The state of the American computing profession

A profession redefined by consequence

For most of the last three decades, the defining question about the American computing workforce was one of supply: were enough people entering the field, and were they entering it fast enough? That question has not disappeared, but it is no longer the one that keeps chief technology officers, regulators and university deans awake. The pressing question in 2026 is one of consequence. Computing professionals now design and operate systems that allocate credit, triage patients, route autonomous vehicles through city intersections, schedule electricity across an independent grid, and decide which of thousands of manufacturing wafers is scrapped. The cost of an unexamined assumption has moved from inconvenience to public harm.

Our study of 48,000 members, 190 accredited degree programs and 640 employer respondents describes a profession absorbing that shift unevenly. Where responsibility has grown quickly, formal competence assurance has lagged. Roughly three in five respondents told us they had signed off on a system whose downstream effects they could not fully characterize. Two in five said no one in their organization held documented accountability for that sign-off.

“Employers can hire people who build systems. They struggle to hire people qualified to judge whether those systems are fit to operate.”

What the numbers show

Compensation continues to reward specialization far more than seniority. Practitioners holding a recognized professional credential in security, assurance or safety-critical engineering reported a median premium of 14 percent over peers with equivalent experience and no credential. That premium roughly doubles in regulated sectors — health, finance, energy and transport — where an employer must demonstrate to an external body that a competent person reviewed the work.

Mobility tells a different story. Practitioners move between employers readily, but move between domains rarely. Only 11 percent of respondents had changed sector in the previous five years, and those who did most often moved into, rather than out of, artificial intelligence work. The result is a labor market that looks liquid in aggregate and is stubbornly siloed in practice.

• Median credentialed premium: 14 percent, rising to 27 percent in regulated sectors.
• 58 percent of respondents have approved a system whose effects they could not fully characterize.
• Only 11 percent changed domain in five years; AI is the dominant net importer of talent.
• Employers report the longest hiring times for assurance, verification and safety-case roles.

The Texas Triangle as a natural experiment

The region around the Society's Austin headquarters offers an unusually complete picture of the profession, because almost every computing discipline is represented within an hour's drive. Advanced semiconductor fabrication, hyperscale data center construction, autonomous vehicle operations, launch vehicle avionics and grid-scale energy management all recruit from the same regional pool. That density produces a labor market in which a verification engineer can plausibly move from wafer yield analytics to flight software, and occasionally does.

It also produces the sharpest version of the competence problem we found anywhere in the country. Employers competing across those sectors described identical difficulty: they can hire people who build systems, and they struggle to hire people qualified to judge whether those systems are fit to operate. The scarcity is not in construction. It is in assurance.

What the Society recommends

Our recommendations are deliberately narrow, because the profession does not need another broad call for more training. It needs a small number of load-bearing changes. First, employers should name an accountable, credentialed professional for every system whose failure would reach the public, and record that accountability where an auditor can find it. Second, accredited programs should treat evaluation and assurance as first-class subjects rather than final-semester electives. Third, professional registration should recognize demonstrated judgment under uncertainty, not only years served.

None of this requires new legislation. It requires the profession to hold itself to the standard the public already assumes we meet.