In this article
Introduction
Many IT procurement teams lived through the 2020-2023 chip shortage and came away with a clear lesson: diversify suppliers, build in buffer stock, and don't assume just-in-time ordering will always work. That lesson still applies, but it's not the whole story anymore. A new shortage is underway in 2026, and manufacturers and analysts are describing it in fundamentally different terms than the last one: not a temporary spike to wait out, but a structural reallocation of capacity that may not meaningfully ease for years. This guide covers what's actually different this time, what's driving it, and what it means for how businesses should plan hardware procurement going forward.
What Happened in 2020-2023
The earlier shortage was, at its core, a demand-forecasting and logistics problem. Automakers canceled chip orders early in the pandemic, expecting reduced demand, then found themselves at the back of the queue when demand rebounded faster than anyone predicted. Combined with broader pandemic-related manufacturing and shipping disruptions, and supply chains optimized for lean, just-in-time inventory rather than resilience, the result was widespread shortages across multiple industries, with losses to the automotive sector alone estimated at roughly $500 billion globally. Critically, this was a temporary shock. Once logistics stabilized and new capacity came online, the shortage eased.
Why 2026 Is Different: A Structural Shortage, Not a Demand Shock
Two different shortages
2020-2023 shortage
- Root cause
- Pandemic demand shock, logistics failure, fragile just-in-time supply chains
- Nature
- Temporary spike
- Resolution path
- Resolved once logistics stabilized
- Primary driver
- Auto industry demand miscalculation
2026 shortage
- Root cause
- Structural, growing AI infrastructure demand
- Nature
- Ongoing capacity reallocation
- Resolution path
- Manufacturers indicate no clear near-term resolution
- Primary driver
- Memory and advanced chip capacity redirected toward AI servers
Memory manufacturers including Samsung, SK Hynix, and Micron have been reallocating fabrication capacity away from conventional DRAM and NAND production toward high-bandwidth memory (HBM) and server-class DDR5, the components AI infrastructure depends on. This isn't a temporary reprioritization while supply catches up; it reflects where the most profitable, highest-demand orders currently are, and manufacturers have been explicit that consumer and enterprise memory supply is likely to stay constrained as long as AI demand keeps growing at its current pace. The underlying economics of that demand are covered in the economics of training large AI models.
What's Actually Being Squeezed
Current market pressure
quarter-over-quarter increase in server DRAM contract prices forecast for Q3 2026 (TrendForce)
lead times for power semiconductors built on mature nodes, as capacity gets squeezed across the board
Intel CEO Lip-Bu Tan's public assessment of the broader semiconductor supply environment
SK Hynix's own warning on how long memory chip constraints could persist
The squeeze isn't limited to memory. Advanced packaging and back-end capacity, substrate availability, and enterprise storage are all reported to be under pressure as AI accelerator production consumes capacity across the wider component supply chain, not just the chips going directly into AI systems.
The Real Cost to IT Procurement Teams
A March 2026 survey of electronics procurement professionals by Fuld & Company and Accuris (n=439) found:
The procurement pain, by the numbers
of organizations report annual reactive decision costs exceeding $50,000
experience between 3 and 10 costly supply disruptions per year
increase in counterfeit electronic components reported in 2024 versus 2023, a risk that tends to rise alongside genuine shortages, as buyers turn to unverified sources under pressure
Lessons From 2020-2023 That Still Apply
Old lessons vs new lessons
Lessons that still hold
- Supplier strategy
- Diversify suppliers rather than relying on one vendor
- Inventory
- Build buffer stock for critical components
- Ordering model
- Avoid pure just-in-time ordering for critical hardware
- Channel
- Maintain relationships with multiple distributors
New lessons this shortage demands
- Supplier strategy
- Diversify by component architecture, not just vendor, since the shortage spans multiple chip categories
- Inventory
- Plan for structurally longer lead times, not just temporary padding, especially for memory-heavy hardware
- Ordering model
- Assess exposure even on hardware with no AI component, since capacity competition now extends across the wider supply chain
- Channel
- Watch counterfeit component risk specifically, which rises when legitimate supply tightens
A Practical Procurement Framework for Uncertain Supply
Procurement checklist for an uncertain chip market
- 01Extend lead-time assumptions for any hardware refresh involving memory-intensive components, and plan purchase timing well ahead of prior norms.
- 02Diversify sourcing across multiple vendors and, where feasible, multiple underlying chip architectures, rather than depending on a single supply relationship.
- 03Verify component authenticity through trusted distributors, given rising counterfeit risk in a tightening market.
- 04Reassess hardware refresh cycles: extending the life of existing equipment may be more realistic than assuming replacement hardware will be readily available on the old timeline.
- 05Build supply chain risk into IT budgeting conversations explicitly, rather than treating hardware availability as a given.
Hardware availability is now a planning variable rather than a background assumption, and it belongs in the same conversation as infrastructure strategy more broadly, including where workloads actually run.
Sources
- TrendForce, semiconductor and memory market pricing data, 2026
- Fuld & Company / Accuris, Electronic Parts Intelligence Survey, March 2026 (n=439)
- Public statements from SK Hynix, Micron, and Intel leadership, 2026 earnings commentary
This article reflects publicly available industry reporting as of July 2026. Semiconductor supply conditions are changing quickly, verify current lead times and pricing directly with suppliers before finalizing procurement plans.
Frequently Asked Questions
Is there really a new chip shortage in 2026?+
Yes. Memory manufacturers have reallocated fabrication capacity toward AI server components, tightening supply for enterprise and consumer components, with TrendForce data showing sharp price increases through 2026.
How is the 2026 chip shortage different from 2020-2023?+
The earlier shortage was a temporary demand shock from pandemic logistics failures. The current one is driven by structural, growing AI infrastructure demand that manufacturers don't expect to ease soon.
What components are most affected right now?+
Memory, particularly server DRAM and high-bandwidth memory, is the fastest-moving segment, with power semiconductors, advanced packaging, and enterprise storage also under pressure.
Does my business need AI chips to be affected by this shortage?+
No. Ordinary servers, laptops, and networking equipment can face longer lead times simply from competing for the same underlying component capacity AI infrastructure is consuming.
How long is this shortage expected to last?+
Estimates vary, with Intel's CEO citing no relief until 2028 and SK Hynix warning constraints could persist past 2030. New fabrication capacity isn't expected to yield meaningful output until 2027-2028.
What should IT procurement teams actually do differently?+
Extend lead-time assumptions, diversify suppliers and architectures, verify component authenticity, and reassess hardware refresh cycles rather than assuming availability will return to prior norms.