The Semiconductor Workforce Paradox: 157,000 Jobs We Cannot Fill While Thousands Are Laid Off

America is simultaneously laying off experienced semiconductor workers and warning that it cannot find enough people to run its new fabs. How can this be? 

Something does not add up.

The United States is undertaking its fastest semiconductor manufacturing expansion in generations. New fabs are rising in Arizona, Texas, Ohio, New York and Idaho. Advanced packaging capacity is expanding. AI is driving extraordinary demand for logic, memory, interconnect and increasingly sophisticated silicon.

Yet the country may be short as many as 157,000 semiconductor and microelectronics workers by 2030, according to analysis from McKinsey and the SEMI Foundation. Only 3% of U.S. engineering graduates who enter engineering roles choose the semiconductor industry, while 73% of chip employers report significant difficulty filling engineering positions.

At the same time, semiconductor companies have laid off thousands of experienced people.

So which is it?

Does America have too few semiconductor workers?

Or too many?

The answer appears to be both! Maybe the apparent contradiction disappears once we stop treating the semiconductor workforce as a single pool of interchangeable people.

Maybe we do have experienced people in the wrong disciplines, companies, and locations. Maybe new graduates choose software and AI over hardware because those jobs seem more dynamic and offer better growth potential. Maybe it's because advanced-manufacturing knowledge is concentrated in Asia.

We're building fabs faster than experience can be created. And companies are removing the very people who once carried knowledge across organizational boundaries. And we are trying to solve all of that while AI changes the work itself.

There Is No Single “Semiconductor Worker”

The first mistake is treating semiconductor employment as one homogeneous labor market. It is not. And this is where the large think tanks we rely on for these macro insights can create more confusion than clarity. A process-integration engineer who can bring up a leading-edge fab is not interchangeable with an RTL designer. An experienced lithography technician cannot simply become a physical-design engineer. A verification engineer cannot automatically operate advanced-packaging equipment. An analog designer is not necessarily the person needed to maintain an EUV process. And someone with 25 years of semiconductor product, systems, or customer experience cannot simply be dropped into a 12-hour manufacturing shift in Arizona. Yet all of these people appear in statistics under the broad heading of the semiconductor workforce.

The earlier Semiconductor Industry Association and Oxford Economics study projected that the U.S. semiconductor workforce would grow from about 345,000 workers to 460,000 by 2030, with 67,000 positions potentially going unfilled. The newer National Network for Microelectronics Education analysis suggests the gap may be substantially larger. The problem is therefore not simply:

We need more semiconductor people.

It is:

We need particular semiconductor capabilities, in particular locations, at particular moments—and we need the knowledge around them to move.

That is a very different problem.

Geography Is Part of the Skills Gap

There is another uncomfortable reality hidden inside national employment statistics:

Certain types of jobs do not move. People have to.

America's semiconductor expansion is creating concentrated demand in Arizona, Texas, Ohio, New York, Idaho and Indiana. A highly experienced engineer losing a job elsewhere does not automatically solve that shortage. They may have a spouse with a career, children in school, a house with a 3% mortgage, deep community ties, or simply no desire to relocate across the country for fab work. From a spreadsheet perspective, America has an available semiconductor engineer. From a hiring manager's perspective in Arizona, it still has a vacancy.

Some of the Shortage Is Economic

The word “shortage” can also conceal a price problem. The reporting behind the latest workforce analysis places typical U.S. semiconductor salaries at roughly $127,000 to $187,000, with senior roles exceeding $238,000. Meanwhile, semiconductor workers in parts of Asia have won exceptional bonus packages as employers compete to retain scarce expertise. That comparison is not exact; salary structures, taxes, seniority, and bonus systems vary significantly by country. But it exposes a basic labor-market truth:

A company cannot declare talent scarce while assuming the historical price and working conditions of that talent will remain unchanged.

Semiconductor employers compete not only with one another, but with AI companies, software businesses, adjacent engineering industries and entrepreneurship. Fab operators must also persuade a new generation to accept continuous-manufacturing schedules that can include demanding 12-hour shifts. Some people who could work in semiconductors have better alternatives. That is not the absence of talent. It is talent making a choice.

America Is Rebuilding Knowledge, Not Just Fabs

The most revealing response to the shortage is not a job advertisement. It is the movement of people across the Pacific. Samsung and SK Hynix are temporarily bringing experienced South Korean workers to the United States to help start new facilities. Samsung has sent U.S. employees to South Korea for months of hands-on training. TSMC used a similar approach during the development of its first Arizona fabs. The reason is straightforward: much of the world's leading-edge manufacturing expertise has accumulated in Asia over decades. That expertise does not arrive with a piece of equipment, a process manual, or a university degree.

It lives in practiced judgment:

• knowing which variation is harmless and which is an early warning;
• recognizing when nominally correct equipment is behaving incorrectly;
• understanding how one process adjustment will ripple into yield, reliability or downstream packaging;
• and knowing whom to call when the documented procedure no longer fits reality.

America is not simply trying to hire a semiconductor workforce. It is trying to reconstruct an advanced-manufacturing knowledge base that migrated to Asia over several decades—and do it in a handful of years. That is why the new educational pipeline, while essential, cannot be the whole answer.

Purdue reports about 2,500 students per semester in chip-related courses and provides hands-on equipment training. Arizona State University converted a former Motorola fab into a cleanroom training facility and hosts a TSMC technician program. Intel has launched apprenticeships and education programs, while more than 80 community colleges have introduced or expanded semiconductor programs since the CHIPS Act. This is meaningful progress. But the timing problem remains:

The United States is building the semiconductor education pipeline while also building the fabs that will consume its output. Fabs can be built faster than experience.

Why Layoffs and Shortages Coexist

Layoffs do not necessarily mean there is no work to be done. They can mean a company can no longer support its existing cost structure. Products are canceled. Markets move. Acquisitions create overlap. Management layers accumulate. Investment priorities change. A firm can have organizational surplus in one area and acute technical scarcity in another. That is why a company can reduce thousands of positions while still competing fiercely for process technologists, equipment engineers, advanced-packaging specialists, verification experts or AI infrastructure talent. But this explanation is incomplete, because layoffs do more than reduce headcount. They can also damage the invisible system through which a company knows what it knows.

The Missing Layer: Connectors

In The Connector's Playbook, we argue that modern organizations rarely fail because they lack talent, data, or technology. They struggle because the most important work happens in the spaces between them:

• between teams that no longer share context;
• between systems that were never designed to communicate;
• between technical specialties with different languages and incentives;
• between experienced people and the generation expected to replace them;
• and between signals arriving faster than the organization can interpret them.

We call these spaces seams. Semiconductor companies are full of specialists, as they must be. But specialists alone do not create a functioning system. Somebody has to connect process technology to design rules, architecture to verification, requirements to implementation, equipment behavior to yield, customer problems to product decisions, and historical failure patterns to today's choices. Those people are Connectors.

They may be senior engineers, applications engineers, program managers, architects, product leaders, technicians, or customer-facing veterans. Their defining capability is not a job title. It is their ability to cross boundaries, translate, preserve context, recognize patterns, and help other specialists act coherently. Their value often remains invisible because their output appears as an avoided mistake, a better question, a faster handoff, a new engineer brought up to speed or a decision corrected before the cost becomes visible. That invisibility makes Connectors particularly vulnerable during restructuring. A spreadsheet can count their salary. It may not count the technical dead ends they prevent, the relationships they maintain or the institutional memory that leaves with them.

The result is a second paradox:

A company can improve its near-term headcount efficiency while making its remaining expertise harder to use.

This is the semiconductor version of the paradox of plenty. The industry can possess more tools, more data, more specialized teams and more investment than ever - and still be constrained because the connections among them are weak.

The Real Scarcity Is Transferable Experience

Semiconductors have an unusually long expertise curve. A university can teach semiconductor physics, process fundamentals and design methodology. It cannot provide 20 years of tapeouts, process excursions, yield learning, customer escalations or post-silicon failures. Much of the industry's most valuable expertise is tacit. It resides in the pattern recognition of people who have already seen several versions of a problem.

This is why the workforce challenge cannot be solved by counting graduates alone. The critical question is whether experience can be transferred, connected and multiplied before it walks out the door. Traditional organizations often treat mentoring as an informal courtesy performed around the edges of “real work.” In the current environment, it is production infrastructure. An experienced engineer who teaches ten younger engineers, codifies failure patterns, documents decision logic and connects teams may create more long-term capacity than one additional individual contributor whose knowledge remains isolated. The semiconductor industry therefore needs not only more specialists, but more people and systems capable of turning individual expertise into organizational capability.

AI Changes the Equation - But Not in the Simplistic Way

AI is increasing the demand for chips while beginning to transform how chips are designed, verified, manufactured and supported. The first wave is already visible in code generation, specification analysis, verification assistance, documentation, debugging, test generation, design-space exploration and knowledge retrieval. This does not mean experienced semiconductor engineers become unnecessary. It means their judgment can operate with greater leverage.

AI can help an experienced practitioner:

• search decades of technical history;
• find contradictions across specifications and change requests;
• surface weak signals across projects and fabs;
• turn recurring decisions into reusable playbooks;
• compare scenarios and dependencies;
• and make hard-won knowledge accessible to the next person facing the problem.

But AI does not know, by itself, which anomaly matters, which result is plausible, when a model is confidently wrong or which tradeoff will create a failure two years later. Where AI generates possibilities, experienced humans must still apply context. Where systems create volume, Connectors create meaning. The right objective is not to automate experts out of the system. It is to make scarce expertise findable, teachable and reusable.

What the Industry Should Do Now

If the shortage is really a knowledge-allocation problem, the response must go beyond recruiting.

1. Map capabilities, not job titles

Companies should know which technical capabilities, relationships and forms of tacit knowledge they possess - not merely how many people sit in each organization. Before a restructuring removes a role, leaders should understand what knowledge routes through that person.

2. Treat knowledge transfer as funded work

Mentoring, apprenticeship, rotational assignments, decision logs and technical playbooks need schedules, owners and measurable outcomes. They cannot remain side projects performed by already overloaded experts.

3. Build Connector paths alongside specialist paths

The industry needs deep specialists. It also needs people rewarded for crossing boundaries, translating between disciplines and raising the coherence of the entire system. Breadth should not be mistaken for lack of depth.

4. Connect displaced experience to new capacity

Regional talent exchanges, remote advisory roles, short-term fab assignments and structured expert networks could connect veterans who cannot relocate with teams that urgently need their judgment. Not every valuable contribution requires permanent residence beside the fab.

5. Design AI around expertise multiplication

The best use of AI is not indiscriminate headcount reduction. It is capturing context, revealing dependencies, preserving technical history and helping one experienced person guide many others without becoming a human bottleneck.

6. Improve the economic proposition

If semiconductor work is nationally strategic and technically scarce, compensation, schedules, career mobility and workplace design must reflect that reality. The industry cannot market yesterday's employment proposition to tomorrow's workforce.

The Question We Should Be Asking

The question is not:

How can there be semiconductor layoffs when America is short 157,000 workers?

The better question is:

What semiconductor capabilities does America need for the next 20 years, and how do we connect the workforce we have to the industry we are building without discarding decades of accumulated knowledge?

That means retraining before replacing. It means creating technician careers that do not require four-year engineering degrees. It means making semiconductor work compelling to a new generation. It means building bridges between displaced experts and new manufacturing regions. It means preserving the logic behind decisions, not merely the final documents. And it means using AI to amplify human judgment rather than treating it as another way to remove people.

I personally believe that America does not need 157,000 more semiconductor workers.

It needs the right knowledge in the right place at the right time - and the human infrastructure capable of moving that knowledge to where it matters.

The paradox of layoffs and a 157,000-person shortage exists because the people making business decisions and the people analyzing our future needs are both wrong in different, fundamental ways. And if they keep driving big business decisions, we will struggle to build a sustainable, profitable industry with fulfilling, satisfying careers for our semiconductor workforce.

The scarce resource is not simply people.

It is connected experience.

Quick answer: Why are semiconductor companies laying people off during a workforce shortage?

Because “semiconductor worker” is not one interchangeable occupation. Companies can eliminate roles in declining products, duplicated organizations, or management layers while facing acute shortages in process engineering, equipment maintenance, advanced packaging, verification, and fab operations. Geography, compensation, and working conditions further restrict the available pool. Layoffs can also remove experienced Connectors who transfer knowledge between specialists, making the remaining workforce less effective. The result is not a simple labor shortage but a mismatch of skills, location, experience, and connection.

Suggested sources

• National Network for Microelectronics Education: National Landscape Analysis
• McKinsey: Reimagining labor to close the expanding U.S. semiconductor talent gap
• SIA/Oxford Economics: Chipping Away at the U.S. semiconductor labor gap
• CNBC: U.S. chipmakers face deep labor shortage
• Intel: First U.S. apprenticeship program for manufacturing technicians