Skilled older workers are now training artificial intelligence systems because they ran out of money.
Let that sit for a second. Not because they wanted to pivot careers. Not because they saw an emerging field and got excited. Because they hit rock bottom and discovered that the companies building the future will pay them to feed it the present — their accumulated knowledge, their pattern recognition, their expertise — one labeled dataset at a time.
This is what demand for AI infrastructure looks like when you strip away the venture capital mythology. It's not a shortage of talent. It's a supply of desperation. And Meta, which owns nearly half of Scale AI (the outfit running these programs), has built a machine that converts human vulnerability into training data.
The implications sprawl everywhere. First: the narrative that AI will "create new jobs" is collapsing in real time. Not in theory — in practice. A 20-year engineer or a doctor or an economist doesn't retrain as an "AI labeler" because they see a bright future. They do it because the previous future closed. The labor market isn't expanding into new frontiers. It's compressing, and the compression is being captured as margin.
Second, and weirder: this is deflationary pressure disguised as inflation. Meta is outsourcing quality control to people so desperate they'll work for pennies relative to their former earning power. The cost of training the models gets lower just as the models become more capable. The company captures all the upside. The workers get survival wages. And because this labor is diffuse and invisible — scattered across freelance platforms, hidden in corporate earnings — it doesn't show up in employment statistics or wage growth. But it's happening at scale.
This connects to something the market keeps missing: the geopolitical pause everyone's celebrating (the two-week Iran ceasefire, the "only virtual talks" before Trump-Xi meet) is doing something perverse to incentive structures. When risk feels manageable, corporations stop preparing. They don't invest in contingency. They optimize for now. Which means they've already moved capital toward the paths of least resistance — like outsourcing training labor to whoever will do it cheapest.
The market sees reduced geopolitical risk and prices in calm. But underneath, the economic architecture is getting more fragile, not less. Companies are leaning harder into the efficiency play — which works great until it doesn't.
The wild part: this is exactly the opposite of what the pre-meeting messaging suggested would happen. Remember — "no new investment push" before the summit. Read that again in this context. Not "no big capital commitments." No new bets at all. Which means the capital that would normally create wage pressure upward is sitting still, waiting. And into that vacuum flows the desperation labor. The system is cannibalizing its own future labor supply to fund present margins.
If this pattern holds, what you're watching isn't a recovery. It's a hollowing. And it moves too slowly for traders to see, which makes it exactly the kind of shift that blindsides markets when it finally surfaces.
PREDICTION: The desperation labor narrative will accelerate hiring announcements from AI training firms (Scale, Outlier, etc.) in the next 48 hours, as the calm from the Iran ceasefire creates a false sense of hiring safety.