Nearly 140,000 jobs have been cut across US technology companies in the first seven months of 2026. A striking number of the announcements named artificial intelligence as the reason.
Whether AI is actually doing the work of the people who were let go is a different question, and the evidence for it is thinner than the announcements suggest.
The scale, and who accounts for it
TechCrunch’s running tally, updated 25 July 2026, puts the total near 140,000 for the year to that point. Four companies — Amazon, Oracle, Meta and Microsoft — account for roughly 50,000 of them.
The individual announcements:
- Oracle: 21,000 roles over the twelve months to June 2026, about 13 per cent of staff.
- Meta: 8,000 roles on 20–21 May, around 10 per cent of the affected division, while moving 7,000 staff into AI-focused positions.
- Microsoft: 4,800 roles on 9 July, 2.1 per cent of its global workforce. Chief financial officer Amy Hood: “AI is changing how work gets done.”
- Cloudflare: 1,100 staff on 7–8 May — 20 per cent of the workforce — announced after record quarterly revenue.
- Monday.com: 600 employees on 25 July, about 20 per cent of staff.
Two of those deserve a second look, because they cut in opposite directions.
Cloudflare: the most specific claim anyone made
Cloudflare cut a fifth of its workforce after posting record revenue. Chief executive Matthew Prince said most of those cut were “measurers” — middle managers whose function was monitoring and reporting on other people’s work, a task he argued AI tooling now performs.
This is the most substantive version of the AI-displacement claim, because it identifies a mechanism. Not “AI made us more efficient” in the abstract, but: here is a category of work — aggregating status, compiling reports, tracking progress — that software now does adequately.
It is also testable, in a way vaguer claims are not. If middle-management layers are genuinely being automated, that should show up in occupational data over the next few years, not just in press releases.
Monday.com: the company that said the opposite
Monday.com cut 600 people and its co-founder Eran Zinman explicitly denied the fashionable explanation, saying the move “was not made to reduce costs or replace people with AI.”
That denial is informative precisely because it was unnecessary. When a company goes out of its way to say a layoff was not about AI, it tells you the default assumption had become that it was.
The number that complicates the story
TechCrunch noted something that sits awkwardly with the “markets reward AI-forward companies” thesis: firms citing AI as a layoff reason underperformed the Nasdaq by roughly 10 per cent in the thirty trading days after announcing.
If invoking AI were purely a way to please investors, you would expect the opposite. One plausible reading is that investors distinguish between a company restructuring from strength and one using a fashionable justification for cuts it needed anyway.
The case that this is “AI-washing”
The sceptical argument has three parts.
First, the timing fits an older explanation. The technology sector hired aggressively through the pandemic years on demand assumptions that did not hold. Correcting that overhang produces layoffs whether or not AI exists. AI arrived as an available narrative at the moment the correction was due.
Second, the tooling is not obviously ready. Deploying AI systems that genuinely replace a role — reliably, at production quality, with the integration and oversight that implies — is a multi-quarter project. A significant number of the companies announcing AI-driven cuts have not been running such systems long enough for the substitution to have happened.
Third, the people who study labour markets are not seeing it. Grant Zallis, Managing Partner at DHR Global, put it plainly: “I’m still not seeing a correlation between AI productivity and those layoffs.”
There is also a straightforward institutional incentive. “We are restructuring around AI” is a story of transformation. “We over-hired and the market has softened” is a story of managerial error. The first is easier to tell to staff, press and shareholders.
The case that it is real
Against that, Cindi Howson, Chief Data and AI Strategy Officer at ThoughtSpot, argues displacement is genuinely occurring and that markets do reward companies seen as AI-forward. CNBC reported on 5 June 2026 that AI had become the leading reason companies give for cutting jobs in some tracking surveys.
And the “it’s just over-hiring” explanation weakens over time. Post-pandemic correction was a reasonable account in 2023 and 2024. By late 2026, a sector still shedding six figures of jobs annually is describing something structural rather than a one-off adjustment.
What the ILO actually said
The International Labour Organization published a research brief on 17 April 2026 that is more careful than most of the commentary citing it.
Its central methodological point: AI “exposure” indicators are early-warning signals, not job-loss forecasts. An occupation being exposed to AI capability means tasks within it could in principle be automated. It does not mean they will be, or that the job disappears — roles typically get reshaped rather than deleted.
The ILO’s substantive finding is the notable one. Earlier automation waves fell hardest on routine, lower-skilled work. Newer AI-capability measures point instead at white-collar, cognitive occupations: analysis, drafting, coordination, summarising. Exactly the “measurer” work Cloudflare described.
The ILO does not claim AI is currently causing the job losses. It says the exposure profile has moved up the skill ladder, which is a different and more consequential statement.
The wider labour market
This is not happening in a strong economy. US nonfarm payrolls fell by 23,000 in July 2026, and May and June figures were revised down by a combined 103,000, according to the Bureau of Labor Statistics.
A softening labour market matters for interpretation. When hiring is strong, a laid-off engineer finds another job and the disruption is individual. When it is weak, the same layoff compounds.
Why it is hard to settle
The honest answer is that the data required does not yet exist in usable form.
Establishing that AI caused a job loss requires knowing what the person did, what the AI system now does, and what would have happened otherwise. Employers hold the first two and rarely publish them. Official statistics record occupations and headcounts, not the reason a role was eliminated. And the stated reason comes from the party with the strongest interest in how it sounds.
What can be said with confidence: the layoffs are real and large; the AI attribution is frequently asserted and rarely demonstrated; and the exposure of cognitive white-collar work is a documented shift regardless of what is driving this particular year’s cuts.
What to watch
- The BLS employment situation report for August 2026, due 4 September — whether the softening continued into late summer.
- Productivity data. If AI is replacing labour at scale, output per worker should rise measurably. That is the check on the narrative.
- Occupational detail. Whether middle-management categories specifically shrink, which would support the “measurers” thesis.
- Rehiring. Companies quietly rebuilding roles they cut would suggest the automation did not hold.
Why productivity data is the test that settles this
There is a clean way to check whether AI is genuinely replacing labour, and it does not depend on trusting anyone’s press release.
If a company sheds ten per cent of its workforce and maintains output, output per worker rises. That is measurable, reported in national statistics, and difficult to spin.
If AI were substituting for labour at the scale implied by the announcements, aggregate productivity growth should be visibly accelerating. Historically, general-purpose technologies do eventually produce exactly that signature — but with a lag, because the gains require reorganising how work is done, not merely installing software.
The economist Robert Solow’s line about the earlier computer era — that the age was visible everywhere except in the productivity statistics — described precisely this lag, which took roughly a decade to resolve.
So a reader watching productivity data has two hypotheses to distinguish, and patience is required for both. Sustained acceleration would vindicate the substitution story. Continued flatness alongside continued layoffs would support the view that something else is driving the cuts and AI is the label attached to them.
What “AI does this job now” actually requires
The gap between a capable model and a replaced role is wider than the announcements imply, and it is worth itemising.
Integration. The system must connect to the actual data, tools and workflows the job uses. In most organisations that is the expensive part, and it is where deployments stall.
Reliability at the tail. A model that handles 80 per cent of cases well does not replace a role; it changes it. Someone must still handle the remaining 20 per cent, and identifying which cases those are is itself work.
Accountability. Someone must be answerable for the output. In regulated contexts, that person must be able to explain a decision. This is often the binding constraint regardless of model capability.
Maintenance. Deployed systems drift, break on input changes, and need monitoring. The staff doing that are a cost that offsets part of the saving.
None of this argues that displacement will not happen. It argues that the timeline between capability and displacement is measured in years, and that announcements made months after adopting a tool are describing an intention rather than an accomplished substitution.
Who is actually exposed
The ILO’s finding that exposure has shifted toward cognitive white-collar work inverts the assumption most people carry about automation, and it is worth taking seriously.
Previous waves displaced physical and routine work: manufacturing lines, clerical processing, retail checkout. The consistent advice was to move up the skill ladder into knowledge work.
Current capability profiles point at that knowledge work: drafting, summarising, first-pass analysis, coordination, translation, code generation. These are tasks performed by people with degrees, in offices, who assumed they were on the safe side of the divide.
Two qualifications keep this from being a prediction of collapse. Exposure is measured at the level of tasks, and most jobs are bundles of many tasks — automating several changes a role rather than eliminating it. And the tasks least exposed are those requiring physical presence, direct human relationship, or accountability that cannot be delegated to software — which is not the same list as the ones that pay best.
What a worker can reasonably conclude
Two things, and neither is comfortable.
The first is that the current wave of layoffs is happening in a softening labour market regardless of its cause. Whether AI or over-hiring is responsible does not change the position of someone who has been let go, and it does change how long they are likely to look.
The second is that the AI attribution, whatever its accuracy, has consequences of its own. A layoff framed as technological restructuring is treated differently by employers, by policymakers and by the affected workers than one framed as a business error. It shapes what retraining is offered, what support is thought appropriate, and how a gap on a CV is read.
That is why the accuracy of the attribution matters beyond bookkeeping. The story a company tells about why it cut jobs becomes part of the environment the people it cut have to work in.
Sources
- TechCrunch, “Monday.com is the latest tech company to blame AI for layoffs — here are 20 others,” updated 25 July 2026 — techcrunch.com
- Fortune, “Cloudflare posted record revenue, then cut 20% of its workforce,” 21 May 2026 — fortune.com
- International Labour Organization, “New ILO brief explains what AI exposure indicators reveal about jobs,” 17 April 2026 — ilo.org
- CNBC, “AI is now the leading reason companies give for cutting jobs, says new report,” 5 June 2026 — cnbc.com
- TechTarget, “AI, economic pressures and other causes of Big Tech layoffs,” 2026 — techtarget.com
