After the graphics card crunch and the memory shortage, the semiconductor supply chain has found another chokepoint: the humble CPU.
In recent weeks, both Intel and AMD have quietly warned customers of delivery delays that would have seemed unthinkable six months ago. Previously, the average lead time for a CPU was around one to two weeks, but now the wait time has prolonged to an average of eight to 12 weeks, with some stretching to six months, according to manufacturer executives cited by Nikkei Asia. Manufacturers like HP and Dell are now saying that the number of processors that get delivered to them no longer matches the required volume, with some sources saying that the situation is getting worse compared to some months back.
The situation is dire enough that industry executives expect the situation to deteriorate further in the April-June quarter. Prices are climbing as well. Costs are rising by an average of 10% to 15%, if not more.
This is not a random bottleneck. The root cause is structural and familiar by now: artificial intelligence. Surging demand for agentic AI systems, which perform complex multi-step operations beyond simple chatbot functionality, has further strained supply, as these advanced applications require significantly more CPU processing power than traditional workloads.
Let's be real: CPUs are the unglamorous workhorses that power data centre operations. Compared to GPUs, they have received little attention during the AI boom. Tech companies bought up as many GPUs as they could when AI LLMs were popularised by ChatGPT, resulting in the GPU shortage between 2023 and mid-2025. But as AI moves beyond training large language models toward systems that reason and act autonomously, the CPU's role has become critical. Data centres now need CPUs alongside GPUs to execute the multi-step operations that agentic AI demands.
The supply crisis has two components. Intel faces manufacturing yield problems at its own fabrication plants. AMD, which outsources to Taiwan's TSMC, faces a different problem: TSMC has prioritised AI chip manufacturing and left limited capacity for CPUs. Both companies are rationing what they can produce.
The consequences are becoming visible. One gaming PC company executive told Nikkei Asia that supply for consumer PCs is being squeezed as chipmakers prioritise server processors. This reflects a rational but harsh economic reality. Enterprise and hyperscaler customers willing to pay premium prices get priority; the consumer market gets the remainder.
There is a genuine tension here between competing interests. Hyperscalers investing hundreds of billions in AI infrastructure represent the future of computing, and they need CPU supply to build out their systems. But consumer PC makers depend on affordable, reliable processor supply to manufacture products for ordinary users. The supply crunch puts manufacturers in an impossible position: either absorb higher costs or pass them on to consumers.
Intel has struggled to ramp up production amid persistent manufacturing yield challenges, a problem that cannot be solved quickly. TSMC's capacity constraints are even more fundamental; expanding cutting-edge foundry capacity takes years and billions in capital investment. The shortage is unlikely to ease significantly before late 2026 at the earliest.
For PC manufacturers and consumers, the takeaway is straightforward: expect continued pressure on CPU availability and pricing through the remainder of 2026. The AI boom has restructured the semiconductor industry's priorities, and traditional consumer computing sits lower on that new hierarchy. This is the cost of the AI supercycle.