Nebius buys Inferize in $100-150M stealth AI GPU play

Nebius has acquired Inferize, a 10-month-old Israeli AI startup, for $100-150 million. The deal, first reported by CalcalisTech, brings Nebius a team of 17 engineers—former Granulate executives—and technology designed to reduce idle GPU capacity during fluctuating AI workloads. TechGraph notes the purchase is aimed at bolstering Nebius’ Token Factory, its suite of tools for scaling AI models more efficiently.

Inferize had no public funding, no product launch, and no press coverage before this deal. That Nebius paid nine figures for a pre-revenue team suggests it sees GPU optimization as a critical differentiator, not just another feature.

Nebius, a cloud infrastructure provider, has been positioning itself as an alternative to dominant cloud platforms. Unlike some of its rivals, Nebius doesn’t manufacture its own AI chips. Instead, it’s betting on software layers to squeeze more performance out of existing GPU fleets. Inferize’s technology—dynamically reallocating idle capacity when demand spikes—fits neatly into that strategy. It’s a play for cost efficiency, not hardware innovation, and one that could appeal to customers wary of vendor lock-in with established players.

The timing is interesting. Just last week, another AI startup secured significant funding for energy-efficient AI inference, signaling that the market is still hungry for alternatives to dominant hardware providers. But Nebius isn’t chasing hardware—it’s doubling down on the software stack, where margins may be thinner but the barriers to entry are lower. That could be a smart move if cloud customers prioritize flexibility over raw performance.

Still, questions remain. Nebius hasn’t disclosed how Inferize’s technology will integrate with Token Factory, or whether it will be offered as a standalone product. The lack of public details about Inferize’s tech makes it hard to gauge how much of the $100-150 million price tag is for the team versus the IP. And while GPU optimization is valuable, it’s not clear how defensible Inferize’s approach is—competitors have been tackling similar problems for years.

For Nebius, the acquisition is less about immediate revenue and more about signaling ambition. If that’s true, Nebius may have just bought itself a seat at the table. If not, it’s an expensive gamble on a team that hasn’t yet proven its worth in the wild. Either way, the deal is a reminder that in AI, even the quietest companies can command big prices—if they’re solving the right problem.

Sources: calcalistech.com · techgraph.co