The global semiconductor shortage began in 2020 as a collision of pandemic disruptions, surging demand, and years of underinvestment in manufacturing capacity. What started as a supply chain hiccup snowballed into a crisis affecting everything from cars to gaming consoles, and while the worst of that initial crunch has eased, new pressures from artificial intelligence are creating fresh shortages in 2025.
COVID-19 Set Off a Perfect Storm
The pandemic hit semiconductor supply chains from both sides at once. Factories in South Korea and Taiwan, where most of the world’s chips are made, faced shutdowns and slowdowns due to lockdowns and worker illness. At the same time, global demand for PCs jumped 13% as millions of people shifted to working and learning from home. That combination of constrained supply and spiking demand created a bottleneck that took years to untangle.
But COVID-19 was really just the accelerant. The underlying problem was that chipmaking capacity had been running near its limits for years, with demand growing faster than new factories could be built. The pandemic exposed how fragile the system already was. On top of that, the U.S.-China trade war disrupted established supply relationships, and a severe drought in Taiwan, the worst rainfall the island had seen since 1947, threatened water supplies at chip factories that consume enormous quantities of ultrapure water in manufacturing.
Just-in-Time Inventory Left No Buffer
Many industries, especially automakers, had adopted “just-in-time” inventory systems designed to minimize the cost of holding extra parts. Instead of stockpiling chips, they ordered them as needed and kept only days or weeks of supply on hand. This works beautifully when supply chains are stable. When they aren’t, it’s catastrophic.
When auto factories shut down early in the pandemic, they cancelled chip orders. Chipmakers quickly redirected that capacity to meet booming demand from consumer electronics companies. When car production restarted months later, automakers found themselves at the back of the line with no inventory buffer to fall back on. This triggered what supply chain experts call the bullwhip effect: small shifts in demand at the consumer end snowballed into massive fluctuations upstream, creating shortages that rippled across industries for the next two years.
Older Chip Technology Was Hit Hardest
Most of the shortage headlines focused on cutting-edge processors, but the real bottleneck was in older, less glamorous chip designs. The chips that control your car’s brakes, run your washing machine, or manage power in medical devices are typically built on manufacturing processes from the 2000s and 2010s, using what the industry calls “mature nodes” at sizes like 28 or 40 nanometers.
Nobody had been rushing to build new factories for these older chips because the profit margins are thin compared to advanced processors. A 2024 report from the U.S. Bureau of Industry and Security found that there are no commercially available, cost-effective foundries in the United States capable of producing chips at 22 or 40 nanometers with sufficient capacity. China is expected to account for nearly half of all new mature-node manufacturing capacity over the next few years, and some chip companies have reported they simply cannot find non-Chinese foundries willing to take on their orders for these older designs.
This creates a geopolitical vulnerability on top of the supply problem. The same chips that go into American cars, appliances, and defense systems are increasingly dependent on Chinese factories, with capacity for chips in the 20 to 40 nanometer range expected to more than triple in China based on announced investments.
New Factories Take Years to Build
You can’t solve a chip shortage quickly because semiconductor factories, called fabs, are among the most complex and expensive facilities ever built. A single advanced fab costs tens of billions of dollars and takes three to five years to go from groundbreaking to producing chips. The equipment inside is equally constrained. ASML, the Dutch company that makes the machines used to print chip designs onto silicon wafers, produces a limited number of its most advanced systems each year. These extreme ultraviolet lithography machines cost around $200 million each.
Even when companies commit to building new capacity, delays are common. Samsung announced a $17 billion factory in Taylor, Texas, originally set to begin production in 2024. That timeline slipped to 2026, with Samsung delaying deliveries of ASML equipment and holding off on orders to other suppliers. Workers deployed to the site were sent home while the company reassessed its plans.
Government Efforts to Expand Capacity
The shortage prompted governments worldwide to invest in domestic chip production. In the United States, the CHIPS Act provides $50 billion to the Department of Commerce for semiconductor research, development, and manufacturing, along with authorization for up to $75 billion in loans and loan guarantees. As of January 2025, the program had made 20 awards totaling up to $33.7 billion in direct funding and $5.5 billion in loans, with an additional $8.3 billion going toward a national semiconductor technology center and research projects.
These investments are significant, but they represent a long-term fix rather than an immediate one. Most of the new fabs funded through these programs won’t reach full production for several more years.
AI Is Creating a New Wave of Shortages
Just as the original pandemic-era shortage began to ease in 2023, a new source of demand emerged. The explosion of generative AI has created enormous appetite for advanced chips, and chipmakers have pivoted their production to meet it. That pivot is now squeezing supply for everything else.
The most acute pressure point in 2025 is memory chips. The AI systems running in data centers require specialized high-bandwidth memory, and manufacturers have shifted production away from the standard memory chips used in phones, PCs, and USB drives. The results are stark: average inventory levels for the main type of computer memory fell from 13 to 17 weeks in late 2024 to just two to four weeks by October 2025. Samsung told customers it planned to end production of an older variety of memory chips used in PCs and servers. Chinese smartphone makers have warned of coming price increases.
The shortage now spans nearly every type of memory, from flash chips in consumer devices to the high-bandwidth chips feeding AI servers. Industry analysts have described it as a macroeconomic risk rather than just a component-level concern, warning it could slow AI-driven productivity gains, delay hundreds of billions of dollars in digital infrastructure, and add inflationary pressure across economies already navigating tariffs and price instability.
Where Things Stand Now
The semiconductor industry is in an uneven recovery. For advanced chips used in AI, demand far outstrips supply and factories are running above 90% capacity utilization. For mature-node chips used in cars, appliances, and industrial equipment, the picture is more mixed. After a period of oversupply and correction in 2024, utilization rates at older factories are climbing back toward 75 to 76%, driven by gradual recovery in consumer electronics and inventory restocking in the automotive sector.
The fundamental problem hasn’t changed: the world relies on a small number of factories in a small number of countries to produce chips that go into nearly everything electronic. Building new capacity is slow, expensive, and subject to delays. And each new wave of technology demand, whether it’s remote work PCs or AI servers, has the potential to tip the system back into shortage. The semiconductor supply chain is more resilient than it was in 2020, but it remains one unexpected disruption away from the next crunch.