Tech Apprenticeships Data: 58,000 Served, Too Few to Replace Junior Hiring

The Department of Labor says more than 58,000 registered apprentices were served across technology, cybersecurity and AI-related occupations in 2025. That headline includes the military. In the one published split we found, 83% of cyber-related apprentices were in a military programme. On that base, tech apprenticeships data points to roughly 2,100 civilian completers a year, which is not a replacement junior developer pipeline.

Can registered apprenticeships replace junior developer hiring? Not at current scale. DOL data shows about 10,293 tech-adjacent apprentices outside the military in the one split published. At national rates that implies about 4,551 starts and 2,116 completers a year, against a modelled shortfall of 20,616 junior seats.

What the tech apprenticeships data actually counts

A registered apprenticeship is a paid job with structured training, written into a programme registered with the Department of Labor or a state agency. The system is tracked mainly in RAPIDS, DOL's apprenticeship database. It was built for trades, and the public numbers still show it.

That matters because tech apprenticeships are now being pitched as the answer to shrinking junior hiring. DOL launched its Tech Registered Apprenticeship Innovation Network in April 2026, and the headline figure gets quoted as proof of scale. Before you plan around it, read what the number measures.

Served is not the same as active or new

DOL's tech figures use the word served. In its Trendlines reporting, DOL says total served includes active apprentices plus those who exited during the fiscal year. A served count therefore runs higher than an active count, and it says nothing direct about how many people started.

That is why I would not compare 58,000 with any annual hiring figure. A person in year two is counted, and so is someone who quit in month three. It is a throughput-flattering measure of a system's reach, and it should be read as one.

Three scopes, none of them labelled the same way

DOL has published at least three tech-flavoured measures in recent years. One is an industry measure, built on NAICS Code 51, the Information sector. One is a cyber-related occupation count. The newest is an occupation count across technology, and the 2025 wording adds cybersecurity and AI-related work.

These do not form a clean series. The 2024 technology count was over 64,800, and the 2025 count, with wider wording, was more than 58,000. That looks like a 10.5% decline, but DOL does not explain the change, and a shifted definition is at least as likely. I would not report it as a fall.

The official ladder has not moved

The BLS Occupational Outlook Handbook, modified in August 2026, still lists a bachelor's degree as typical entry education for software developers, with no on-the-job training. It does not mention apprenticeship anywhere on that page. Apprenticeship.gov does list Software Developer among its high-demand occupations, so the route exists. It has not become the expected one.

The DOL trend table for tech apprenticeships

Here is every series we could open, in one place. The left columns cover all occupations. The right column is the closest tech measure available for that year, and it changes scope, so read down it with care.

Sources: DOL national results pages for FY2019 to FY2021; DOL fact sheets for 2022 to 2024; Work Shift for FY2024; Nevada Apprenticeship Council report for FY2025, quoting national totals at 30 September 2025. Dashboard only means DOL publishes the value in an interactive dashboard we could not read as text. Retrieved 6 November 2026.

The whole system grew. Active apprentices rose 11% between FY2019 and FY2025, and new apprentices rose 23%. DOL's goal is 1 million active apprentices, and Work Shift reported 714,000 in August 2026, on a trend about 150,000 short even by 2030.

Tech growth looks more dramatic only when the base is tiny. DOL's 221% five-year rise in the IT industry count implies a starting point near 1,963 and an end point of 6,300. The broader technology count rose 29% over five years, from an implied base near 50,200. That is respectable, and it is not a breakout.

The FY2021 row is the warning. The federal table a year earlier printed 2,951 active apprentices in Information, and the FY2021 RAPIDS table printed 189, beside 59,411 with no industry recorded. Those tables have different scopes, so this is not a like-for-like drop. It does show how little weight an industry code can carry here.

The military share inside the tech apprenticeship headline

This is the finding I think matters most, and it is sitting in a footnote. DOL's 2025 technology fact sheet says its total includes United Services Military Apprenticeship Program apprentices. USMAP serves active-duty Navy, Coast Guard and Marine Corps members, and DOL describes it as the largest registered programme in the country.

The only published split we found is on the cyber side. For 2023, DOL reported 60,676 cyber-related apprentices, with 50,383 in USMAP and 10,293 outside it. That is 83% military, and the cyber count is nearly as large as the whole technology count a year later.

Across all occupations, USMAP was a big minority rather than a majority. In FY2019 DOL showed 109,878 USMAP actives against 633,476 nationally, or 17%. The cyber share is close to 5 times that. So a system-wide reading of apprenticeship will understate how military this particular slice is.

Military training is not worthless to a software employer. Veterans enter civilian technical work, and that is a real pathway. It is simply a different pipeline with a different clock, because those apprentices are serving personnel and their intake is set by force needs rather than by your hiring plan.

The civilian count is a proxy, and we say so

DOL reports USMAP in aggregate only, and its FY2019 page says that data cannot generally be broken out in greater detail. That is why no all-tech civilian figure exists in public. We use 10,293 as the best available proxy, with the 6,300 Information sector count as a lower reference and the full 58,000 as a ceiling.

Completion cuts the tech apprenticeship pipeline roughly in half

A pipeline is measured by what comes out of it. Work Shift reported a national completion rate of 46.5% in August 2026. Social Finance found 46.8% for the 2017 starting cohort within 6 years, with the military and federal prison programmes excluded, so the two figures agree from different directions.

DOL has its own beta completion dashboard covering FY2013 to FY2024, and it gives national averages by year. We could not find a national rate for technology occupations published anywhere. For a category being sold as a workforce answer, the absence of that one number is telling.

Tech programmes can be shorter than a multi-year trades apprenticeship, and a shorter programme has fewer months in which someone can quit. Social Finance also notes that cancellations spike in the early months. So I would expect a tech rate to differ from the national one, and possibly to beat it.

We use 46.5% anyway because it is the measured figure. To keep that honest, the scale table below also shows the starts, which is the case where every apprentice finishes.

The junior developer pipeline drop, expressed in seats

The other side of the comparison has its own problem. Nobody publishes a national count of junior developers hired per year, so every headline figure is a percentage of something unstated. The strongest figures are covered in our earlier piece on the junior developer pipeline, and they disagree in useful ways, as the comparison of conflicting entry-level hiring data sets out.

Stanford's payroll study reports that workers aged 22 to 25 in the most AI-exposed occupations fell about 11% in levels and sit about 19% below their counterfactual. SignalFire reports new graduate hiring down about 65% against 2019, at 12 named large technology companies. Indeed reports that entry-level roles were 4.5% of US software development postings in the first quarter of 2026.

None of these is a national software junior count. Stanford covers all exposed occupations, and SignalFire covers 12 firms. We use all three as scenarios rather than choosing one, because the conclusion should survive whichever you trust.

The baseline we use for a junior class

BLS names a bachelor's degree as typical entry education, so we use degrees as the class size. US institutions conferred 108,503 computer and information sciences bachelor's degrees in 2021-22. As a cross-check, BLS projects about 106,100 openings a year for software developers and testers, at all levels, over 2025 to 2035. These two unrelated figures land close together, which is the only comfort this baseline offers.

Software developer headcount itself has not collapsed, as our breakdown of BLS occupation data shows. The squeeze is at the door, and that is the part an apprenticeship would need to replace.

The scale gap between junior seats and tech apprentices

This is the calculation we could not find published elsewhere, with every step shown. Change any assumption you dislike. The order of magnitude is what holds.

Supply. Take three starting counts: 6,300, 10,293 and 58,000. Estimate annual starts using the national ratio of new to active in FY2025, which is 311,035 divided by 703,475, or 44.2%. Then estimate completers by multiplying starts by 46.5%.

Demand. Apply each scenario to the class of 108,503: an 11% fall gives 11,935 missing seats, a 19% shortfall gives 20,616, and a 65% fall gives 70,527. The last is a stress test, because SignalFire's figure describes 12 firms and was never a national rate.

Zan Digital estimates. Illustrative scenarios, not measured apprentice flows. Starts apply an active-count ratio to served counts, which overstates starts somewhat. The right-hand column is the only one that clears any bar, and it counts serving military personnel as civilian supply.

On the central case, civilian tech apprenticeships produce about 2,116 completers against 20,616 missing seats. That leaves 18,500 seats unfilled, or about 10 for every completer. If every starter finished, coverage would still reach only 22%.

My read is that the pipeline is real and it is small. It is a useful programme for the people in it, and I would not call it a replacement for anything at this scale. To close the middle-row gap on starts alone, the civilian system would need to grow about 4.5 times, assuming the gap itself stopped growing.

Where this argument is weakest

This post has more proxies than I would like, and you should know exactly where they sit.

The civilian figure may be low

The 10,293 is cyber-related and from 2023, so it is not the full civilian tech count. A sensitivity check helps. The 2024 technology total exceeded the 2023 cyber total by 4,124. Treating all of that as civilian gives 14,417, which yields about 2,964 completers and 14% coverage of the 19% scenario. The picture moves, but it does not change.

Unregistered programmes are invisible

Employer-run apprenticeships that never register do not appear in any of this. An Urban Institute and Mathematica evaluation of two DOL grant programmes reported IT as the most common industry among their nontraditional apprenticeships, and that activity shows real interest. The registered system may undercount the actual activity by an amount nobody can state. This is the strongest objection, and I have no number for it.

The demand side is a scenario, not a measurement

Degrees are not hires, 2021-22 is old, and Stanford's exposed-occupation shortfall is not a software figure. The authors say their divergence persists when technology firms and computer occupations are excluded, and they call the patterns descriptive rather than causal. Applying that percentage to computing graduates is our choice, not their finding.

The gap may not need filling

This is the objection I take most seriously. If employers cut junior seats because they want fewer juniors, an apprenticeship route does not bring those seats back. An apprentice is an employee first. A programme needs an employer willing to hire, so apprenticeship can change who gets a seat, and how they are trained, more easily than how many seats exist. On that view, the scale gap understates the problem.

Frequently asked questions

How to use this in a headcount plan

If you run an engineering or delivery team, do not write apprenticeships into your plan as the replacement for a junior intake you have cut. Write down the seats first. An apprenticeship is a way to fill a junior seat with a different training model, so it only helps once you have decided the seat exists.

Then ask any programme provider for 4 numbers, by cohort: starts, completions, conversion to permanent roles, and the date each was measured. Plan on the 46.5% national rate until a provider shows you better. Providers who answer with a retention percentage and no starting count have not given you a pipeline figure.

The first-rung work those apprentices would do has changed too, which is why our reading of public engineering career ladders is worth having open when you write the role.

Related reading