Meta's buying more compute from Nebius while cutting headcount, and that gap between spending up and staffing down tells you everything about where AI work is actually happening.
The Signal
Meta just expanded its deal with Nebius, the neocloud that spun out of Yandex, right as reports surface about significant layoffs across the company. This isn't contradictory. It's the pattern. The company is reallocating spend from human labor to machine compute, and they're doing it through a relative unknown in the infrastructure space rather than the usual suspects. Nebius matters here because they're positioned as GPU-rich alternative infrastructure when AWS, Google Cloud, and Azure are all capacity-constrained. Meta needs the compute density to train and run their AI models, and they need it now, not on a waitlist.
Meanwhile, OpenAI is pursuing a joint venture with private equity to push enterprise adoption. That's code for: consumer AI has saturated faster than expected, margins are getting squeezed, and the real money is in selling to companies that can afford six-figure contracts. The PE structure suggests they need distribution muscle and enterprise credibility they don't have in-house.
Nvidia's GTC timing underscores all of this. Jensen Huang isn't just selling chips anymore. He's selling the entire compute substrate for this transition, and everyone from Meta to OpenAI to whoever PE backs next is racing to claim territory in an economy where the marginal cost of intelligence is collapsing toward zero.
The Implication
Watch the neocloud players. When tier-one companies start spreading infrastructure bets beyond the big three clouds, it signals real capacity constraints and a coming reshuffling of where AI actually runs. For workers, Meta's simultaneous compute expansion and workforce reduction is the template. Your company will spend more on AI infrastructure while employing fewer people. The question isn't if, it's how fast.
Source: Bloomberg Tech