The People Who Build AI Are Turning on the Labs

DeepMind's UK staff voted 98% to unionize while Meta cut 8,000 jobs and raised AI spending toward $145 billion. The workforce is done being quiet.

The Q2 headline was capital — trillion-dollar valuations, record chip margins, IPOs. Underneath it ran a quieter and arguably more important shift: the people who actually build these systems started organizing against the companies that employ them.

In May, Google DeepMind's UK staff voted 98% to unionize — a first for a frontier lab. Not a fringe group: the researchers at one of the most important AI labs in the world, voting almost unanimously to bargain collectively.

The agentic era, charged to the workforce

It happened against a backdrop of cuts. Meta cut 8,000 jobs while raising AI capex toward $145 billion. Cloudflare cut 1,100. One day alone brought roughly 11,000 layoffs across the sector — many explicitly justified by "the agentic era," the same automation the cut workers had helped build.

The juxtaposition is the story: record spending on AI infrastructure, record cuts to the humans, and morale to match — Meta posted record profit alongside record-low employee morale and an AI reorg its own CTO reportedly called "atrocious."

Why the builders have leverage

Here's what makes this more than a labor footnote: frontier AI is bottlenecked on a few thousand people, and those people are now showing they will act collectively. When the workers who train the models can shape what the models are used for — military contracts, safety practices, what ships — that stops being an HR question and becomes a governance one.

The workforce was one of six forces we mapped in The State of AI · Q2 2026 — every claim sourced, a poll in each section.

— Alexis