Every CTO who signed off on a reduction-in-force this year was told the same thing by the same market: there are too many people. Yet when ManpowerGroup asked 39,063 employers across 41 countries what they can't hire, the answer that came back first, for the first time ever, was AI skills. AI model and application development (20 percent) and AI literacy (19 percent) now top the global hardest-to-hire list, edging out engineering at 19 percent. Traditional IT and data skills, the category that used to anchor this survey, dropped to seventh at 17 percent.
Here's the number sitting right next to it that makes this genuinely strange instead of just interesting: depending on which tracker you trust, between 128,000 and 209,000 U.S. tech workers have been laid off in 2026 — Layoffs.fyi counted 128,536 across 299 companies as of September 10, while broader trackers put it north of 200,000 across 365 events. Hold the methodology constant and the direction is the same either way: Layoffs.fyi's 2026 count has already passed its own full-year 2025 total of roughly 124,000, with a quarter still to go.
The company-level numbers are blunter. Amazon alone has cut roughly 30,000 corporate roles across two major rounds since October 2025, plus further undisclosed cuts through mid-2026. Oracle cut 30,000 in a single event. It's the worst year for tech headcount since the 2023 peak.
Read those two facts side by side and the obvious question is which one is wrong. Neither is. That's the part most hiring plans get backwards.
A talent shortage and a layoff wave aren't contradictory signals from the same market — they're the same signal, described from two different vantage points. ManpowerGroup isn't measuring "not enough tech workers." It's measuring "not enough of a specific two-part skill": the ability to actually build AI systems, and the literacy to apply them with judgment inside a real workflow. The people who lost their jobs this year overwhelmingly did not have that skill, or worked in roles where that skill wasn't the job requirement. Companies aren't short on people. They're short on the very narrow slice of people whose work an AI system can't yet draft a passable version of, and whose judgment the company still needs to catch what the system gets wrong. Everyone else is in a labor market with a surplus, competing for a shrinking set of roles. That narrow slice is in a market where one in five employers worldwide says it can't find them.
ManpowerGroup's own chief executive says the quiet part, if you read it closely:
"Companies are responding with upskilling and more flexible workforce models, recognizing they will need to hire for potential; while building AI literacy across their workforce so people can apply new technology with judgment and confidence. AI is not replacing jobs, it is reshaping work, and companies that connect productivity gains with opportunity and career growth will be best positioned to compete in a talent-scarce world."
— Jonas Prising, Chair & CEO, ManpowerGroup, news release
Reshaping, not replacing. That's the whole argument in three words, and it's coming from the firm that just counted the shortage.
This is exactly why "we're doing layoffs" and "we can't find anyone" show up in the same earnings call without anyone in the room noticing the tension. They're not competing narratives. They're the same restructuring, and the org chart just hasn't been redrawn to say so out loud. One WARN filing out of Amazon's SEA106 building in Bellevue lists 49 names, and the titles run from entry-level software development engineers straight up to a VP of legal. One sheet of paper, both ends of the org chart. That isn't proof AI is shrinking headcount uniformly — it's proof the company already knows which slice of its own workforce is in surplus and which slice it would fight to keep. It just hasn't said which slice is which out loud, because that's a harder conversation than "de-bureaucratizing."
The hiring plans that cross my desk keep making the same mistake. They treat this as a headcount problem instead of a skill-mix problem. A VP who approves "reduce corporate headcount by X percent" and a recruiter who's told "we have an AI skills gap, go fill it" are often working the same restructuring without realizing it. The company doesn't need fewer people and more AI people. It needs to know, role by role, which of its current people are doing work an agent can already draft, and which are doing the specific judgment work — deciding when the AI's output is wrong, what to build with it, when to trust it — that one in five employers worldwide now says it can't source. You can't solve that with a blanket RIF and a separate blanket AI hiring initiative running in parallel. That's two headcount decisions fighting each other on the same P&L line.
If you're planning next year's budget, run the audit before you run either motion. Don't ask "how many roles can we cut" and "how many AI engineers can we hire" as separate exercises — ask, for every role on the chart, whether the person in it is doing work that's already commoditized or work that's now the hardest thing in the world to source. The layoff number and the shortage number aren't in tension. They're the same answer, sorted two different ways. Sort your own org chart before someone else's press release does it for you.
VC5 Consulting sources the specific slice of talent ManpowerGroup's survey says is now hardest to find — the AI build and judgment skills that don't show up in a headcount-reduction spreadsheet. If your 2026 plan has a layoff line and an AI-hiring line that were never reconciled against each other, let's talk.