Nebius

Nebius Group, publicly traded on the NASDAQ under the ticker symbol NBIS, represents the latest evolution of a technology company originally founded in 1997 by Arkady Volozh and Ilya Segalovich. In its current iteration, Nebius Group was officially formed in July 2024 following a massive corporate restructuring. The company finalized the divestment of its Russia-based operations (including its search, ride-hailing, and e-commerce platforms) to a consortium of local investors—including the energy firm Lukoil and members of local management—in a transaction valued at approximately $5.4 billion.

The company maintains its formal corporate headquarters in Amsterdam, Netherlands. It operates major research and development and commercial hubs across Europe and the Middle East, with its largest concentrations of operational footprint in Tel Aviv, Paris, and London. The physical GPU clusters that power its cloud infrastructure are strategically located in datacenters in Mäntsälä, Finland, and Paris, France.

Nebius is a publicly traded entity on the NASDAQ exchange. Trading of the company’s shares had been suspended in February 2022 due to geopolitical fallout surrounding the Russia-Ukraine war. Following the successful completion of the divestment and a comprehensive rebrand, the NASDAQ approved the resumption of trading on October 21, 2024, under the NBIS ticker. Upon its return to the market, the company's valuation hovered between $3 billion and $4 billion. Crucially, the firm emerged from the restructuring entirely debt-free, carrying a cash reserve of roughly $2.5 billion, primarily settled in Chinese Yuan and Class A shares during the divestment.

As of late 2024, the company employs approximately 3,000 individuals worldwide across four distinct business units. Roughly 400 of these employees are located in Israel, with the remainder distributed throughout Europe and North America. The workforce is heavily weighted toward systems engineering, hardware infrastructure, and machine learning research.

The single-sentence product description for its core offering is: an AI-native cloud computing platform that provides bare-metal and managed GPU-as-a-service infrastructure specifically engineered for training and deploying large language models.

The most consequential event in the company's past 12 months was the July 2024 completion of the $5.4B divestment of its Russian assets. This restructuring paved the way for founder Arkady Volozh to officially return as CEO, following his successful legal appeal that resulted in his removal from the EU sanctions list in March 2024.

Nebius Group is not a subsidiary; it operates as an independent holding company overseeing four distinct businesses: Nebius AI (cloud compute infrastructure), Toloka AI (data generation and labeling), Avride (autonomous driving and delivery robots), and TripleTen (technical education bootcamps).

The primary product line driving the company's commercial strategy is Nebius AI. This platform provides direct access to massively parallel clusters of NVIDIA graphical processing units, offered as an infrastructure-as-a-service model to developers requiring extreme compute density.

The specific domain problem Nebius solves is the acute global scarcity of available, high-performance GPU compute required to train generative AI models. Building and orchestrating clusters of thousands of GPUs requires exceptionally rare expertise in thermal management, InfiniBand networking, and power provisioning—complexities that Nebius abstracts away for its customers by offering fully managed cluster environments.

The buyer profile is highly specific. Nebius sells to foundational model builders, generative AI startups, academic research laboratories, and enterprise ML engineering teams. The platform does not target traditional web-hosting or general-purpose cloud buyers; its entire go-to-market strategy is focused exclusively on AI developers requiring massive parallel compute power.

The sales motion incorporates both direct enterprise sales and a self-serve platform. For massive cluster reservations, the sales-led team negotiates long-term capacity agreements directly with AI labs. Conversely, individual ML researchers can use a self-serve portal to spin up single GPU instances or small virtual machines for localized inference testing and fine-tuning.

The pricing model is firmly rooted in a per-GPU-hour metric. The company offers on-demand pricing for short-term workloads, but heavily incentivizes reserved instances. Customers willing to commit to 12, 24, or 36-month contracts for NVIDIA H100 clusters receive substantial discounts, securing predictable compute pipelines while providing Nebius with guaranteed, recurring cash flows.

The company's most tangible technical moat resides in its physical infrastructure, specifically its datacenter in Mäntsälä, Finland. Built in 2014, this facility operates at a highly efficient Power Usage Effectiveness (PUE) rating of 1.09. It utilizes free cooling from the Nordic climate and captures server heat exhaust, which is then sold back to the local municipality to heat residential homes. This dramatically lowers operational energy costs compared to competing datacenters in warmer climates.

On a day-to-day basis, the engineering organization focuses heavily on hardware orchestration. The team builds and maintains managed Kubernetes environments optimized for GPU workloads, implements Slurm workload managers for batch training jobs, and optimizes NVIDIA Quantum-2 InfiniBand networking hardware running at 400Gbps to minimize latency between server nodes during massive training runs.

Nebius Group acts as an umbrella for four distinct named products and services. The first is Nebius AI, the core cloud infrastructure platform. This division provides scalable access to NVIDIA H100 Tensor Core GPUs, interconnected via non-blocking InfiniBand fabrics, and utilizes VAST Data storage architectures designed specifically for the extreme read-throughput requirements of AI training datasets.

The second product is Toloka AI. This platform focuses on data generation and evaluation, employing a vast global network of crowdsourced workers to produce high-quality datasets. Toloka specializes in Reinforcement Learning from Human Feedback (RLHF), providing the essential human-in-the-loop annotations necessary to align foundational models and minimize AI hallucinations.

The third entity is Avride, the autonomous vehicle division. Avride develops self-driving algorithms and operates fleets of autonomous delivery robots and robotaxis. The hardware and software stack rely heavily on computer vision and machine learning applied to real-time Lidar and camera data. The robots are currently being tested in controlled environments, including partnerships for autonomous food delivery in select global cities.

The fourth division is TripleTen, an educational technology platform. TripleTen operates intensive coding bootcamps focused on software engineering, data science, and quality assurance. Aimed primarily at career switchers in the United States and Latin America, the platform offers six-to-ten-month remote curricula designed to bridge the tech talent gap.

Among these offerings, Nebius AI is the undeniable flagship product and the primary destination for the company's capital expenditure. The remaining three units operate as adjacent legacy spin-offs that were retained during the 2024 divestment and now function semi-autonomously.

The most notable recent product expansion occurred in late 2024 when Nebius AI deployed fresh clusters of NVIDIA H200 GPUs in its newly established Paris datacenter. This launch marked a significant milestone in expanding its hardware footprint beyond the original Finnish facility, laying the groundwork for upcoming Blackwell architecture integrations.

Regarding sunset products, the company completely divested all of its Russian-focused services in July 2024. Localized search engines, ride-hailing applications, streaming services, and e-commerce platforms were permanently removed from the Nebius portfolio. These assets now operate under completely separate ownership with no technical or financial ties to Nebius Group.

The Nebius AI platform features deep integrations with the most prominent open-source machine learning frameworks. Engineers can deploy workloads utilizing PyTorch, Ray, and Hugging Face directly on the infrastructure, minimizing the need for custom configuration. Additionally, the Mäntsälä datacenter holds ISO 27001 (Information Security) and ISO 14001 (Environmental Management) certifications, providing necessary compliance benchmarks for European enterprise clients subject to strict data governance regulations.

Nebius operates in a hyper-competitive subset of the cloud market. Its most direct named competitors are specialized GPU cloud providers such as CoreWeave, Lambda Labs, and Together AI. While CoreWeave has aggressively expanded its footprint in the United States through massive debt financing backed by its hardware inventory, Nebius distinguishes itself by utilizing its deep cash reserves to target the European market without the burden of heavy debt servicing.

Within the broader competitive landscape, Nebius positions itself as an emerging challenger in the "AI Sovereign Cloud" space. Unlike hyperscalers such as AWS, Microsoft Azure, and Google Cloud—which offer hundreds of generic IT services—Nebius is attempting to win market share by exclusively optimizing its network topology and software stack for AI workloads, theoretically offering better performance-per-dollar for specific machine learning tasks.

The company's pricing positioning is highly competitive. Because Nebius operates with zero debt and a multi-billion dollar cash reserve, it possesses the financial flexibility to undercut competitors on long-term hardware reservation contracts, aiming to aggressively capture market share among well-funded AI startups.

In terms of notable contracts, the company has announced strategic compute partnerships with several prominent European AI startups, most notably participating in the infrastructure ecosystem for France's Mistral AI, which utilizes localized European datacenters for model training.

The trajectory of the company points toward massive infrastructure expansion. Following the corporate restructuring, management announced a dedicated $1.5 billion capital expenditure plan to be deployed over the next two years. This funding is explicitly earmarked for building out new datacenter capacity in Europe and aggressively procuring the next generation of NVIDIA hardware.

The primary sector tailwind propelling the company is the unprecedented surge in generative AI development. The demand for H100, H200, and upcoming Blackwell GPUs far outstrips global supply. Any provider capable of securing hardware allocations from NVIDIA and providing the requisite power and cooling infrastructure finds itself in a highly advantageous seller's market.

Since the completion of the corporate restructuring in July 2024, Nebius has not executed any notable acquisitions. The firm's current strategy is strictly focused on organic growth, directing its capital toward hardware procurement and datacenter construction rather than purchasing external intellectual property.

Nebius Group maintains a highly strategic physical footprint in Israel, centralized in Tel Aviv. The company's offices are located in the Alon Towers on Yigal Alon Street, adjacent to the Sarona business district. This location serves as one of the most critical engineering and leadership hubs outside of the Amsterdam headquarters.

The Israeli workforce consists of approximately 400 full-time employees. This team does not handle general IT or localized sales; rather, it is responsible for the deepest technical challenges across the organization. The Tel Aviv office houses core R&D teams for Nebius AI, machine learning researchers for Toloka, and the autonomous driving algorithm engineers for Avride.

Over the past 24 months, the Israel office underwent a significant organizational transition. Previously operating under the Yandex Israel umbrella—which included local services like Yango and Yango Deli—the office was restructured throughout 2023 and 2024. The complex divestment process ensured that the core machine learning and systems engineering talent was retained within the newly formed Nebius entity, while localized commercial operations were spun off or separated.

The connection between Israel and the company's executive leadership is exceptionally strong. Founder and CEO Arkady Volozh holds Israeli citizenship and has resided in Tel Aviv continuously since 2014. Having the chief executive based in the country elevates the Tel Aviv office from a standard satellite R&D center to a primary decision-making hub for the global organization.

The hiring profile in Israel is heavily skewed toward deep-tech engineering. The company routinely recruits backend developers proficient in C++ and Go, network engineers with specific expertise in InfiniBand and RDMA protocols, low-level Linux kernel developers, and advanced machine learning researchers. There is minimal hiring for traditional SaaS marketing or sales roles in this specific geography.

A vital strategic element of the Israel office is its proximity to NVIDIA's massive R&D presence in the country (formerly Mellanox). Nebius's network engineering teams in Tel Aviv work in close physical and professional proximity with NVIDIA Israel's engineers, facilitating rapid optimization and troubleshooting of the Quantum-2 InfiniBand networking hardware that serves as the backbone of Nebius's cloud datacenters.

The organizational culture in the Tel Aviv office is deeply intertwined with the Russian-speaking expat community. The company has successfully integrated hundreds of immigrant engineers with formidable backgrounds in mathematics and physics, resulting in a highly academic and technically rigorous work environment that relies on demanding algorithmic assessments during the interview process.