Between May 1 and May 10, 2026, more than 37,000 workers across the United States learned their jobs were gone. A budget airline that once carried 50 million passengers a year shut down for good. A cybersecurity company filed paperwork to cut a fifth of its staff in pursuit of what it called an “agentic AI-first operating model.” And a Gartner research report circulating widely this month delivered a finding that has drawn scrutiny across business media: companies that have laid off workers in the name of artificial intelligence have not yet realized measurable cost savings from doing so.
The layoffs keep coming anyway.
An airline collapses, a tech firm pivots
Spirit Airlines, once the largest ultra-low-cost carrier in the country, began winding down operations in early May after concluding it could no longer survive surging fuel prices. The airline employed roughly 17,000 people when it announced the shutdown, according to an AP News report on the closure. Flight crews, gate agents, mechanics, and corporate staff all received notices within days.
Spirit had filed for Chapter 11 bankruptcy protection in late 2024 and spent months trying to restructure, but volatile oil markets made the economics unworkable. This was not an AI story. It was a fuel-cost story, and it eliminated one of the most recognizable names in American air travel. But it matters here because it illustrates a broader reality: whether the cause is commodity prices or automation strategy, the result for workers is the same.
Days later, Cloudflare filed a Form 8-K with the SEC disclosing plans to eliminate approximately 20% of its workforce as part of its AI pivot. Based on the company’s most recently reported headcount in its annual filing, that translates to roughly 860 positions. The 8-K spelled out expected restructuring charges of $140 million to $150 million: between $105 million and $110 million in cash severance, plus $35 million to $40 million in non-cash charges tied to equity vesting acceleration.
That number deserves attention. Cloudflare will spend well over $100 million just to execute the layoffs, before a single dollar of AI-related savings shows up on a balance sheet. The restructuring is a bet on a future operating model, not evidence that one already works. The filing does not specify how many eliminated roles will eventually be replaced by AI agents, how many will simply disappear, or on what timeline the company expects automated systems to absorb the work those employees were doing. Cloudflare did not respond to a request for comment on those details.
Where the 37,000 figure comes from
Spirit and Cloudflare account for roughly 18,000 of the early-May total. The remaining job losses come from smaller announcements tracked by Layoffs.fyi, a crowdsourced tracker of technology-sector layoffs, and monthly reports compiled by the outplacement firm Challenger, Gray & Christmas. Not every one of those cuts is backed by a regulatory filing or detailed corporate disclosure, and some headcount figures from smaller firms may be revised as more information surfaces. The 37,000 number should be understood as an informed estimate assembled from those public trackers and filings, not a precise census. Readers seeking to verify the figure can cross-reference the Layoffs.fyi tracker and Challenger’s monthly job-cut reports for May 2026.
For context, the pace stands out. Challenger data from the first quarter of 2026 showed U.S. employers announcing roughly 60,000 to 65,000 job cuts per month. If early May’s trajectory holds through the rest of the month, May 2026 could rival or exceed those figures, which suggests the rate of workforce displacement is not slowing.
What Gartner found, and what remains unclear
The Gartner report drawing the most attention this month concluded that companies pursuing AI-driven workforce reductions have not realized net cost savings from those decisions. The finding has been cited across business media, and it aligns with what Cloudflare’s own SEC filing implies: spending $140 million or more on severance and restructuring is not a shortcut to efficiency.
There are important caveats. The specific report title and publication date have not been confirmed in a freely accessible primary document. Without knowing the sample size, how the firm defined “AI layoffs,” or the time horizon it measured, it is difficult to assess whether the conclusion applies broadly across industries or reflects a narrower group of early movers. The “not a cent” framing is striking, but it cannot be independently verified against raw data until the complete analysis is released. The link above points to Gartner’s newsroom rather than to the report itself, because the full study does not appear to be publicly available at this time.
What can be said with confidence is that the pattern Gartner describes matches the financial reality visible in public filings. Companies are absorbing large upfront costs to restructure around AI, and the efficiency gains they expect remain prospective. Whether those gains eventually arrive, and at what scale, is a question the current data cannot answer.
What the filings leave out
Cloudflare’s 8-K is specific about costs but vague about what comes next. The phrase “agentic AI-first operating model” describes a direction, not a staffing plan. The filing does not disclose whether the company intends to hire AI-focused engineers at comparable scale, redeploy existing employees into new functions, or permanently shrink its headcount. It also does not address how Cloudflare will maintain service quality and operational continuity during the transition, a gap that customers and investors will likely press the company to fill during upcoming earnings calls.
It is also unclear how representative Cloudflare’s move is of the broader technology sector. Some companies are deploying AI to augment existing teams rather than replace them. Others are trimming headcount for conventional reasons, like slowing revenue growth or shifting strategic priorities, without explicitly tying those decisions to automation. Without a comprehensive public dataset linking layoff announcements to AI adoption plans, the share of current tech job losses that is genuinely automation-driven versus traditionally motivated remains an open question.
Meanwhile, the people losing these jobs face a labor market that offers uneven support. The Bureau of Labor Statistics reported the national unemployment rate at 4.2% in its most recent release, a level that suggests the broader economy is still absorbing workers. But aggregate numbers obscure the reality for specific groups: mid-career tech employees, airline workers in regional hubs, and support staff whose roles are being automated face a much harder path to re-employment than the headline unemployment rate implies.
The bill comes before the savings
The clearest takeaway from the available evidence is not that AI will never deliver on its efficiency promises. It is that the transition is expensive, disruptive, and costly for workers long before any of those promises are tested. Spirit Airlines did not fail because of technology. It failed because jet fuel got too expensive for its business model to survive. Cloudflare is not failing at all. It is choosing to spend more than $100 million now on the belief that AI will make the company leaner and faster later.
In both cases, the immediate consequence is identical: tens of thousands of people are out of work, and the companies making these decisions have offered little public detail about retraining support, transition assistance, or accountability for the human cost involved. The Gartner report, whatever its methodological fine print, has landed hard because it names something many displaced workers already suspect: the savings that justified their layoffs may not exist yet, and may never arrive at the scale that was promised.
More than 37,000 jobs gone in ten days. Over $100 million in restructuring costs at a single company. And a growing body of evidence suggesting that the AI-efficiency narrative, however compelling in a boardroom presentation, has yet to prove itself where it counts: on a balance sheet.
