TL;DR

US tech layoffs over the past year have passed 165,000, according to tracker Layoffs.fyi, with Microsoft cutting 15,000, Amazon shedding 30,000 in six months, and Block removing more than 4,000 — about 40% of its workforce. Companies overwhelmingly point to AI productivity gains as the rationale. Economists and AI researchers interviewed by the Guardian are sceptical that the technology is actually doing the work the headcount cuts imply.

The Hype, the Cuts and the “AI-Washing” Question

Block CEO Jack Dorsey explicitly tied his February layoffs to AI productivity gains, and the company’s stock jumped 20% on the news before retreating 6% two weeks later as analysts questioned execution risk. Google has credited AI for 50% of its code in recent earnings, and Block’s head of engineering said 90% of code submissions were AI-assisted by November. The market’s instinct is to reward layoffs as a productivity story.

But Yale Budget Lab research director Ryan Nunn told the Guardian his team sees nothing “differentially happening with the AI-exposed labor market”, and MIT Sloan professor Thomas Malone warned that “many people are overestimating the rate at which jobs will change”. Marc Andreessen, normally an AI booster, conceded on a podcast this week that big tech firms “were overstaffed” and now have “the silver-bullet excuse: ah, it’s AI.”

Reliability Is Still the Ceiling

Princeton post-doc Stephan Rabanser, who has co-written on AI agent reliability, argues “reliability will be a key limiting factor” on job transformation, with even leading models producing inconsistent outputs from identical prompts. Wharton’s Ethan Mollick noted that some firms are now running so-called “dark factories” — code shipped without human review — despite the well-documented failure modes.

Looking Forward

For UK businesses, the takeaway is to read the Wall Street layoff narrative carefully. Cuts framed as “AI replacing roles” often look more like overhang from pandemic-era hiring bumping into weaker consumer demand. The other risk is structural: even where AI genuinely lifts coding throughput, firms are now falling behind on code reviews because reviewers cannot keep pace with generated output. UK firms tempted to mirror the US headline approach should price in both the AI-reliability ceiling and the second-order workload that comes with it before cutting deeply.