At a Glance: CPU prices are up 10–15% across Intel and AMD product lines since January 2026. Global supply is meeting only 65–75% of demand. Average lead times have stretched from 1–2 weeks to 8–12 weeks, with enterprise Xeon and EPYC orders now facing waits of up to six months. Source: Nikkei Asia, March 11, 2026.

The New Bottleneck

For much of 2025, the semiconductor component that system builders and data center operators feared running short of was memory — specifically High Bandwidth Memory (HBM) and DRAM, both of which constrained AI server assembly across the industry. In the first quarter of 2026, that constraint has migrated upstream: the new bottleneck is the CPU itself.

Intel and AMD central processing unit prices have increased by up to 15% since January 1, 2026, according to industry sources reporting to Nikkei Asia on March 11. The price increases span both consumer and enterprise product lines — affecting the Core i7/i9 and Xeon Scalable families from Intel, and the Ryzen 7/9 and EPYC 9004/9005 series from AMD.

Major PC brands including HP and Dell have issued internal warnings to channel partners about extended lead times and upward price adjustments, with enterprise customers seeking new server deployments for AI infrastructure now receiving quotes with delivery windows of 90 to 180 days.

Scope and Scale of the Shortage

Industry analysts estimate that current CPU production capacity is meeting only 65–75% of global demand for high-performance processors in Q1 2026. The gap between demand and supply is most acute in two segments: mid-to-high-end desktop CPUs — the Core i7/i9 and Ryzen 7/9 tiers used in gaming PCs, creative workstations, and developer machines — and server-grade processors including Intel's Xeon Scalable and AMD's EPYC 9004 and 9005 series, which are critical for data center and AI training deployments.

Why CPUs Are Running Short

The shortage is the product of several converging pressures that have each individually been building for over a year, and have now arrived simultaneously.

AI Server Demand

The dominant driver is the explosive demand for AI-accelerated servers. Training and inference workloads require not just AI accelerators (GPUs, TPUs) but high-core-count CPUs to manage data pipelines, preprocessing, and orchestration. The buildout of hyperscale AI infrastructure — accelerated by initiatives like Project Stargate and similar efforts from Microsoft Azure, Google Cloud, and Amazon Web Services — has created a sustained, large-volume pull on server-class CPUs that manufacturing capacity was not configured to meet at this pace.

The council of technology executives appointed to the White House PCAST AI advisory body this week — which includes both Jensen Huang of NVIDIA and Lisa Su of AMD — reflects how central compute infrastructure has become to federal technology policy. The shortage they are now convening to advise on is partly a product of the infrastructure buildout their own companies have been driving.

PC Market Recovery

The traditional PC market has been in a prolonged post-pandemic correction since 2022, with consumer demand running well below the abnormal peaks of 2020–2021. That correction has largely run its course. Consumer PC demand has been recovering steadily since mid-2025, adding a concurrent pull on desktop and notebook CPU inventory at the same time that AI server demand is surging — a combination that supply chains were not positioned to absorb.

Advanced Node Constraints

AMD's Ryzen and EPYC processors are manufactured on TSMC's 4nm and 3nm process nodes — capacity that is simultaneously in demand for smartphone SoCs, AI chips, and discrete GPUs. TSMC has continued expanding, but the pace of new fab capacity coming online has not kept pace with the breadth of competing demand across its customer base.

Intel is in an additional bind: its next-generation 18A and 20A manufacturing processes are still in their ramp phase, meaning the company is absorbing the costs and logistical complexity of a major process transition while simultaneously trying to fulfill elevated demand on its existing facilities.

Compounding Memory Shortages

CPU scarcity has arrived on top of — not instead of — the memory chip shortages that dominated industry coverage in late 2025. DRAM and HBM availability remains constrained. Server OEMs now report that CPU allocation has surpassed HBM as the primary bottleneck in fulfilling enterprise orders. System builders are, in some cases, sitting on memory inventory they cannot turn into complete systems because the processor to pair it with is unavailable.

Impact on PC and Server Manufacturers

For PC manufacturers, the cost pressure is working its way toward consumers. Notebook and desktop systems in the $800–$1,500 range — covering mainstream gaming PCs, professional workstations, and mid-range laptops — are projected to see retail price increases of 5–12% by Q2 2026 as manufacturers pass on elevated component costs. Budget system tiers below $600 are less immediately affected, as that segment relies more heavily on lower-end SKUs with somewhat less constrained supply.

For enterprise customers, the impact is more severe. Data center operators planning new deployments for AI workloads are being quoted 90-to-180-day delivery windows. For organizations trying to meet quarterly infrastructure targets — a standard operating cadence for cloud and enterprise IT — that timeline represents a significant operational disruption. Some have begun exploring AMD-as-Intel-substitute and vice versa where architecture compatibility allows, attempting to source from whichever vendor has shorter lead on a given week.

The Broader Picture: A $627 Billion Industry Under Strain

The CPU shortage is the latest chapter in what has become a multi-year, multi-commodity semiconductor supply chain recovery. Global semiconductor revenue reached $627 billion in 2025, up 19% year-over-year — a figure that reflects real demand strength but also highlights the scale of the capacity gap that the industry has been racing to close since the 2021–2022 crunch.

New fabrication capacity — from TSMC's Arizona expansion, Intel's Ohio and Germany fabs, and Samsung's Texas buildout — is progressing. But advanced node fabs take three to five years to build and qualify from groundbreaking to volume production. The capacity being added today reflects decisions made in 2021 and 2022; the decisions being made today will not reach the market until 2028 or later.

Geopolitical factors add additional friction. Ongoing export control frameworks restrict the transfer of certain advanced semiconductor manufacturing equipment — primarily extreme ultraviolet lithography machines — to Chinese foundries, keeping the global advanced-node supply base concentrated among a small number of manufacturers in Taiwan, South Korea, and the United States.

The net result is a supply chain that is structurally unable to respond quickly to demand spikes in any single commodity. When HBM ran short, CPUs were sufficient. Now CPUs are the constraint. The question industry analysts are beginning to ask is not which component runs short next — but whether the cadence of single-commodity bottlenecks has become a permanent feature of the AI infrastructure buildout rather than a temporary correction.