> ## Content Index > Fetch the complete content index at: https://www.implicator.ai/llms.txt > Use this file to discover other available public pages before exploring further. # Nebius Buys Eigen AI for $643 Million to Strengthen Token Factory - URL: https://www.implicator.ai/nebius-buys-eigen-ai-for-643-million-to-strengthen-token-factory/ - Published: 2026-05-01T12:26:16.000Z - Updated: 2026-05-01T12:30:29.000Z - Description: $32 million — Per-employee price for Eigen AI's 20-person team, the premium Nebius is paying for inference know-how - Author: Marcus Schuler - Tags: AI News, #one-number, Funding [Nebius agreed Friday to acquire Eigen AI](https://nebius.com/newsroom/nebius-agrees-to-acquire-eigen-ai-strengthening-nebius-token-factory-as-a-frontier-inference-platform?ref=implicator.ai), a California inference-optimization startup, for about $643 million in cash and stock. The deal would fold Eigen's serving, system and kernel work into Nebius Token Factory, Nebius's managed service for running open-source AI systems in production. Nebius said the transaction is expected to close in the coming weeks, subject to customary conditions including antitrust clearance. Key Takeaways - Nebius agreed to buy Eigen AI for about $643 million in cash and stock. - Eigen's optimization work will move into Nebius Token Factory after closing. - The deal follows Nebius's February agreement to acquire agentic-search startup Tavily. - Nebius is pairing software acquisitions with a large AI data center buildout. AI-generated summary, reviewed by an editor. [More on our AI guidelines](https://www.implicator.ai/about/). ## Eigen moves inside model serving Eigen has 20 employees, Bloomberg reported, and was co-founded by alumni of MIT's HAN Lab, a research group known for efficiency research. Nebius said co-founders Ryan Hanrui Wang and Wei-Chen Wang will help establish its engineering and research presence in the San Francisco Bay Area. The price will be paid through cash and Nebius Class A shares, based on the company's 30-day weighted average stock price at signing. Nebius described Eigen as a way to improve the number of tokens generated per Nvidia chip when customers run AI models. ## Token Factory gets the target The acquisition follows a March [Nebius and Eigen partnership](https://nebius.com/blog/posts/nebius-and-eigen-ai-partner-to-accelerate-frontier-open-source-ai-inference?ref=implicator.ai) around optimized open-weight systems in Token Factory. That earlier work covered DeepSeek, GLM, GPT-OSS, Kimi, Llama, MiniMax and Qwen, and Nebius cited Artificial Analysis rankings for multiple fast implementations. One cited benchmark put Eigen's GPT-OSS-120B implementation at 911 output tokens per second. Another listed Qwen3 Coder 480B at 244 output tokens per second on a 10,000-token general workload and 374 output tokens per second on a 1,000-token coding workload. Nebius launched [Token Factory in November](https://finance.yahoo.com/news/nebius-launches-nebius-token-factory-124100461.html?ref=implicator.ai) as a production serving service for open and custom deployments. The release listed autoscaling endpoints, fine-tuning pipelines, access controls and deployment tools with Nebius's GPU infrastructure. Track AI Infrastructure Deals Strategic AI news from San Francisco. No hype, no "AI will change everything" throat clearing. Just what moved, who won, and why it matters. Daily at 6am PST. Email address Subscribe Check your inbox. Click the link to confirm. No spam. Unsubscribe anytime. ## Nebius keeps buying product layers The Eigen deal is Nebius's second acquisition agreement in three months. In February, Nebius [agreed to buy Tavily](https://nebius.com/newsroom/nebius-announces-agreement-to-acquire-tavily-to-add-agentic-search-to-its-ai-cloud-platform?ref=implicator.ai), an agentic-search provider, to add real-time web access to its AI cloud platform. CTech reported that the Tavily deal started at $275 million and could rise to $400 million if milestones were met. Roman Chernin, Nebius's co-founder and chief business officer, told Bloomberg Nebius wants products and features closer to direct customer usage. That means Nebius is trying to sell more than GPU rental capacity at a time when inference demand is shifting from experiments toward production applications. ## Capacity still controls the bet Nebius is also building the physical side of the business. Nebius announced a 310-megawatt AI factory in Lappeenranta, Finland, in March and said it is targeting more than 3 gigawatts of contracted power by the end of 2026\. CNBC reported that Europe faces higher energy prices and grid-connection challenges as AI data center plans multiply across the region. The strategy puts Nebius in the same neocloud category as CoreWeave and [Nscale](https://www.implicator.ai/nscale-raises-2-billion-in-record-european-round-adds-sandberg-and-clegg-to-board/), companies that rent specialized AI compute to model developers and enterprise customers. Nebius is betting Eigen gives it a software layer for each deployed chip. The next test is whether customers run enough production workloads to justify the capacity buildout now underway. Frequently Asked Questions What did Nebius buy? Nebius agreed to acquire Eigen AI, a California startup focused on inference and model optimization, for about $643 million in cash and stock. Why does Eigen AI matter to Nebius? Eigen works on serving, system and kernel optimizations that can help AI models generate more tokens per Nvidia chip, lowering serving costs for production workloads. What is Nebius Token Factory? Token Factory is Nebius's managed service for running open-source and custom AI systems in production, with autoscaling endpoints, fine-tuning pipelines and deployment tools. How does this fit Nebius's acquisition strategy? The Eigen deal follows Nebius's February agreement to acquire Tavily, an agentic-search provider. Both moves add product layers above raw GPU capacity. What still has to happen? Nebius said the Eigen transaction is expected to close in the coming weeks, subject to customary conditions including antitrust clearance. AI-generated summary, reviewed by an editor. [More on our AI guidelines](https://www.implicator.ai/about/). [AWS Bets the Farm on Vertical Integration, From Silicon to AgentsAt AWS re:Invent 2025, Matt Garman made a declaration that would have sounded absurd five years ago. "Agents are the new cloud," the AWS CEO announced during his keynote. Not chatbots. Not copilots. AThe Implicator![](https://www.implicator.ai/content/images/2025/12/freepik__op-art-illustration-in-the-style-of-roy-lichtenste__71398.png)](https://www.implicator.ai/aws-bets-the-farm-on-vertical-integration-from-silicon-to-agents/) [Anthropic's $50 Billion Bet: Infrastructure Arms Race or Strategic Bluff?Anthropic announced a $50 billion data center investment Wednesday. The timing reveals more than the number. Days earlier, OpenAI floated federal guarantees for its compute deals. Trial balloon, meetThe Implicator![](https://www.implicator.ai/content/images/2025/11/2025-11-12-10.38.27-Anthropic_Revenue_3@2x.webp)](https://www.implicator.ai/anthropics-50-billion-bet-infrastructure-arms-race-or-strategic-bluff/) [DeepSeek Cuts Inference Costs by 10xDeepSeek's sparse attention technique cuts long-context AI inference costs by 10x while slashing API prices over 50%.The Implicator](https://www.implicator.ai/deepseek-cuts-inference-costs-by-10x/)