Photo: Laura Ockel / Unsplash
Every conversation about AI eventually becomes a conversation about chips, and every conversation about chips eventually becomes a conversation about people. Companies have committed more than $800 billion to new semiconductor plants in the United States since the 2022 CHIPS Act, but the buildings do not run themselves. As The Associated Press reported this month, community colleges are racing to fill the gap with short, hands-on programs that move people from zero experience to the fab floor in a matter of weeks. It is one of the most hopeful workforce stories of the year, and it carries a lesson for every founder struggling to hire.
What is happening
AP told the story through Ethan Mock, a 23-year-old in his fifth and final year of a business administration degree at Northern Arizona University. He had worked part time at a gym for three years and taken on more than $100,000 in debt. In July he heard on the radio about a tuition-free, three-month semiconductor crash course at Rio Salado College in Tempe, Arizona, which came with interviews at major Arizona chipmakers. He signed up with no background in the field. Two months later he was operating the advanced machinery that turns thin slices of silicon into chips for phones, robots and data centers. In his words, “I do want to be in a booming industry.”
His classmates are not who you might picture: a 64-year-old middle school teacher, a recently laid-off software developer in his 50s, gig workers without degrees, and younger students looking for an alternative to a traditional four-year path.
Rio Salado consulted with TSMC and Intel and condensed courses that once took up to two years into a free 12-week program open to the public this fall. About a dozen students are enrolled. They learn fabrication procedures first in virtual reality headsets, then put in about 20 hours of hands-on work in Arizona State University clean rooms, suited up head to toe. Similar programs that lead to industry interviews run at Austin Community College in Texas, West Valley College in Saratoga, California, and many other schools. The industry group SEMI says more than 200 community colleges and universities have coordinated with it to prepare workers for new plants.
The scale of the shortage
The urgency is real. Shari Liss, who leads talent initiatives at SEMI and manages a CHIPS Act-funded workforce program, told AP the U.S. industry could be short as many as 157,000 full-time workers by 2030. Experts expect the worst gaps in Texas, Ohio, California, New York and Arizona, and they warn that shortages could delay production and undercut the country’s push to lead in AI.
The money behind the training is a mix of public and private. The National Science Foundation and companies including TSMC, Samsung and SK Hynix are funding community college partnerships, helping schools buy equipment and advising them on which skills matter most. TSMC’s $265 billion project in northern Phoenix is the largest foreign direct investment in U.S. history. Pearl Chang Esau, an education and workforce consultant for TSMC Arizona, estimated it will create about 25,000 jobs over roughly the next decade, half of them for locally hired technicians.
In the meantime, the chipmakers are also bringing in hundreds of experienced engineers and technicians from Taiwan and South Korea to run new facilities in Arizona, Texas and Indiana, and TSMC’s workforce consultant says their job is to train local workers.
Why founders should pay attention
You may never build a fab. But this story upends three assumptions that quietly shape how most startups hire.
- The degree is not the job. About 15 years ago, according to Rick Vaughn, who manages technology and engineering courses at Rio Salado, Intel required a two-year associate degree for entry-level fab jobs. Now some of the world’s most complex manufacturing is recruiting people after 12 weeks of focused training. If chipmakers can drop the credential, so can you.
- Career changers are an underpriced talent pool. A laid-off developer and a veteran teacher bring judgment, reliability and work habits that are hard to teach. Many startups screen them out by default.
- Training is a competitive weapon. The companies that help design the curriculum get first look at the graduates. That is a hiring advantage money alone cannot buy later.
For hardware, robotics, energy and deep-tech founders, there is a more direct implication: you will be competing with giants for the same technicians. The time to build your pipeline is before you need it.
A founder’s playbook for building a talent pipeline
- Write job descriptions around skills, not pedigree. List the tasks someone must do in their first 90 days and the skills required. Cut degree requirements that the work does not actually demand.
- Call your local community college. Ask the workforce development office what programs exist, what equipment they lack and whether employers can advise on curriculum. Even a small company can offer guest lectures, site visits or a project for students.
- Offer paid apprenticeships. Short, structured, paid on-ramps let you test fit before committing to a full hire and give candidates a reason to choose you over a bigger name.
- Recruit career changers on purpose. Post roles where they look: workforce boards, veteran programs, adult education centers.
- Build training into onboarding. Document your processes so a motivated newcomer can learn them. Fabs run on hundreds of tightly controlled steps; your company runs on its own version of that.
- Show the ladder. Mock is already thinking about starting as a technician and later combining that experience with his business degree. Make your career paths visible so people can picture a future with you.
- Treat precision and safety as culture. Vaughn tells his students that precision, safety and reliability are the foundation of fab work. That is good advice for any team shipping things that matter.
Questions to ask before your next hire
- Which of our current requirements actually predict success in this role?
- Could someone with strong fundamentals learn this job in three months with the right training?
- What local institutions are already training people with adjacent skills?
- Are we competing for the same small pool everyone else is fighting over, or growing our own?
What to watch next
- Whether short programs can scale. AP questioned whether accelerated courses can grow fast enough to meet hiring demand from TSMC, Samsung and SK Hynix.
- Federal workforce funding. Programs backed by the CHIPS Act and the National Science Foundation are central to this effort. Changes in Washington could speed it up or slow it down.
- The regional race. Texas, Ohio, California, New York and Arizona face the most acute shortages. Watch which states move fastest to expand training.
- Knowledge transfer. How well the engineers brought in from Taiwan and South Korea pass their expertise to local hires will decide whether these plants become truly American operations.
The bottom line
The AI economy is often described in terms of models and megawatts, but it is built by people in protective suits repeating careful steps on microscopic layers of silicon. The good news is that many of those people do not need a decade of schooling to get there. They need a door that opens and a company willing to train them. With Thanksgiving approaching, I am grateful for the instructors and community colleges doing this unglamorous, life-changing work. Founders, the best thing you can do is pick up the phone and join them.
Source: The Associated Press
Quick check
How much did you catch? Tap a question to see the answer.
1. How large could the U.S. chip industry’s worker shortage be by 2030, according to SEMI’s Shari Liss?
Answer: As many as 157,000 full-time workers.
2. How long is Rio Salado College’s free semiconductor program?
Answer: 12 weeks, condensed from courses that once took up to two years.
3. About how many jobs is TSMC’s Phoenix project expected to create, per its workforce consultant?
Answer: About 25,000 over roughly the next decade, half for locally hired technicians.
