Why the Numbers Don’t Match
If you’ve been following AI layoff coverage in 2026, you’ve seen at least four different aggregate figures for Q1 alone. They aren’t wrong. They’re measuring different things, using different methodologies, across different time windows. The problem isn’t bad data. It’s that “AI layoffs” isn’t a single category with a single count.
Here’s what each major source is actually reporting.
The Four Competing Q1 Figures
Source 1: Challenger, Gray & Christmas (15,341)
Challenger’s March 2026 report attributed 15,341 of 60,620 total U.S. job cuts to AI as the employer-stated reason. That’s 25% of all March layoffs. This is the most conservative figure because it counts only one month and only includes cuts where the employer explicitly named AI.
Source 2: Challenger Q1 Aggregate (27,000)
Challenger’s quarterly roll-up reports approximately 27,000 AI-attributed cuts for full Q1 2026, up 40% year-over-year. This extends the same methodology across three months but remains limited to employer-stated AI attribution.
Source 3: Challenger via Outlets (52,050)
Multiple outlets reported Challenger’s broader Q1 tech-sector total at 52,050. This figure includes all tech layoffs where AI was cited as a contributing factor in reporting, not just as the employer’s stated primary reason. The gap between 27,000 and 52,050 is the difference between “employer said AI” and “reporting context included AI.”
Source 4: Tom’s Hardware / MedhaCloud (78,557-85,000)
Tom’s Hardware reported 78,557 total tech layoffs in Q1 with approximately 47.9% attributed to AI (~37,600). MedhaCloud’s tracker cited 85,000+. These are the broadest counts, including companies where AI was mentioned in any coverage context, not necessarily as the stated cause.
What the Hub Tracks: Confirmed Company-Level Events
The Hub’s Job Displacement Tracker doesn’t compete with aggregate estimates. It tracks confirmed company-level events with sourced attribution. Here’s what’s verified as of April 24, 2026:
AI-Direct (employer explicitly cited AI):
• Block: ~4,000 (40% of workforce). CEO Dorsey stated AI replacing tasks.
• Atlassian: ~1,600 (10%). CEO cited “changed skills mix for AI era.”
• Challenger March aggregate: 15,341 employer-stated AI cuts.
AI-Adjacent (AI cited in reporting context, not confirmed as primary cause):
• Meta (April 23): ~8,000 confirmed cuts + 6,000 frozen roles. Internal memo cites “efficiency”; AI CAPEX context is real but not the stated cause.
• Oracle: “Up to 30,000” reported; company hasn’t confirmed exact total or AI attribution.
• Crypto.com: ~12% of workforce; AI adoption cited, causation debated.
Projections (not actual cuts):
• NBER/CFO Survey: 502,000 projected AI-driven cuts for full-year 2026 (9x the 2025 baseline of ~55,000). This is a survey-based forecast from 750 CFOs, not a count of actual layoffs.
• Brookings/NBER: 37.1 million workers in high-exposure roles. This is exposure data, not displacement data.
The Attribution Problem
Every aggregate number above requires a judgment call: does “AI layoff” mean the company said AI was the reason, or that reporting mentioned AI in the same context? The gap between those definitions explains most of the numerical disagreement.
The Hub uses a four-tier attribution framework: ai-direct (employer stated), ai-adjacent (AI context without confirmed causation), mixed (multiple factors), and business (no AI connection). When you see different numbers in different Hub briefs, check which tier the source is using. A brief citing Challenger’s employer-stated data will always show a lower number than one citing Tom’s Hardware’s broader methodology.
How to Use This Brief
This is a living reference. When a new layoff brief publishes on the Hub, it links here for aggregate context rather than asserting its own total. The Job Displacement Tracker maintains the event-level data. This brief maintains the methodology reconciliation.
If you’re citing AI layoff numbers in your own analysis, name the source and the methodology. “Q1 AI layoffs” isn’t one number. It’s at least four, and the one that’s right depends on what you’re trying to measure.