in short
A McKinsey and SEMI Foundation analysis published in July 2026 puts the US semiconductor workforce shortfall at up to 157,000 people by 2030. Only about 3 percent of US engineering graduates go into chipmaking. And nearly three-quarters of chip employers say engineering roles are hard to fill. The industry can buy a lithography machine, sign a power contract and lease land in a quarter. It cannot buy people who already know how to run the factory.
Every other bottleneck in this series can, in principle, be bought. Capital can order more machines, finance more purification plants, and fund more water reclamation. The talent gap is different in kind: you cannot order experience. It has to be built one person at a time, on a clock that doesn't respond to the size of the check. That is why a shortfall measured in tens of thousands of workers can quietly decide whether a multi-billion-dollar fab opens on schedule or sits half-staffed.
This piece looks at what the 157,000 figure is actually made of, why it splits into two very different problems, and what happens when the obvious workaround, flying in experts from abroad, runs into the rest of the system.
What the 157,000 Is Actually Made Of
The headline number hides a structure. The SEMI Foundation and McKinsey estimate that US semiconductor manufacturing, design, materials and advanced packaging will need roughly 189,000 additional workers between 2026 and 2030, and that the gap between needed workers and available supply will be between 127,000 and 157,000 (Startup Fortune). The category breakdown matters more than the total. The report's table shows demand for 104,300 process and design engineers against an available supply of 16,300, a shortfall of 88,000, and demand for 72,400 equipment operations technicians against a supply of 8,900, a shortfall of 63,500 (BigGo Finance).
The number has also grown. An earlier study by the Semiconductor Industry Association and Oxford Economics, published in 2023, projected a shortfall of 67,000 technicians, engineers and computer scientists by 2030 (SIA). The two studies measure somewhat different scopes, so they aren't directly comparable. But the direction is unmistakable: three years of fab announcements have more than doubled the estimated gap, and the pipeline feeding it hasn't grown at the same pace. McKinsey's data shows only 3 percent of US engineering graduates enter semiconductors (Startup Fortune), and 73 percent of chip employers report serious difficulty hiring engineers (Briefs).
The Airline Analogy
Think of an airline ordering a hundred new aircraft. The planes can be built in a couple of years, financed in a few months, and delivered on a schedule. The captains to fly them cannot. A pilot takes years to train and more years to be trusted with a full command. An airline can poach captains from a rival, but that doesn't add a pilot to the world. It only moves one from one cockpit to another. A fab is the aircraft order. The workforce is the captain's seat, and the industry has just placed a historic number of orders.
Technicians: The Fast Pipeline
The more encouraging half of the gap is the technician side, because technicians can be trained relatively quickly and without a four-year degree. TSMC Arizona has built an apprenticeship system for exactly this. Its process technician track runs two years and combines roughly 600 hours of coursework with 4,000 hours of on-the-job training at the fab (NAU), and its equipment and facilities tracks run roughly 12 to 24 months depending on the specialty, with some tracks longer (TSMC Arizona). The company has told CNBC it needs to fill around 6,000 roles at its first three Arizona fabs, and has been recruiting from a dozen universities and its own hundreds of interns to do it (CNBC). In May 2026, Arizona State University and TSMC launched a new equipment technician program, adding a fifth school to the pipeline, with plans to hire more than 100 equipment technicians from it before year-end (Tucson Local Media).
This is genuine progress, and it took about two years of deliberate institution-building to assemble. It also shows the scale mismatch. A pipeline that produces on the order of a hundred equipment technicians from a new program, set against a national need measured in tens of thousands, is the right design running at a fraction of the required volume.
Engineers: The Slow Pipeline
The engineering gap is the harder problem, because it can't be shortened by an apprenticeship. A process engineer typically needs a four-year degree, often an advanced one, plus years of hands-on experience before being trusted with a production line. The numbers explain why the shortfall there is so large: a required 104,300 against an available 16,300 means demand is more than six times the available supply (BigGo Finance).
Competition makes it worse. The same AI boom creating the demand for chips is pulling engineers toward software and AI labs, which pay more and carry more prestige. Startup Fortune's summary of the SEMI findings notes that AI labs are competing for the same talent pool and driving pay higher (Startup Fortune). The industry isn't only short of engineers. It's competing for them against the sector its output exists to serve.
The Know-How That Isn't Written Down
Think of a sourdough starter. The recipe is simple and fully documented. But the loaf depends on a living culture, maintained for years by someone who knows by feel when it's ready. Copy the recipe without the starter, or without the baker, and you get bread that isn't quite right. Fab know-how works the same way. A process is not just a document. It's a body of tuned judgment held by the people who ran it.
That is why the standard solution for opening a fab in a new country is to move the people. TSMC flew hundreds of engineers from Taiwan to staff its Arizona facilities (The Tech Buzz). Samsung and SK Hynix have told CNBC they are bringing workers over from South Korea temporarily to launch their new US fabs, and Samsung sent its own US employees to Korea for months of hands-on training, echoing a similar move by TSMC in 2021 (CNBC). The knowledge transfer is physical. It travels with the people.
The Workaround Meets the System
Importing expertise works only as far as the immigration and labor systems around it allow, and in the past year that boundary has become visible. On September 4, 2025, federal agents detained 475 people at a Hyundai and LG Energy Solution battery plant under construction in Georgia, among them roughly 300 Korean nationals (NBC News). That was a battery plant, not a chip fab, but its effects reached chipmaking quickly. Samsung and SK Hynix both moved to cap business trips to the United States under the visa waiver program (Korea JoongAng Daily), and a Korean equipment supplier to Samsung's Taylor, Texas foundry described a looming visa bottleneck for sending technicians to install tools there (Korea JoongAng Daily).
The friction predates the raid. Korean companies had reported persistent trouble getting visas for technical staff for some time, with workers describing the H-1B route as extremely difficult and some short-term business visas rejected without explanation (Reuters, via Yahoo; Atlanta Journal-Constitution). The other side of the tension is local: when TSMC requested 500 visas for Taiwanese engineers to install equipment at its Arizona plant, local labor unions protested (Korea JoongAng Daily). Both positions have legitimate logic. Local workers want the jobs the new fabs create, and the fabs need specialists who don't yet exist locally. The point for this piece isn't to adjudicate that debate. It's that the workaround for the talent gap is itself constrained by systems that weren't designed for a surge of equipment-installation specialists.
The Money Isn't the Problem
Wages show the market has noticed. TSMC has offered signing bonuses above $50,000 to lure technicians from rivals, and entry-level US fab technicians now start above $60,000 (The Tech Buzz). In the memory sector, CNBC reports that when thousands of workers threatened strikes at Micron, Samsung and SK Hynix, bonuses ballooned to upwards of $500,000 (CNBC). IEEE Spectrum's 2023 reporting found that Intel had struggled since the pandemic to find enough operators and technicians with two-year degrees to keep its fabs running (IEEE Spectrum).
Notice what higher pay does and doesn't do. It moves workers between employers, which is useful for whoever wins the bidding and neutral for the industry as a whole. It can encourage some people to enter the field, but that effect arrives on the training clock, not the pay clock. In the short run, a bonus can redistribute the talent that exists. It can't create more.
Why It Slows Everything Else
The talent gap rarely shows up as a headline of its own. It shows up as a delay somewhere else. TSMC's Arizona buildout has faced construction setbacks attributed to a shortage of skilled engineers (BigGo Finance), and Intel's Ohio site needed roughly 7,000 construction workers in a region that labor officials said did not have them (CSIS). A fab that has secured its lithography allocation, its materials and its water can still sit idle if the people to commission it aren't there. That is the quiet way this bottleneck compounds the others: it determines how quickly every scarce resource the fab has already paid for starts earning its keep. McKinsey's analysis notes that the timeline from groundbreaking to initial production typically runs about 24 months in the United States against 12 to 16 months in Taiwan and China, with labor accounting for roughly half of the cost gap (BigGo Finance).
The newest entrants face the same arithmetic. SpaceX says Terafab will create at least 1,800 permanent jobs at its Grimes County site, many of which it expects to fill locally (Wikipedia; Spectrum News). A project that wants to build the largest fab on Earth will be hiring in the same market, for the same scarce process engineers, as every established chipmaker.
The Challenge: This May Be the Most Solvable Bottleneck of the Four
Fairness requires acknowledging that on paper, the talent gap looks like the most tractable constraint in this series. Technician training takes 12 to 24 months, not decades. TSMC Arizona built a multi-school training system in roughly two years. Employers are now partnering with community colleges, running paid apprenticeships and expanding programs at pace. If the gap were only technicians, it is easy to imagine it closing within the decade.
Two things keep that from being the whole story. First, the engineering shortfall, the larger half, runs on a multi-year degree-and-experience clock that no apprenticeship shortens, and the engineers needed for fabs opening in 2026 through 2028 would have had to begin training years ago. Second, even the technician pipeline faces a timing mismatch: the capacity coming online now was announced years before the training systems matched it, and the gap between those two clocks is exactly where delays occur. Whether automation could eventually narrow the labor requirement is a fair open question, but none of the sources reviewed for this piece quantify it, so it remains a variable rather than a plan.
Conclusion
The talent gap is the one bottleneck where the scarce resource is a person's accumulated judgment, and that can only be built over time. It is partly solvable and partly not, and the part that isn't, experienced process engineers, is the part the fabs need first. Money can move the existing talent around, imported expertise can bridge a few years of gap, and training programs can eventually raise the supply. None of these changes the underlying arithmetic: the factories are being built faster than the people who run them are being made.
This piece expands the third bottleneck in The Four Things Standing Between Us and Infinite AI Compute. Its companions cover the materials chokepoint and the water problem.
Sources
source
covers
SEMI Foundation/McKinsey 127,000–157,000 shortfall; 189,000 additional workers; AI-lab competition
Category-level demand, supply and shortfall table
3% of engineering graduates; 73% of employers struggling to hire
2023 estimate of a 67,000-worker shortfall
TSMC's 6,000 Arizona roles; Korean workers temporarily in US fabs; strike-driven bonuses
Two-year program structure (600 hours coursework, 4,000 hours on-the-job)
Apprenticeship tracks and timelines
ASU equipment technician program; five-school pipeline
TSMC flying in Taiwanese engineers; limits of that approach
Signing bonuses; starting wages
September 2025 Georgia plant detentions
Samsung and SK Hynix travel caps
Visa bottleneck for Taylor; TSMC's 500-visa request and union protests
Visa difficulties for Korean technical workers
Pre-raid visa struggles
Intel's difficulty finding two-year-degree technicians
Terafab's stated job creation
Local-hiring expectations at Terafab
