Amsterdam, May 1, 2026, 22:07 CEST
• Nebius is picking up Eigen AI in a cash-and-stock transaction that puts the deal’s value around $643 million.
• The acquisition is all about AI inference—the deployment of trained models into live applications.
• Shares of Nebius jumped, with the market digesting the move before the May 13 earnings release.
Nebius Group N.V. is set to buy California-based Eigen AI for roughly $643 million, a move that folds the AI optimization startup into the Amsterdam cloud provider’s portfolio and boosts its presence in production-ready AI software.
This deal is grabbing attention as the AI industry shifts past the phase of training large models and dives into inference — that’s deploying those trained models to respond to prompts, generate predictions, or power customer-facing products. Deloitte projects that inference will drive about two-thirds of the demand for AI compute, or data-center processing, by 2026.
Nebius isn’t just focused on adding to its stockpile of graphics processing units—the chips essential for training and operating AI. The company is also pushing to squeeze as much performance as possible out of every chip, a key concern now that cloud firms are all scrambling for limited power, hardware, and clients prepared to commit to long-term deals.
According to Bloomberg, Nebius is set to acquire Eigen AI for roughly $98 million in cash, plus 3.8 million Nebius shares, pegged to Nebius’s 30-day weighted average. The closing is expected in the coming weeks, pending standard approvals like antitrust review, the company said.
Nebius climbed 11.3% to $153.84 late in the U.S. session, having earlier reached as high as $155.99.
Roman Chernin, co-founder and chief business officer at Nebius, described the AI landscape as a “capacity-scarcity world.” Pairing Eigen’s optimization tools with Nebius’s infrastructure, he said, should lead to better model performance and stronger unit economics. For Eigen AI, CEO and co-founder Ryan Hanrui Wang pointed to the deal making life easier for developers—letting them “run models reliably in production” without having to manage infrastructure themselves. Nebius
The founding crew at Eigen AI comes out of MIT’s HAN Lab. Nebius plans to tap their expertise to set up an engineering and research team in the San Francisco Bay Area. Bloomberg put Eigen’s headcount at around 20, describing it as a startup focused on boosting open-source AI model performance.
Nebius is looking to climb higher up the AI infrastructure stack with this acquisition, shifting from simply renting out raw compute to delivering managed services for model deployment and tuning. The company’s Token Factory offering targets customers interested in running and customizing open-source models, minus the need to create that supporting layer on their own.
The field is getting crowded. Reuters calls Nebius and CoreWeave specialist AI cloud outfits, noting their strong focus on tech clients who need GPU and power resources—unlike the more general-purpose cloud giants. Over at Investor’s Business Daily, CoreWeave and Lambda show up as key cloud-infrastructure startups catering to those building AI models and apps.
Nebius’s rapid growth has been fueled by hefty infrastructure deals. Back in March, Reuters said Meta lined up a $12 billion purchase of AI computing capacity from Nebius, stretching out to 2027. There’s also an option on the table that might boost the deal to $27 billion. As for Nvidia, Reuters reported a $2 billion investment for an 8.3% slice of Nebius.
The company’s been pulling in more cash to back its expansion. On March 23, Reuters said Nebius sealed a $4.34 billion convertible debt deal and was targeting between $16 billion and $20 billion in capital spending for 2026. That growth, according to the report, would be bankrolled with customer prepayments plus a blend of equity and debt.
Risks remain. The Eigen transaction is still awaiting approval, and M&A outcomes often falter if integration issues crop up—whether it’s people or products. Nebius itself points to uncertainties: execution on deals, hanging onto clients, funding requirements, market volatility, and fast-moving technology. All of it matters in a market racing to add capacity without much clarity on future demand.
Nebius, for its part, has scheduled the release of its first-quarter 2026 results for May 13, before the opening bell, and plans a conference call later that day. Investors will be watching the report for any signs that the company’s recent string of deals is delivering revenue at a pace that keeps up with its hefty outlays.
