standard refresh from 3 to 4-5 years for non-power users; the biggest device saving is buying fewer, not cheaper.
Cut the catalogue to a few standard configs by persona; config sprawl kills volume leverage and inflates support.
Evaluate Device-as-a-Service and certified buy-back/ITAD; residual value and secure disposal are real money usually left on the table.
Buy through the right channel/region and time orders against component cycles; direct-vs-reseller and FX move price more than another 2% of haggling.
Make AI do the work
AI's device angle isn't the laptop - it's the NPU premium you're now being upsold. Decide which personas genuinely need on-device AI silicon versus cloud AI, and don't let a blanket 'AI PC' refresh inflate your standard build. Separately, use asset data to predict refresh and failure so you replace on need, not on the calendar.
Reactive vs proactive (category playbook)
| Reactive | Proactive | ||
|---|---|---|---|
| Buying | Buy at list when needed | → | Lock LTA/allocation ahead of shortage |
| Spec | Over-spec RAM/SSD by default | → | Standardise configs to the task |
| Model | Capex refresh on a cycle | → | DaaS/consumption where it fits |
| Supply | Single OEM, spot orders | → | Secured allocation + buffer stock |
| Lifecycle | Dispose at end of life | → | Circular/refurb; capture residual |
How AI is changing the category
- AI should-cost models decompose the bill of materials (CPU, DRAM, panel, battery) so you negotiate components, not list price.
- Market-intelligence AI tracks DRAM/NAND and FX to time buys around the 2026 memory-shortage spikes.
- Demand-forecasting AI predicts refresh cycles and right-sizes volume commitments.
- Lifecycle/asset-management AI extends device life and optimises ITAD and resale value.
What to do about it
- Build an AI should-cost/BOM model to challenge OEM quotes line by line.
- Use demand-forecast AI to pre-commit volumes ahead of shortage-driven price rises.
- Run AI market scans to buy large volumes at component-price troughs.
- Apply lifecycle analytics to lengthen refresh cycles and cut total cost of ownership.
Each number carries source, as-of date and a tier (T1 official · T2 index/survey · T3 analyst/estimate). 2026 is dominated by a global DRAM/NAND shortage: AI-server & HBM demand is diverting capacity from PC/consumer memory, pushing device prices up and lead times out into 2027. Memory contract prices move quarterly - validate the live TrendForce print and your allocation before committing.
Full written analysis
Written live for Hardware & devices and your region, grounded on the figures above. Public data only.
Written live for this category and region when the page opens in a browser.
Negotiation playbook
Negotiation playbook
Suppliers have pricing power until memory supply eases, likely after 2027. Your leverage is committed volume, a standard catalogue and flexibility on specification and timing.
Pick the right commercial model
| Long-term agreement with allocationUsual best | Committed volume in exchange for guaranteed supply and price rules. The priority in a shortage. |
| Index-linked pricing | Price moves with a published memory index, within a cap and floor. Fairer than open-ended increases. |
| Device-as-a-service | Monthly per-device fee including refresh and support. Smooths cost; check the financing premium. |
| Spot purchases | Buy as needed. The most expensive and least reliable route in 2026-27. |
Questions to ask suppliers
- Allocation
What quantities and delivery dates will you commit to for 2027?
Good answer: committed quantities and dates for 2027. - Price rule
Which memory price index will our prices follow, and will you cap the increase?
Good answer: link to a named memory index, with a cap. - Quote validity
Will you hold quotes for at least 30 days on committed configurations?
Good answer: at least 30 days on committed configurations. - Spec flexibility
Can we switch configurations without penalty if supply changes?
Good answer: right to switch configurations without penalty. - Buy-back
What buy-back value will you guarantee for returned devices?
Good answer: guaranteed value for returned devices.
Your leverage
- Your committed annual volume
- A standard catalogue (fewer configurations)
- Willingness to sign 2-3 year agreements
- Two qualified OEMs
Watch out for
- Price increases applied to open quotes
- Spec downgrades in 'equivalent' models
- Allocation cuts at short notice
Global - 2026 outlook
| Component | Change / forecast | Source |
|---|---|---|
| DRAM contract price (Q4'26 QoQ, incl. HBM) | +15-20% | T2TrendForce · 30 Sep 2026 |
| NAND flash contract price (Q4'26 QoQ) | +15-20% | T2TrendForce · 30 Sep 2026 |
| DRAM contract price (Q1'26 QoQ - the shock) | +90-95% | T2TrendForce · Q1 2026 |
| PC average selling price (2026) | +6-8% | T2IDC · 2026 |
| PC shipments (volume, 2026) | -10.4% | T2Gartner · 2026 |
| IT spending (overall, 2026) | +10.8% | T2Gartner · Feb 2026 |
Pricing models (category playbook)
In a memory shortage the commercial model is about securing supply, not shaving unit price.
| Long-term agreement / locked allocationUsual best | Lock price and allocation ahead of 2027 tightening - in a shortage, guaranteed supply beats a lower unit price. |
| Capex outright purchase | Own the asset; best with cash and a long refresh cycle, but you carry residual and obsolescence risk. |
| Device-as-a-Service (subscription) | Opex per-seat with refresh and support bundled - smooths budget and offloads lifecycle, at a financing premium. |
| Consignment / buffer stock | Supplier holds inventory you draw down - hedges lead-time risk in a shortage; negotiate who owns the buffer. |
Benchmarks to bring to the table - Global
- Memory is the whole story: TrendForce sees Q4'26 conventional DRAM +10-15% QoQ (+15-20% incl. HBM) and NAND +15-20% - after Q1'26's record +90-95% DRAM. AI-server/HBM demand is starving PC & consumer DRAM into 2027T2TrendForce (30 Sep 2026) · Q4 2026F
- Street-level proof: a 32GB DDR5 retail kit went ~$400 (Apr)→~$500 (Oct 2026); 64GB DDR5 server RDIMM $873→~$1,586 projected by Q4 (Citi); retail SSD prices ~+303% YoYT3TrendForce / Citi / market trackers · Oct 2026
- PC ASP +6-8% while shipments fall -10.4% (2026) - buy-timing and RAM/SSD spec control beat vendor rate negotiation; lock long-term agreements and standardise configs before 2027 tightensT2Gartner / IDC · 2026
- SK Hynix says HBM/DRAM/NAND essentially sold out for 2026; its chair warns the shortage could run to 2030 - treat memory as a secured-supply problem, not a price negotiationT3Industry (Tom's Hardware) · Oct 2026
AI disruption
Tailwinds for buyers
- AI fleet management predicts failures, supporting longer device life.
- On-device AI may reduce some cloud costs over time.
- AI tools help right-size specifications by role from usage data.
- AI-driven demand forecasting cuts over-buying of devices.
- AI tools automate asset tracking and refresh planning.
- Device-as-a-service bids are easier to compare with AI analysis.
Headwinds for buyers
- AI servers take priority on DRAM and flash supply, raising PC and phone costs.
- AI PC premium adds cost for users who don't need it.
- HBM demand keeps memory tight through 2027.
- Memory shortage from AI data centres raises PC and server prices.
- AI PCs carry a price premium that isn't always needed.
- Vendors steer components to AI servers, tightening PC supply.
How to capture it
Buy AI PCs only for roles that use on-device AI; use usage data to right-size memory; extend life with predictive fleet tools.
Who is changing the market
- Refurbished devices at lower prices with warranty
- Lets buyers reduce new-device spend
- Modular laptops are repairable and upgradeable
- Longer device life cuts refresh spend
- Devices and infrastructure as a service, paid monthly
- Moves hardware from capital spend to subscriptions
- Pay-per-use infrastructure on premises
- Consumption pricing competes with public cloud
Category playbook view
Tailwinds
- Asset-usage AI predicts refresh and failure, so you replace on need, not on the calendar - fewer devices bought.
- Config and demand analytics expose over-spec'd builds and SKU sprawl before they inflate the order.
- AI-driven ITAM recovers more on buy-back and disposal - residual value usually left on the table.
Headwinds
- The 'AI PC' / NPU premium is an upsell - on-device AI silicon few personas actually use.
- Component pricing is AI-optimised by suppliers; the quote already reflects the demand you signalled.
- 'AI management' features arrive as a per-seat line on refreshes that didn't need them.
Capture it
Decide which personas truly need on-device AI before a blanket 'AI PC' refresh inflates your standard build - and keep buying fewer, not just cheaper.
Risk radar
Risk radar
Exposure by dimension, 0-100, Global
Risk heat matrix
Impact against likelihood, category playbook; register numbers follow
- Component/memory shortage
- Geographic concentration
- Counterfeit / grey market
- E-waste compliance
Top risk flags
- HighComponent/memory shortage & allocation risk
- HighGeographic concentration (Taiwan/Asia fabs)
- MediumCounterfeit / grey-market parts
- MediumE-waste & ESG compliance
Risk register
DRAM +90-95% in Q1 2026 alone; still rising +10-15% in Q4.
PC DRAM supply may shrink as memory makers shift to servers.
Vendors are cutting configurations and cheap models; Gartner sees sub-$500 PCs disappearing by 2028.
Vendors re-price open quotes as costs move.
Delaying refresh too long raises failure and security risk.
Market pulse
What the biggest suppliers are reporting
From their latest results and investor material.
Q1 FY2026/27 (Aug 2026). Raised PC prices twice in 2026; expects shortage to persist; leading PC share.
Early 2026: memory costs doubled in one quarter; configurations halved; long-term supply agreements signed.
Q2 FY27 (+58%). AI servers drive growth; components redirected from PCs to infrastructure.
Q3 2026. 55.4m PCs shipped, lowest since early 2023; higher memory and storage costs now fully in prices.
Latest developments
Dated, newest first. Checked 16 Nov 2026.
Client revenue +20%; component shortages persist.
So what: PC supply may be tight; order early and lock prices.
Constellation Research →Reports say Samsung and SK hynix could cut 2027 PC DRAM output by 20-40%.
So what: Lock 2027 PC volumes now; flash may ease before DRAM does.
TrendForce →Devices $916bn (+8.1%); data centre systems +55.5%; memory prices more than doubled in 12 months.
So what: Device budgets rise on price, not volume.
Gartner →62.7m units; early stockpiling and memory costs; HP -30.9%, Apple -11.3%.
So what: Some short-term discounts on stock, but prices stay high.
IDC (preliminary) →AI server demand keeps memory undersupplied; PC OEMs buying aggressively for 2027.
So what: Expect another price increase round in early 2027.
TrendForce →Q1 FY2026/27 revenue $26.94bn; Dell, HP and Super Micro raised prices 10-30%.
So what: All major OEMs are passing costs through.
CGTN →What this means for your next negotiation
- Memory costs peak around Q4 2026 (HP): avoid big PC orders at the peak if you can wait.
- Vendors are moving scarce components to AI servers; secure allocation in contracts.
- Lock price protection for 2027 refreshes.
Signals to watch next quarter
PC pricing and supply.
Whether the memory squeeze eases.
HP expects 60-70%.
Curated developments (category tool)
TrendForce →
Tom's Hardware →
Citi / market trackers →
Gartner / IDC →
IDC / Storyboard18 →
Benchmarks and forecasts - Global
- Memory is the whole story: TrendForce sees Q4'26 conventional DRAM +10-15% QoQ (+15-20% incl. HBM) and NAND +15-20% - after Q1'26's record +90-95% DRAM. AI-server/HBM demand is starving PC & consumer DRAM into 2027T2TrendForce (30 Sep 2026) · Q4 2026F
- Street-level proof: a 32GB DDR5 retail kit went ~$400 (Apr)→~$500 (Oct 2026); 64GB DDR5 server RDIMM $873→~$1,586 projected by Q4 (Citi); retail SSD prices ~+303% YoYT3TrendForce / Citi / market trackers · Oct 2026
- PC ASP +6-8% while shipments fall -10.4% (2026) - buy-timing and RAM/SSD spec control beat vendor rate negotiation; lock long-term agreements and standardise configs before 2027 tightensT2Gartner / IDC · 2026
- SK Hynix says HBM/DRAM/NAND essentially sold out for 2026; its chair warns the shortage could run to 2030 - treat memory as a secured-supply problem, not a price negotiationT3Industry (Tom's Hardware) · Oct 2026
Sources
- Gartner, worldwide IT spending forecast · 21 Oct 2026
- TrendForce, 4Q26 memory contract prices · 30 Sep 2026
- IDC (preliminary) · Oct 2026
- Omdia, memory share of PC cost (via TechRadar) · Oct 2026
- Dell, HP and Lenovo results and calls · 2026
Figures checked 16 Nov 2026. Tags: Official = published by the source; Derived = calculated from published figures; Judgement = The Procurement Lens view.