Nebius Acquires Clarifai Team, Licenses Inference IP

Amsterdam-based AI cloud provider Nebius announced on May 12, 2026, that the core engineering and research team from Clarifai is joining the company. The deal includes a license to Clarifai's inference and compute orchestration technology, marking a significant expansion of Nebius's full-stack inference platform.

Clarifai founder and CEO Matthew Zeiler will join Nebius as Senior Vice President of Research. Zeiler, a recognized pioneer in machine learning who has worked alongside researchers including Geoffrey Hinton, Jeff Dean, Rob Fergus, and Yann LeCun, will lead a team focused on frontier AI innovation. His research areas include multimodal agentic reasoning, world models, token efficiency, and long-term memory.

According to the official Nebius press release, a select group of Clarifai engineers and researchers will also join Nebius's infrastructure teams. They bring more than a decade of expertise in inference optimization and machine learning to the organization.

This transaction follows Nebius's recently announced acquisition of Eigen AI. The company is positioning itself with a clear division of labor: Eigen AI optimizes at the model level, while Clarifai's technology optimizes the system. The combination creates the end-to-end infrastructure required to run complex AI models reliably in production.

Roman Chernin, co-founder and Chief Business Officer of Nebius, explained the strategic logic. "We are building a complete inference ecosystem, because delivering efficient execution at scale is a system optimization game: model optimization, system design, and compute orchestration all have to work together." The integration of Clarifai's advanced system-building capabilities will further strengthen Nebius Token Factory, offering customers the infrastructure they need to run models reliably and cost-effectively in production.

Alongside the talent acquisition, Nebius will acquire Clarifai's patent portfolio covering AI inference, compute orchestration, and related technologies. The company will receive a non-exclusive, perpetual license to Clarifai's modern AI inference and reasoning technology stack. Commercial terms of the agreement were not disclosed.

The scope of the license has important limitations. It is limited to Clarifai's modern AI inference and compute orchestration technology. It does not include Clarifai's legacy computer vision models, any intellectual property, products, services, or commercial arrangements associated with Clarifai's US government and defense programs. This carve-out suggests Nebius is targeting commercial enterprise customers rather than defense contracts.

Zeiler's statement frames the deal in infrastructure terms. "The future of AI — from agentic systems to physical AI — depends on the infrastructure powering it. Nebius is building the ultimate foundation to become the next hyperscaler." He emphasized that combining deep experience in compute orchestration and system-level optimization with Nebius's massive compute capacity will give developers the jointly optimized hardware and software stack they need to deploy AI at scale.

Independent reporting from HPCwire corroborates the timeline and scope of the changes. The coverage confirms the May 14, 2026 announcement date and the core details of the team integration and technology licensing.

For developers, this means the physical experience of deploying AI models should improve. Think about the current friction: waiting for containers to spin up, debugging memory leaks that only appear under load, watching latency spike during peak hours. Nebius is attempting to eliminate that friction by optimizing the entire stack from silicon to API endpoint (a problem that has plagued users for years, frankly).

The timing matters. Nebius is listed on Nasdaq (NASDAQ: NBIS) and headquartered in Amsterdam. The company is building a full-stack platform for developers and companies to take charge of their AI future — from data and model training to production deployment. Founded on deep in-house technological expertise and operating at scale with a rapidly expanding global footprint, Nebius serves startups and enterprises building AI products, agents and services worldwide.

This deal represents a shift in how AI infrastructure companies compete. Rather than just selling compute hours, Nebius is bundling optimized software stacks with hardware capacity. The value proposition moves from "here's a GPU" to "here's a production-ready inference pipeline." That distinction matters when you're running models that need to respond in milliseconds, not minutes.

The integration of Clarifai's system-level optimization expertise addresses a specific pain point in AI deployment. Many companies can train models but struggle to run them efficiently at scale. The difference between a model that works in a notebook and one that handles thousands of concurrent requests is often invisible until you're paying for overprovisioned infrastructure or dealing with degraded performance.

Nebius Token Factory is positioned as the central platform for this work. The name suggests a focus on token-level efficiency, which aligns with the research areas Zeiler will lead. Token efficiency directly impacts cost — fewer tokens processed means lower compute bills and faster response times. For enterprise customers running agentic systems that make hundreds of API calls per task, this optimization compounds quickly.

The deal structure is notable for what it excludes. By not acquiring Clarifai's legacy computer vision models or defense programs, Nebius is making a clean break from Clarifai's historical business. This suggests the company is focused on the inference technology itself rather than Clarifai's existing customer relationships or product lines. It's a technology acquisition, not a customer acquisition.

Whether this integration delivers measurable improvements for end users remains to be seen. The press release contains forward-looking statements subject to risks and uncertainties. Actual results may differ materially from the results predicted or implied by such statements, according to the company's SEC filings.

Integration of acquired teams and technology stacks rarely goes smoothly. There's the technical challenge of merging codebases, the cultural challenge of bringing together engineers from different organizations, and the business challenge of delivering on the promised synergies. Nebius has experience with acquisitions, having recently added Eigen AI, but each integration presents unique challenges.

The competitive landscape for AI infrastructure is intensifying. Major hyperscalers continue to expand their AI offerings, while specialized providers like Nebius attempt to differentiate through full-stack optimization. The question isn't whether Nebius can build the technology — the real question is whether customers will migrate their workloads to a new platform rather than stick with established providers.

For now, the announcement signals Nebius's commitment to becoming a comprehensive AI infrastructure provider. The combination of Eigen AI's model-level optimization and Clarifai's system-level expertise creates a theoretically complete stack. Whether that translates to competitive advantage in the marketplace depends on execution, pricing, and customer adoption.

Whether users actually pay for it remains the real question.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn