Skip to content
NewsletterAI Learning, Keeping It HumanCall647-641-0643
Christopher Ross

Guides

AI is in the budget more than in the headcount

By 12 min read

AI is hitting early-career hiring. The evidence that it is driving this year's layoffs is thinner: employers named it in about 21 percent of announced U.S. cuts, and independent data shows little economy-wide effect so far. My reading is that AI is in the budget more than in the headcount.

Key takeaways

  • AI is showing up in early-career hiring. Stanford researchers find employment among workers aged 22 to 25 in highly AI-exposed occupations about 19 percent below where it would be had it tracked similarly aged workers in less-exposed jobs, and it appears to come mainly from reduced hiring. They call it a descriptive pattern and make no causal claim.
  • The evidence that AI is driving the layoffs themselves is thinner. Employers cited AI in about 21 percent of announced U.S. job cuts through September 2026, but that counts what employers say, and nobody has audited the cause. Independent measures of the whole labour market show little economy-wide AI effect so far.
  • My reading is that AI shows up in the budget more than in the headcount. Meta cut about 8,000 jobs in May 2026, as Fortune reported, and its capital spending guidance in July was $130 to $145 billion, while its headcount on 30 June was down 1 percent from a year earlier. That fits cuts tied to AI spending. It does not show it.

I drafted a short opinion for LinkedIn. It said I don't think AI is responsible for recent job losses, at least not to the extent companies are reporting, and that Facebook has been bleeding users for a decade, so "AI took the jobs" is an easier story to tell than "the market has shifted under us." Then I went and read the numbers. The Facebook claim was wrong, and seeing why changed the argument. I had been treating legacy brands in decline as the real story and AI as the cover. What the evidence supports is narrower: "AI layoffs" bundles two different claims. One is that AI is doing work that people used to do. The other is that AI spending is crowding out other spending, payroll included. The first has support in one place, early-career hiring. The second has a plausible case at one company.

There are real people in these numbers, and I wrote about that side of it in July, in the piece on what Cape Breton's dying coal towns taught me about the AI layoffs. This one is about what the evidence supports, which is a separate question from whether anyone is hurting.

What the announcements count

Challenger, Gray & Christmas has counted announced U.S. job cuts for decades. Its September 2026 report says employers cited AI in 120,136 announced cuts for the year to date, "approximately 21% of all cuts," out of 573,195. AI "remains the leading reason year-to-date." Technology accounts for 165,925 of all announced cuts, about 29 percent. Challenger's July report put it this way: "Tech remains the center of gravity for this year's cuts, and AI is still the reason companies give."

That number counts reasons given in announcements. It is not a measurement of why anyone lost a job, and Challenger makes judgment calls inside it: a separate bucket, "Technological Update (possibly AI)," holds announcements where new technology is the stated reason and AI is only implied. I go through how the count works in its own piece.

Two things follow from that. The share swings: AI was 33 percent of July's announced cuts and 9 percent of September's. And the total shrank, down 39 percent on the same period of 2025, which makes any slice look bigger. The AI count itself did rise, from the 4.5 percent Oxford Economics found for 2025 to 120,136 cuts in nine months of 2026, so employers are saying "AI" far more often than they did a year ago. The question is whether the labour market looks like it agrees.

Where the wider data lands

Two research groups looked at the whole labour market this year and landed close together. A third looked at one slice of it and found something different.

Oxford Economics published a briefing on 7 January 2026. It called AI-related losses "still relatively limited" and made a productivity argument I find hard to dodge: "If AI were already replacing labour at scale, productivity growth should be accelerating. Generally, it isn't." It also suggested a motive for the framing, which some now call "AI washing": "We suspect some firms are trying to dress up layoffs as a good news story rather than bad news, such as past over-hiring." Cuts tied to "market and economic conditions" ran about 245,000 in the same period, roughly four times the AI figure (as reported by Fortune, 7 January 2026).

Yale Budget Lab looked at how the mix of jobs has shifted since ChatGPT arrived and at how long highly AI-exposed workers stay unemployed. Its conclusion, as Fortune reported it on 2 February 2026, was "stability, not major disruption at an economy-wide level." Its executive director, Martha Gimbel, said attributing layoffs to AI can help executives avoid explaining trouble with tariffs, immigration limits and policy uncertainty, and added: "it's not helpful to declare that it's here before it's here."

Stanford's Digital Economy Lab looked at one slice, workers by age and by how exposed their occupation is to AI. In an August 2026 update, Erik Brynjolfsson, Bharat Chandar and Ruyu Chen report that employment among workers aged 22 to 25 in highly AI-exposed occupations is about 19 percent below the level it would otherwise have reached, relative to similarly aged workers in less-exposed occupations. The shortfall was 15 percent in the July 2025 data. Experienced workers show no comparable gap. The adjustment "appears to operate primarily through reduced hiring of young workers rather than increased separations." The authors describe their results as "descriptive patterns, not causal estimates of the effect of AI," and note that some differential trends predate widespread generative AI use.

That is the only place I found where the data and the headline line up, and it describes a hiring story, not a layoff story. The 22-year-old who never got the interview has no announcement to point to. I take it up in its own piece.

Read the memos

The more useful evidence is what executives wrote when they cut. In November 2022, Mark Zuckerberg wrote that "I got this wrong, and I take responsibility for that," after predictions of a permanent pandemic acceleration in e-commerce did not hold (Meta, 9 November 2022). In January 2023, Sundar Pichai wrote that Alphabet had "hired for a different economic reality than the one we face today" (Google, January 2023). Both memos explained the cuts by what the company had done. I lay those out, with the interest-rate timeline, in a separate piece.

The 2026 explanations are different, and Meta's are the most documented. Fortune's reading of Zuckerberg's May memo is that it stressed competition and that "success isn't a given" in the AI race. Reuters reported that the cuts were partly meant to offset rising AI infrastructure costs, and Fortune's article does not cite AI replacing workers as a reason (Fortune, 21 May 2026). Meta's second-quarter filing supplies the numbers. Costs rose 55 percent to $42.03 billion, including $1.18 billion of severance for the May reduction. Capital spending guidance as of July is $130 to $145 billion. Headcount was 75,472 on 30 June, "a decrease of 1% year-over-year" (Meta Form 8-K, 29 July 2026). The filing does not say how much of the May reduction had taken effect by then, and a year of hiring sits inside that comparison, so I read the headcount as roughly flat and no more than that.

Amazon says something different again. On its October 2025 earnings call, Andy Jassy said the roughly 14,000 cuts announced days earlier were "not really financially driven" and "not even really AI-driven, not right now at least." He called it culture (Fortune, 1 November 2025). The same article reports a Goldman Sachs study of its corporate clients, which found that only 11 percent were actively cutting jobs owing to AI.

Sorting the evidence by the two claims, AI doing the work and AI spending crowding out payroll, is my reading, and I am labelling it as one. On the first, the company filings and memos I read do not say AI is doing the work the laid-off people did. On the second, AI spending plausibly shaped the cuts at Meta, but I could not establish whether the payroll savings are large enough to matter next to the capital spending, and Amazon says its own cuts were about culture. So the crowding-out reading is consistent with the Meta evidence. It is not proven by it.

Where my own draft was wrong

I used the "bleeding users" line to argue that legacy brands are in trouble and AI is a convenient cover story. The evidence does not support the first part. Pew's December 2025 survey of 1,458 U.S. teens found 31 percent use Facebook, "far lower than the 71% in 2014-15, though on par with 2022." Meta's 3.60 billion daily active people in June 2026, up 3 percent from a year earlier, is a figure for its whole family of apps, so it cannot speak to Facebook alone. It does not describe a company losing its audience, though.

So "legacy brands are vulnerable" is a claim about the next generation of users. The Facebook piece has the detail.

A slow market, with or without AI

The wider market is soft. The Federal Reserve raised its target range from 0.25 to 0.5 percent in March 2022 to 5.25 to 5.5 percent in July 2023, and the Bureau of Labor Statistics reported that the economy added 29,000 payroll jobs in September 2026, with unemployment at 4.2 percent (BLS, 2 October 2026). In a market that slow, an employer can hire fewer 22-year-olds without any model involved. The pandemic piece has the rate timeline in full.

Where I land, and what would change my mind

I think the opinion I started with is half right. Employers are naming AI in a far larger share of announcements than the aggregate labour data supports, and the pattern fits a motive Oxford Economics named: it reads better to investors than "we over-hired" or "demand is soft." I am not saying AI has no effect on work. I think its effect on early-career hiring is real enough to take seriously. Neither that nor the spending at Meta is the story the headlines tell, in which a model does the job and the person leaves.

Four things would move me: productivity growth that actually accelerates, layoff filings that name the function an AI system took over with output numbers to match, Stanford's gap opening up for experienced workers, or a company reporting headcount falling while a rising workload is handled by named AI tools. I would like to see any of them, and I would update this piece when I see one.

A household that cancels a gym membership to pay for a new roof still loses the gym. That is how I read much of 2026, and I may be wrong about how much of the roof is AI.

How this was made. AI tools did the source reading and checking for this piece, and AI reviewers argued against the draft before it went out. The question is mine, and so is the decision to publish it.

Related reading

Common questions

Is AI causing layoffs in 2026? Employers say so in a large share of announcements: Challenger counted 120,136 AI-attributed cuts through September, about 21 percent of the total. Independent measures of the whole labour market, including Oxford Economics and Yale Budget Lab, show little economy-wide effect so far. The clearest AI-linked signal is a drop in hiring of workers aged 22 to 25 in highly AI-exposed jobs, which Stanford researchers describe as descriptive, not causal.

What does "AI washing" mean? It is the practice of crediting AI for job cuts that have other causes, such as over-hiring, weak demand or higher costs. Oxford Economics suggested in January 2026 that some firms do this because it conveys a more positive message to investors. The evidence is suggestive across firms, not proof about any single company.

Are companies cutting jobs to pay for AI? Meta cut about 8,000 jobs in May 2026 and guided to $130 to $145 billion of capital spending in July, and Reuters, as relayed by Fortune, reported the cuts were partly meant to offset AI infrastructure costs. Whether the payroll savings are large enough to matter is not established, and Amazon's CEO said its own cuts were about culture and not AI.

References

Checked against the sources above, as listed, on 2026-10-29.