The global chip shortage that began in 2020 has largely eased for everyday consumer electronics, but it has not fully ended. Instead, it has shifted. The original crisis, where automakers and appliance manufacturers couldn’t get basic chips, has mostly resolved. In its place, a new and intense shortage of advanced processors and specialized memory for artificial intelligence has taken hold, and that one likely won’t ease before 2027 at the earliest.
Where the Shortage Stands Now
The chip market in 2025 looks very different from the peak crisis years of 2021 and 2022. Back then, car dealerships sat empty, game consoles were impossible to find, and even washing machine deliveries were delayed by months. Those shortages stemmed from a sudden surge in demand during the pandemic colliding with limited production capacity for older, simpler chips.
That wave has largely passed. Factories caught up, inventories were rebuilt, and most consumers can now buy cars, laptops, and appliances without chip-related delays. But the semiconductor industry hasn’t returned to a comfortable equilibrium. Deloitte’s 2025 industry outlook notes that efforts to diversify where chips are made, including building new factories in the U.S. and Europe, are still in early stages, leaving the industry “vulnerable for the next year or two, at least.” The supply chain is functional but fragile, and a major disruption (a natural disaster in Taiwan, an escalation in trade tensions) could tip parts of it back into shortage quickly.
AI Created a New Bottleneck
The most severe shortage right now involves the advanced chips that power artificial intelligence. The processors used to train and run AI models require cutting-edge manufacturing, enormous amounts of high-speed memory, and sophisticated packaging that bonds memory directly to the processor. Every one of those steps is constrained.
Demand from companies building AI data centers has been staggering. Major cloud providers have signed multi-year contracts locking up the entire output of some memory factories, reportedly claiming around 40% of global memory chip supply. Analysts project that data centers will consume up to 70% of the world’s memory supply by 2026. Memory manufacturers, chasing the higher profit margins on AI-grade chips, have reduced shipments of the standard memory used in gaming graphics cards and high-end PCs. In Japan, some PC vendors have already stopped taking orders for high-end desktops because they simply can’t get the components.
Even when factories produce enough processor chips, turning them into finished products hits another wall: packaging. Modern AI chips use a complex process that stacks memory on top of silicon wafers. The specialized packaging lines needed for this work are fully booked, and building new ones takes longer than adding raw chip-making capacity. So even if more processors roll off the line, there aren’t enough packaging facilities to assemble them into usable products.
New Factories Are Coming Online, But Slowly
Governments around the world have been pouring money into semiconductor manufacturing to reduce dependence on a handful of Asian factories. The most high-profile example: TSMC, the world’s largest contract chipmaker, officially began mass production at its advanced facility in Phoenix in 2025. That plant produces chips at the 4-nanometer level, the kind used in high-end smartphones and AI hardware. TSMC’s even more advanced 2-nanometer chips are entering volume production in Taiwan this year.
In the U.S., the CHIPS Act has funded dozens of projects beyond the TSMC plant. GlobalFoundries received funding to expand its New York facility with first-of-its-kind advanced packaging technology. Infinera is building a new factory in California and a packaging facility in Pennsylvania that will increase its domestic capacity by roughly tenfold. Supporting companies are scaling up too: Corning is boosting production of the ultra-pure glass needed for chip-making equipment, and Edwards Vacuum is building a new facility to produce the specialized pumps that semiconductor fabs require.
These investments are real, but semiconductor factories take three to five years to build and ramp to full output. Most CHIPS Act projects won’t reach high-volume production until 2027 or 2028. They’ll help prevent future shortages and reduce geopolitical risk, but they won’t resolve today’s bottlenecks.
Legacy Chips Carry a Different Risk
The older chip designs used in cars, medical devices, telecommunications equipment, and industrial controls occupy a separate category. These “legacy” or “mature node” chips aren’t cutting-edge, but they’re embedded in virtually every piece of critical infrastructure. The U.S. Bureau of Industry and Security published an assessment highlighting how deeply American industries depend on these chips, and how much of the supply traces back to manufacturers in China.
The immediate shortage of legacy chips has eased, but the supply chain concentration creates ongoing risk. If trade restrictions tighten or a geopolitical event disrupts production, the same industries that struggled in 2021 could face shortages again. This is less about a current shortage and more about a structural vulnerability that hasn’t been fixed yet.
A Realistic Timeline
For most consumers, the chip shortage already feels over. You can buy a car, a laptop, or a game console without waiting months. That part of the crisis ended in 2023 for most product categories.
For AI hardware, the picture is much tighter. The combination of explosive demand, limited high-speed memory, constrained packaging capacity, and the sheer complexity of advanced manufacturing means supply won’t comfortably meet demand until new factories and packaging lines finish ramping. Most industry projections place that somewhere in the 2027 to 2028 range, assuming demand growth doesn’t accelerate even further.
The ripple effects of the AI shortage are already reaching consumers in indirect ways: gaming graphics cards are more expensive because memory is being diverted to data centers, and high-end PC components face periodic supply constraints. Those pressures will likely persist through 2026. If you’re shopping for standard electronics, you’re fine. If you’re in an industry waiting on AI compute capacity, or building systems that depend on the latest processors, plan for tight supply for at least another two years.