AI hiring 2026: heavy spenders add staff faster than peers

Companies making the heaviest investments in artificial intelligence are still hiring faster than their peers, according to Ramp data cited by the Financial Times, a result that complicates one of the market’s cleanest assumptions about the AI trade: that higher model bills and data-centre spending will quickly show up as lower headcount and wider margins.

Across 22,000 US companies tracked by Ramp and Revelio from January 2021 to February 2026, the high-intensity adopters, defined in the reporting as businesses spending about $33.67 per employee each month on AI tools, expanded headcount by 10.2 per cent over two years after adoption. Entry-level hiring rose 12 per cent over the same stretch, Business Insider and TechCrunch reported from the study.

Goldman Sachs reads the same labour market from a colder angle. MarketWatch’s account of Goldman analysts Sarah Dong and Joseph Briggs said 20.6 per cent of US companies were using AI in June, but the labour effect remained visible only in pockets, with weakness in some white-collar functions offset by construction hiring tied to data-centre buildout.

That tension matters for investors. Equity markets have often priced generative AI as a margin story first, a belief that software and models would trim labour before the bills for chips, cloud capacity and debt financing came due. The hiring data suggests a messier sequence. Output is being expanded, workflows are being rearranged and the first hires to benefit may be junior workers who can operate with the tools, not senior workers being replaced by them.

What the hiring data actually shows

Public layoffs made the subtraction story easy to believe. TechCrunch’s running list of tech layoffs where employers cited AI kept growing through June, while CNBC reported that US tech layoffs had reached about 140,000 this year and AI had been the main reason companies gave for cuts for a fourth straight month. That is the backdrop against which the Ramp figures landed, and it helps explain why the study read as a contradiction rather than a refinement.

Ramp’s own warning was blunt.

“If you are reading headlines where CEOs blame layoffs on AI, be skeptical.”

— Ramp study authors, Business Insider

What makes that line important is not that layoffs are imaginary. They are not. Instead, it argues that attribution is doing too much work. Post-pandemic over-hiring, slow demand in parts of tech, cost-cutting pressure and normal restructuring can all sit inside the same announcement. Label the cut an AI efficiency move and the narrative becomes cleaner than the underlying business case.

Junior workers sit at the centre of the dispute. The market’s layoff thesis assumes entry-level roles are the first to go once chatbots and coding assistants start absorbing routine tasks. Yet the same dataset showed the heaviest adopters increasing entry-level hiring. Separate Business Insider reporting on millions of job listings pointed in a similar direction: employers are raising the premium on workers who can use AI, not abandoning early-career hiring altogether. The Atlantic’s analysis of Erik Brynjolfsson’s productivity case carried the other half of the story, warning that a hiring recession can still exist for young workers even when the technology’s long-run promise looks real.

A partial answer for graduates follows from that split. Choose the firms using the tools, but do not confuse that with an easy market.

“If you are a young person entering the labor market, and you are choosing between otherwise similar firms, choose the one that is using AI.”

— Ramp study authors, Business Insider

Firms that deploy AI aggressively appear to want different junior workers, not necessarily fewer of them. Skill requirements rise, screening gets harder and training shifts toward prompt-heavy, software-assisted workflows. None of that feels benign from the applicant side. Still, it is a different claim from broad labour destruction, and the distinction matters for any investor assuming that today’s AI winners will soon enjoy a structurally lower wage bill.

Why the market’s layoff thesis looks early

Sceptics of the hiring story have real ground to stand on. High AI spenders are not a random sample of corporate America. They are more likely to be well-capitalised, venture-backed, tech-forward or already growing faster than peers. In that sense, the study may reveal as much about who can afford AI as about what AI does to labour demand. Goldman’s narrower read of the market, one in which adoption has reached only about one company in five and the employment effects remain uneven, is a reminder that the macro picture still looks thin.

Another problem is that some companies appear to have mistaken software adoption for an automatic staffing blueprint. CNBC’s reporting on employers who later reversed AI-linked cuts said 32 per cent of hiring managers who shed workers after AI deployments ended up rehiring, a sign that the labour-saving math often looks cleaner in a board deck than in an operating model.

“This can lead to duplicated effort, slower decision-making, and diminished productivity gains.”

— Jessica Zhang, CNBC

Zhang’s point lands because generative AI has so far behaved more like a coordination problem than a one-step replacement machine. Managers must decide which workflows can be standardised, which still need review, where compliance risk sits and how quickly employees can absorb new tools. Add poor training or badly chosen use cases and companies can end up paying for software, severance and re-hiring in the same cycle.

Andy Challenger’s sceptical frame fits here as well. AI can be the headline explanation for a cut without being the load-bearing economic reason for it. When investors hear “AI layoffs”, they may picture a permanent step-down in labour intensity. What the current evidence shows instead is a muddle of selective hiring, selective cuts and a broad corporate search for the right mix of people and tools.

Capex first, labour savings later

One reason the market keeps reaching for a layoff payoff is that the spending is already visible. Semafor reported that Amazon was preparing to borrow $25 billion or more as AI buildout costs piled up. In a separate piece, Semafor argued that big tech’s surge in AI spending was collapsing free cash flow and reviving an old question about whether growth now means buying more tokens and compute before it means selling more output.

Seen through that lens, persistent hiring is less surprising. The first phase of the AI cycle demands engineers, sales staff, compliance teams, procurement specialists, construction workers and junior employees willing to work inside new software environments. Goldman flagged data-centre construction as one of the offsets already cushioning white-collar weakness. Even companies warning that office work will change are not yet behaving as if people have become optional.

Amazon offers a useful example of that gap between rhetoric and staffing. Business Insider reported that Amazon Web Services chief Matt Garman said half of white-collar jobs may “change” because of AI, while Amazon was also hiring more than 11,000 software development engineering interns and early-career engineers globally this year. That is not evidence that labour costs will stay high forever. It is evidence that the savings case is arriving later than the capex case.

Markets may eventually get the margin story they want. If AI tools mature from copilots into dependable workflow systems, some functions should need fewer people. Customer support, routine coding, marketing operations and administrative work still look exposed. Yet investors looking for that outcome in 2026 are trying to book the second-order effect before the first-order buildout has finished.

What investors should watch next

Three indicators matter more than headline layoff counts from here. First, revenue and output per employee need to improve in a way that persists after software spending is included, not before it. Second, entry-level hiring at heavy adopters needs to keep rising if the additive thesis is real. A one-off 12 per cent gain tells the market something; a repeat pattern would tell it much more. Third, adoption has to broaden beyond the 20.6 per cent of companies Goldman tracked in June before anyone can claim the labour effects are economy-wide.

Plenty could still break against the optimistic view. A weaker economy would make every labour dataset harder to read. Cost pressure could turn today’s complementary hiring into tomorrow’s rationalisation. Early-career workers may find that firms want AI fluency without offering enough apprenticeship to build it. None of those risks disappears because one study unsettled the dominant storyline.

For now, though, the cleanest reading is that AI’s first labour effect at the companies spending most heavily is additive, not subtractive. Corporate America is buying compute, borrowing for infrastructure and still paying up for people. That does not kill the layoff thesis forever. It does suggest the market is early in treating job cuts as the automatic reward for AI capex.