Money and factories are the most visible parts of the US push to bring semiconductor manufacturing back home.
From Arizona to Texas, and from Ohio to New York, fabs, advanced packaging plants and materials facilities have been announced one after another. In its latest report, the Semiconductor Industry Association, or SIA, said more than $770 billion in semiconductor supply-chain investment has been announced in the US since 2020, spanning 30 states and 160 projects.
Capital can build factories. Equipment can be purchased from global suppliers. The harder issue shows up when production lines are ready to start: whether there are enough people to tune tools, run processes, maintain yield and keep plants operating 365 days a year.
In its 2026 State of the U.S. Semiconductor Industry report, SIA said the sector directly employs about 342,000 workers and supports nearly 2 million indirect and induced jobs. On average, one semiconductor job supports about 5.7 jobs in other industries, and pay in the sector runs more than 30% above comparable roles elsewhere. Even so, the US could still face a shortage of 67,000 semiconductor technicians, engineers and computer-related workers by 2030.
That number is close to one-fifth of the industry’s current direct workforce. At that scale, talent stops being a routine HR problem and becomes a question of whether the reshoring push can actually be delivered.
Where the 67,000 figure comes from
The number is not the result of a new labor census completed in 2026.
It comes from a forecasting model released in 2023 by SIA and Oxford Economics, and the latest report continues to cite it. Under assumptions based on then-current graduation rates and labor-supply trends, the model estimated that the US semiconductor industry would add about 115,000 jobs by 2030, with roughly 67,000 at risk of going unfilled.
So the 67,000 figure does not represent jobs that are already vacant, nor is it a number guaranteed to materialize with precision. AI-driven productivity gains, changes in factory investment, higher levels of automation, market cycles and expanded training programs could all shift the final outcome.
What the estimate still captures is the underlying tension: US capacity expansion may be moving faster than the workforce pipeline can replenish talent.
The shortage is not only about elite chip designers
When people hear “semiconductor talent shortage,” they often think first of top engineers designing CPUs and GPUs. That is only part of the picture.
Based on the SIA and Oxford Economics breakdown, of the projected 67,000-worker gap, about 26,400 are technicians, or 39%; around 27,300 are engineers, or 41%; and another 13,400 are in computer science-related roles, or 20%.
Even within engineering, the needs are split. The projected gap includes about 9,900 positions requiring a bachelor’s degree, 12,300 requiring a master’s degree and roughly 5,100 requiring a PhD. In other words, the US lacks both the technicians who can operate and maintain equipment in cleanrooms and the highly trained engineers needed in process technology, materials, devices and chip design.
The training path for those two groups is very different.
Advanced process R&D, transistor structures, EDA algorithms and materials research often require graduate-level training. Companies cannot produce engineers capable of handling critical research work through a few months of accelerated classes.
Technician roles have a lower formal education threshold. The US Bureau of Labor Statistics said semiconductor processing technicians can often start with a high school diploma, while some jobs require a certificate or an associate degree, followed by on-the-job training. The entry point may look broader, but that does not make the positions easy to fill.
A fab does not run on a handful of engineers in clean suits adjusting parameters from a terminal. Production lines need equipment staff, manufacturing operators, quality personnel, automation teams, facilities staff and maintenance workers. Once a process moves into volume production, those roles determine tool utilization, production cadence and yield.
Top engineers decide how far a process can go. Technicians decide whether a factory runs properly each day. The shortage exists on both sides, even if it appears in different ways.
Why technician jobs remain hard to fill
US Bureau of Labor Statistics data show there were about 31,900 semiconductor processing technicians in the US in 2024. Employment in that occupation is expected to grow 11% by 2034, well above the 3% average for all occupations in the country. Over the next decade, there are projected to be about 3,900 openings a year on average, including both new demand and replacement demand from retirements and career changes.
These are not classic low-wage factory jobs. In 2024, the median annual pay for semiconductor processing technicians was $51,180, above the $45,960 median for production occupations in the US. In semiconductor and electronic component manufacturing, median annual pay for the role was about $52,230, and the top 10% earned more than $87,000.
Pay, though, is only one part of the hiring equation.
Fabs usually run continuously, and shift work, nights and weekends are common. New fabs are also concentrated in specific industrial clusters. Workers are not only deciding whether to join semiconductor manufacturing; they are also deciding whether to move to where the plant is located, and whether local housing, transportation, schooling and childcare are manageable.
That helps explain why the US Department of Commerce has required labor-force development plans in CHIPS-related applications. Companies seeking larger direct subsidies also need to provide plans for affordable childcare access for employees. The requirement may sound far removed from chip technology, but it reflects a practical reality: if factories need rotating shifts and local public services do not match that schedule, strong training programs alone may not keep workers in place.
That is why the technician shortage cannot be reduced to a simple idea that younger workers do not want factory jobs, nor can it be solved just by opening a few vocational training classes.
Companies also have to answer basic questions: whether jobs offer a clear promotion path, whether credentials can be recognized across employers, who pays for training, and whether workers want to remain in the area for the long term. At a fab costing tens of billions of dollars, those details eventually show up in attrition rates and production efficiency.
For engineers, the issue is not simply salary
The engineering labor market looks different.
The US Bureau of Labor Statistics expects employment for electrical and electronics engineers to rise 7% from 2024 to 2034, with roughly 17,500 openings a year on average. In 2024, the median annual wage for electronics engineers in semiconductor and electronic component manufacturing reached $142,760.
Computer hardware engineers were paid even more. In 2024, the nationwide median annual wage for the occupation was $155,020, while the median in semiconductor and electronic component manufacturing was about $162,460. Employment is also projected to grow 7% over the next decade, with about 4,700 openings a year.
Those wages are already attractive. The shortage persists because companies are competing not for generic labor, but for people with highly specific experience.
A process engineer who can improve yield in advanced-node manufacturing cannot be replaced by a general mechanical engineer. Packaging specialists familiar with HBM stacking, hybrid bonding and thermal management are also not easily substituted by new graduates.
There is another constraint. Semiconductor companies are not the only destination for this talent pool. Artificial intelligence, cloud computing, software, aerospace, automotive electronics and the defense industry are all competing for workers trained in electrical engineering, computer science and materials. Chip companies do not have a monopoly on high pay.
As a result, the US engineering gap is both a numbers problem and a matching problem. Universities can increase enrollment in electrical engineering and computer science, but it often takes years for students to finish master’s or doctoral training and then accumulate industry experience. Hiring peaks for new fabs, by contrast, may arrive within the next three or four years.
$770 billion in investment is redrawing the talent map
The US semiconductor workforce has never been spread evenly across the country.
Chip design, wafer fabrication, equipment and materials companies have long clustered in a limited number of hubs. Now that new projects span 30 states, they are expanding traditional bases while also bringing large factories into regions that historically had a thinner semiconductor footprint.
That creates a mismatch that is easy to miss: having enough engineering graduates nationwide does not mean a given project location can hire enough people locally.
A company can recruit dozens of senior managers and core experts on a national basis. High-volume manufacturing at a fab, however, requires thousands of local employees who can stay for the long run. If companies rely heavily on recruiting workers from other states, the bill extends beyond wages to relocation expenses, housing subsidies and the costs of settling entire families.
That is one reason community colleges are taking on a larger role in the US semiconductor workforce strategy.
By the time SIA published its 2024 workforce policy blueprint, more than 50 community colleges in the US had announced new or expanded semiconductor-related programs. Universities mainly train researchers and engineers. Community colleges are better placed to prepare equipment technicians, production technicians and engineering assistants for local fabs. The CHIPS and Science Act also set aside a $200 million Semiconductor Workforce and Education Fund, with the National Science Foundation leading related training efforts.
The model is straightforward in theory: companies define job needs, community colleges train toward those roles, and students move into local plants. That should be more durable than competing for workers across the whole country.
The difficulty is in execution. Courses cannot stop at broad semiconductor concepts. Different fabs use different tools, process platforms and automation systems, so schools need company participation, training facilities and constantly updated curricula. Semiconductor equipment is expensive, and many schools cannot build full cleanrooms on their own. In discussions about future semiconductor workforce development, the National Science Foundation has already identified the lack of costly cleanrooms and advanced equipment as a real barrier for training programs.
That means workforce development is itself an infrastructure investment, not just a curriculum exercise.
Domestic training and international talent are not mutually exclusive
Among SIA’s policy recommendations, alongside expanded STEM education, vocational training and apprenticeships, is reform for high-skilled immigration.
The reason is fairly direct. Technician shortages can be addressed more through community colleges, vocational education and on-the-job training. Engineers at the master’s and PhD levels are much harder to scale quickly. US universities have long attracted international students in engineering and computer-related fields, and some of those graduates have already completed the necessary training in the country. Allowing them to move into industry can help fill near-term high-end talent needs.
Still, international talent is not a complete solution.
Immigration policy carries uncertainty, and companies cannot build an entire talent strategy on overseas hiring. Many advanced manufacturing positions also require employees to stay at specific plant locations for long periods, and not every international graduate wants a factory-based role. Over the long run, the US still needs to expand domestic engineering education, vocational programs and industrial training systems.
There is little value in treating those two paths as opposites. Domestic training addresses long-term supply and broader industrial coverage. International talent can ease near-term pressure in advanced engineering roles. A mature industrial policy needs to deal with both time horizons at once.
What this may mean for China
The US response to semiconductor labor shortages also offers some reference points for China. The more important point is not how much money Washington allocates to training programs, but that workforce supply is being treated as part of semiconductor infrastructure, planned alongside factories, research spending and subsidies.
That has practical relevance for China as well.
In recent years, semiconductor projects in China have expanded across wafer manufacturing, packaging and testing, materials and equipment bases in multiple regions. But project rollout and labor supply do not always line up. Some areas have capital, land and policy support but lack mature engineering teams. Some universities graduate large numbers of related majors, yet their courses and practical training conditions still sit at a distance from actual job requirements in industry. A larger national talent pool does not guarantee that every industrial cluster can find the right people on time.
One part of the US policy approach that stands out is requiring companies receiving industrial support to submit workforce training and employment plans at the same time. For China, project evaluation could move beyond metrics such as investment size and planned capacity and extend to questions including what roles are needed, where talent will come from, how it will be trained and how workers will be retained. For large fabs and advanced packaging projects in particular, stable cooperation with local universities, vocational schools and training institutions may need to begin at the project approval stage rather than in a rush one year before production starts.
The role of vocational education also needs a fresh look.
Semiconductor talent does not mean everyone must hold a master’s or PhD. Advanced-node R&D, chip architecture, EDA and key materials research do depend on highly educated specialists. Manufacturing sites, though, also need large numbers of equipment maintenance workers, production control staff, quality managers and automation technicians. Those positions are better suited to vocational schools, application-oriented undergraduate programs and enterprise training systems. For schools, the question is not only whether to add more integrated-circuit majors, but whether courses resemble actual R&D and manufacturing environments, whether students can work with equipment, process flows and quality systems, and whether companies are willing to stay involved in teaching over the long term.
Workforce policy also has to address the practical side of regional mobility and career development.
Semiconductor plants are often located away from the urban cores where internet and finance talent tends to cluster. Whether employees stay for the long term depends not only on pay, but also on housing, education, healthcare, commuting and room for promotion. A region may be able to attract projects with incentives, but one-off subsidies alone rarely build a stable talent ecosystem. For local governments, schools, training centers and public services may be less eye-catching than signing a major factory project, yet they often matter more once production begins.
China also needs to avoid treating semiconductor workforce policy as a simple contest to poach talent. Hiring people away from other companies and regions at higher salaries may relieve pressure for one project, but it does not expand the industry’s total talent supply. It can instead raise costs and increase turnover. A more durable route is still to expand training capacity, improve engineering education and ask companies to shoulder more responsibility for on-the-job training.
The US projection of a 67,000-worker shortage may not fully materialize, and China is unlikely to capture future demand in one perfectly accurate figure either. AI and automation may reduce some labor needs, while industry cycles and project adjustments will reshape hiring plans. Even so, one point is clear: the more semiconductors become a contest in advanced manufacturing and system-level execution, the less workforce development can lag behind capacity expansion.
For China, a mature semiconductor industry will not be defined only by how many fabs are built or how many tools are imported. It will also depend on whether local systems can keep producing engineers and technicians, support stable factory operations, improve processes over time and form talent ladders that renew themselves from one generation to the next. Building the line is only the start. An industry base is established only when talent supply can sustain itself.
This article was originally published by the WeChat public account TechSugar and written by 谈芯.