AI leads US layoffs as cuts go uncounted

The Bureau of Labor Statistics reported +178,000 nonfarm payrolls in March, beating consensus of 59,000 ¹. Unemployment edged down to 4.3%. Health care added 76,000 positions. Construction added 26,000. Technology was not a growth sector.

Tech sector unemployment rose to 5.8%, the highest since the dot-com bust of 2001-02 ². Annual wage growth fell to 3.5%, the lowest since May 2021. Median tech reemployment time stretched to 4.7 months, up 47% from 3.2 months in 2024. When the same mid-level roles are eliminated across multiple employers, there is no adjacent firm to absorb the displaced.

Two economies are running in parallel: the broad labour market is hiring, and the knowledge-worker sector is contracting at its fastest rate in 25 years.

Oracle began cutting 20,000 to 30,000 jobs on 31 March, roughly 18% of its 162,000 global workforce ¹. The move freed $8 to $10 billion annually in salary costs. TD Cowen estimates the company has committed $156 billion to AI data centre infrastructure, adding to the $650-690 billion capital expenditure commitment from the five largest US tech companies .

Worker Adjustment and Retraining Notification (WARN) Act filings surfaced in Washington state (491 positions) and Missouri (539), but no Massachusetts filing has appeared despite Oracle's Burlington offices. Oracle has joined the pattern the 55% regret rate documented earlier this quarter : cutting on projected capability, not demonstrated return.

At 30,000 positions, Oracle's action is the largest single AI-attributed corporate reduction on record. Its workforce is now a funding mechanism.

Oracle terminated approximately 12,000 of its 30,000 India-based staff on 31 March, a 40% contraction of the company's largest non-US workforce ¹. The cuts arrived by a 6am email from "Oracle Leadership" with no prior contact from HR or direct managers.

These workers will not appear in BLS payroll data, initial claims, or any WARN Act filing . India produces more Oracle employees than any country outside the United States. The cuts fell hardest on Revenue and Health Sciences, SaaS and Virtual Operations Services, and the NetSuite India Development Centre.

If total US WARN filings remain below 3,000 for a 20,000-30,000 cut, it confirms the template: concentrate displacement in countries without disclosure requirements.

Morgan Stanley found that UK firms suffered net AI-driven job losses of 8% over the past year, double the international average, despite reporting identical productivity gains to US peers ¹. American firms with the same AI productivity improvements saw net job creation. British firms saw net job loss. The divergence has no satisfactory policy explanation.

UK vacancies had already fallen 9.5% year on year , and the Office for Budget Responsibility has modelled the worst case: 500,000 additional unemployed, £9 billion in extra government borrowing, no corresponding growth offset ². The Bank of England is planning to war-game an AI shock scenario in its stress tests. UK youth unemployment for 18 to 24 year olds has risen to approximately 14.5%.

Software developer vacancies have fallen 37% since the launch of ChatGPT, compared with 26% for other roles ³. The data so far: same productivity, fewer workers, higher youth unemployment.

A working paper by Seyed Mahdi Hosseini Maasoum and Guy Lichtinger, drawing on 62 million US worker resumes across 285,000 firms, found that entry-level job postings fell 15% in firms adopting AI tools ¹. Senior roles remained flat. The decline is driven by slower hiring, not increased firing. Companies are not sacking juniors. They are simply not replacing them.

This confirms the Federal Reserve Bank of Dallas finding that employment losses concentrate among workers under 25 through collapsed job-finding rates . In three to five years, the juniors never hired will not exist as mid-career professionals. Companies consuming this seed corn face a structural senior talent shortage they are not accounting for.

Fortune and Columbia University research confirmed that roughly 75% of unemployed Americans never file for unemployment insurance ¹. Severance packages delay filing. Recent graduates lack sufficient work history to qualify. Contractors are categorically ineligible.

Initial jobless claims fell to 202,000 for the week ending 28 March, a 10-month low ². New York's updated WARN Act, the world's first law requiring companies to disclose AI's role in mass layoffs, produced zero AI attributions from 162 companies covering 28,300 workers after nearly a year . The nine-senator coalition pushing for expanded BLS data collection is demanding better measurement from agencies whose existing tools are structurally blind to what they are meant to measure.

The National Bureau of Economic Research survey confirmed another dimension: executives use AI only 1.5 hours per week on average, yet project a 0.7% employment decline over three years . Workers at the same firms expect a 0.5% increase. Those planning the cuts and those absorbing them hold irreconcilable forecasts.

The Office for Budget Responsibility modelled a worst-case AI scenario: 500,000 additional unemployed and £9 billion in extra government borrowing, roughly the annual budget of the UK Home Office ¹. No corresponding growth offset appears in the model.

The Bank of England is planning to war-game a full AI shock scenario in its banking stress tests, assessing potential surges in household and company loan defaults. Its March agents' summary found task time savings of 5-20%, rising to 70% for highly automatable activities. Organisations are meeting demand without additional hiring.

The fiscal mathematics connects to the Brookings finding from Update #2 that 75% of US federal tax revenue derives from labour taxation . Official forecasters in both countries have run the displacement scenario. Neither result is informing the policy debate.

The AI Data Centre Moratorium Act introduced by Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez is dead on arrival, killed not by the Republican majority but by the Democratic caucus itself. Senator John Fetterman branded it "China First." Senator Mark Warner called it "idiocy" ¹.

Neither the moratorium nor the earlier robot tax proposal has a legislative path. What survives is measurement. The bipartisan nine-senator coalition led by Warner and Senator Josh Hawley wrote to the Department of Labour, the BLS, and the Census Bureau urging expanded AI workforce data collection. Federal agencies can act on this request without new legislation.

Salesforce cut customer support staff from 9,000 to 5,000 using AI agents and hired no new engineers in its 2026 fiscal year (ending January 2026) ¹. Sales headcount grew 20%. CEO Marc Benioff stated plainly: "I need less heads."

AI now handles roughly half of customer interactions on help.Salesforce.com, addressing 1.5 million inquiries in nine months and lowering support costs by 17%. Support is automated. Engineering is frozen. Revenue-generating sales staff expand. Salesforce has joined the 55% regret rate cohort acting on projected AI capability rather than proven return . The humans who remain are the ones who close deals.

A Fortune survey of more than 350 public-company CEOs managing $19 trillion in combined assets found that 66% plan to freeze or cut hiring through the rest of 2026 ¹. The figure aligns with the Atlanta Fed projection of 502,000 AI-attributed cuts , translating executive intention into corporate consensus.

Some 53% of investors expect AI returns within six months. Yet 84% of CEOs acknowledge meaningful returns require multiple years. Firms are cutting the HR and middle-management roles needed to define future jobs and redesign workflows, stripping out the supervisory capacity that managed previous technology transitions.

The EU Digital Omnibus faces its second trilogue on 28 April . The employer AI literacy obligation, stripped by Parliament on 26 March, remains contested. The final text determines whether EU workers have a guaranteed right to understand AI deployed against them.

China's Ministry of Human Resources and Social Security is preparing a dedicated AI employment policy: job-retention rebates, social security subsidies, and five targeted training programmes for 12.7 million graduates entering the labour market ¹. China faces a shortage of more than 5 million AI professionals, a supply-demand ratio of 1 to 10.

Europe debates disclosure rights. China deploys the state as a workforce intermediary. The United States does neither.

A bipartisan coalition of nine senators led by Senator Josh Hawley and Senator Mark Warner wrote to the Department of Labour, the Bureau of Labor Statistics, and the Census Bureau urging expanded AI workforce data collection.

Federal agencies can act on this request without new legislation. The letter requested specific AI attribution tracking in occupational and displacement surveys. If the BLS responds with a new survey methodology, it would close the measurement gap identified across all three prior updates, and potentially the New York WARN Act silence .