The Best AI Play — A Stock I’d Hold Forever: $NBIS (Investment thesis) Here’s my deep dive into what I believe is the most exciting, underappreciated opportunity in the AI infrastructure space. Over the past few months, $NBIS has surged over 150%, and its presence on X has grown noticeably. Still, many people don’t fully understand what $NBIS actually does, and just how much potential it really has. I won’t waste your time. Let’s get straight into it.

1) History Nebius Group traces its origins to 1989, when it was founded by Arkady Volozh and colleagues as part of the early development of what was Yandex. In May 2011, Yandex went public on NASDAQ. In August 2024, following a full corporate restructuring and divestment from its former Russian operations, the company rebranded as Nebius Group and adopted the ticker $NBIS. Today, Nebius builds full-stack AI infrastructure, including large-scale GPU clusters, developer tools, and AI-centric cloud platforms. Its core business units include Nebius, an AI-first cloud platform; Toloka AI, a global data partner for generative AI; TripleTen, an edtech reskilling platform; and Avride, a developer of autonomous driving technologies. Headquartered in Amsterdam, with R&D centers across Europe, North America, and Israel, Nebius is at the forefront of the global AI revolution. While it emerged from the legacy of Yandex, Nebius is now a fully independent company, with no operational or ownership ties to its former Russian business.

2) What does $NBIS exactly do? (Core business, data centers) Nebius’ core business is a fully integrated cloud chain built for intensive AI workloads. But what does this entail? Nebius begins its fully integrated cloud journey by manufacturing and assembling the core hardware infrastructure needed to build its own data centers. This includes designing and deploying custom GPU servers engineered specifically for performance and energy efficiency. The result? A highly optimized, low-overhead foundation for cloud and AI operations. The company operates one of the most energy-efficient data centers in Europe, with a Power Usage Effectiveness (PUE) of 1.13, outperforming global tech giants like Oracle (1.15), Alibaba (1.20), and IBM (1.46), and well below the global average of 1.58. Lower PUE means less wasted energy, directly improving cost-efficiency and sustainability. Beyond just cost-efficiency, Nebius operates the most powerful commercially available supercomputer in Europe and the second most powerful of its kind globally. By replicating these high standards in every expansion, Nebius ensures that its infrastructure remains competitive, cost-effective, and built to support AI at scale.

2.2) Core business (Hardware) To support high-performance and cost-efficient AI infrastructure, Nebius develops its own servers in-house, from design and prototyping to production and deployment. This end-to-end control over hardware results in superior efficiency, faster innovation cycles, and a lower total cost of ownership. By engineering its hardware internally, Nebius achieves power-efficient cooling, tightly integrated performance tuning, and full visibility across the supply chain. This translates into lower energy costs, better thermal performance, and the ability to scale with predictability and confidence. The on-demand development process allows Nebius to move faster than traditional vendors. With rapid prototyping, custom production, and streamlined shipment, Nebius can bring new server designs to market quickly, matching the pace of innovation in AI workloads. Moreover, the in-house engineering team enables full-stack validation and quick implementation of new features. This agility is critical when supporting evolving AI architectures and adapting to customer needs. Finally, Nebius servers are built on an open-source infrastructure, ensuring security, transparency, and flexibility. This commitment empowers customers to audit, customize, and integrate the platform to their specific requirements, without vendor lock-in. Through this integrated, agile approach to hardware, Nebius ensures its infrastructure is always aligned with the performance, reliability, and transparency needed for cutting-edge AI operations. NOTE: $NBIS does not design and manufactureall components for its data centers in-house.

2.3) Core business (Cloud platform) Built on top of Nebius’ robust hardware and data center layer is the proprietary Nebius cloud platform. This platform is a core pillar of Nebius’ full-stack infrastructure for AI developers, ensuring control over the whole value creation chain. Unlike generic cloud services, it’s engineered for large-scale GPU clusters, making it ideal for training and deploying advanced AI models. The platform removes bottlenecks that often hinder machine learning workflows, offering developers a seamless environment to scale both training and inference with minimal latency or resource fragmentation. The $NBIS Platform is described as "more than 'bare metal'," offering a more flexible and wider product portfolio than generic GPU Cloud solutions. It provides an AI Platform and Applications (PaaS) layer for managing and deploying AI workloads and a foundational AI Cloud (IaaS) layer providing managed infrastructure and scalable resources, including GPU clusters and storage. The platform's overall design improves user experience through a friendly interface, API access, comprehensive documentation, and integrated monitoring. Additionally, Nebius Studio, an API SaaS for open-source models, simplifies application creation. It offers benchmark-backed performance and cost efficiency, providing significant savings and faster processing. Key benefits include ultra-low latency, verified model quality, and expert support.

2.4) Core business (AI expertise and experienced team) Nebius’ success is underpinned by its dedicated team of over 850 professionals, all focused on software, hardware, and data center operations, supported by robust back-up departments. The team boasts exceptional key strengths, including full-stack expertise in AI cloud development and operations, covering everything from server design to running cutting-edge LLMs. This deep expertise is evident in the in-house LLM team, which plays a crucial role in "dogfooding" (using Nebius' own technology) to facilitate adjusting hardware and platform to the real needs of ML/AI practitioners. This strategic alignment of "AI expertise" and an "experienced team" ensures $NBIS has the right number of people with the necessary skills and knowledge in place to support current scaling and future growth, covering essential areas like software development, hardware R&D, data center operations, and product development.

2.5) MOAT So, what differs Nebius’ core business from competitors? 1. Sole Focus on AI Infrastructure and Tailored Solutions: Unlike general cloud providers, $NBIS maintains a precise and sole focus on AI infrastructure. This allows them to provide solutions specifically tailored to the unique needs of AI developers and businesses, including small and medium-sized customers, serving a broader set of customer demands. 2. Full Control Over the AI Infrastructure Value Chain: Nebius has full control over its entire AI infrastructure value chain, from in-house server design to its proprietary cloud platform. This end-to-end control ensures maximum productivity, quick scaling, higher cost efficiency, and low downtime. 3. Strategic Partnerships and Cutting-Edge Hardware: Nebius benefits from long-standing collaborations with $NVDA across hardware and cloud, positioning them as one of the first providers in Europe to offer state-of-the-art, energy-efficient NVIDIA Blackwell platforms to customers by 2025. 4. Strong, Experienced, and Specialized Team: Nebius boasts a leading team of over 850 professionals, including approximately 400 experienced AI/ML/software and cloud engineers. This team has a proven track record of building cloud services and infrastructure from scratch, possesses full-stack expertise in AI cloud development and operations, and includes an in-house LLM team that ensures hardware and platform are adjusted to the real needs of AI practitioners. The team is also strategically sized to support future growth. 5. Differentiated Offering Compared to Competitor Groups: • Versus Bare Metal: Nebius provides more than just bare metal, offering additional resources and services specifically for AI workloads. • Versus GPU Clouds: Nebius offers greater flexibility in GPU consumption and a broader service offering, backed by a larger development team. • Versus Hyperscalers: Nebius provides a more tailored approach specifically for AI developers and businesses, unlike the broader, less specialized offerings from larger hyperscalers.

2.6) Value proposition for clients • User-Centric Approach: $NBIS makes it easy for developers and teams to manage and automate their infrastructure using a variety of modern tools and interfaces, tailored for flexibility and efficiency. Clients benefit from detailed documentation, built-in monitoring tools, and round-the-clock (24/7) technical support, ensuring ease of use and operational reliability. • Resilient Infrastructure: The core infrastructure layers are meticulously designed by the Nebius team with a strong emphasis on speed and resilience. Flexible Capacity Allocation and Pricing: Clients have the flexibility to choose between on-demand consumption and various term reserve consumption modes, allowing for optimized cost management and resource allocation tailored to their specific needs. Fast Onboarding: Nebius ensures rapid deployment with an optimized software stack. A dedicated team of engineers is available to assist clients in quickly setting up infrastructure and implementing their AI solutions, including training, inference, RAG, and other applications. State-of-the-Art Hardware with Immediate Access: Clients benefit from Nebius' strong partnership with NVIDIA, which often grants immediate access to the newest GPUs. This ensures clients gain access to the most powerful and cutting-edge GPU clusters, providing unparalleled computational power for intensive AI workloads.

3) Toloka Besides its core business, $NBIS has 3 subsidiaries, of which one being Toloka. Toloka, however, is set to be deconsolidated from Q2 2025. While Nebius Group will relinquish voting control, it will retain a significant majority economic stake. Toloka AI is a leading provider of high-quality human data operations, crucial for training, fine-tuning, alignment, and evaluation of advanced Generative AI models. In 2023, the company strategically pivoted to focus on providing this essential data for GenAI. Key Competitive Advantages: • Technology & Quality at Scale: Toloka delivers high-quality human and synthetic data, leveraging proprietary technology for automated quality control and expert selection. Its resilient infrastructure ensures reliability and compliance. This consistent high quality and scalability create a significant technical moat. • Diverse Expertise: Toloka draws on a multi-layered approach, including Domain Experts, AI Tutors, and a Global Crowdsourcing network spanning 50+ knowledge areas and 40+ languages, reaching up to 250k+ monthly active users. • Trusted & Recognized: Toloka is a trusted partner for leading frontier AI labs, model developers, and enterprises (e.g., Anthropic, Shopify, Microsoft, Amazon), consistently receiving positive client feedback and industry recognition from analysts like Gartner and Forbes. • Investor Validation & Funding: Toloka's potential is validated by high-profile investors. In May 2025, Bezos Expeditions notably led a new investment round, supported by Mikhail Parakhin (CTO of Shopify), providing significant funding for business scale-up. Market Opportunity & Growth: The expert data market for GenAI is experiencing tremendous growth, projected to expand from $2 billion in 2024 to over $17 billion by 2030 (+40% CAGR). Toloka is strategically positioned to capitalize on this surging demand across various GenAI data user profiles.

4) Tripleten $NBIS second subsidiary that I’ll be discussing is Tripleten, an edtech platform focused on reskilling individuals for careers in technology. It offers high-quality, immersive, and affordable training, combining bootcamp and MOOC formats. The platform features proprietary technology with AI integration, including an AI Tutor and AI-enhanced learning tools. Tripleten delivers strong outcomes, evidenced by an 87% employment rate within six months of graduation, solid median salaries, and positive student feedback. It has been recognized by Fortune as the "Best overall provider" of software engineering bootcamps. Financially, Tripleten doubled its year-over-year revenue in Q1 2025 (+144%) and achieved strong ARR growth, reaching $27.2 million in Q3 2024. It is also expanding into the B2B segment. The global digital education market, its primary opportunity, is projected to grow from $19 billion in 2024 to over $95 billion by 2030. Tripleten's Key Investment Highlights: • It is a top-rated US bootcamp characterized by high employment rates, competitive salaries, and positive student feedback. • The automated, flexible e-learning platform simplifies launching and scaling in new markets, enabling affordable pricing for customers and cost savings for the business. • Tripleten is experiencing a growing presence in both B2B and B2C sectors, opening massive growth opportunities across its key focus markets: the US and Latin America.

5) Avride The third and final NBIS subsidiary I’ll be covering is Avride. It specializes in developing and operating autonomous vehicles and delivery robots, leveraging shared and mutually enhancing technologies. Core Business & Technology: Avride is uniquely positioned by simultaneously developing autonomous driving technology for both cars and delivery robots. Both applications utilize common technological elements, including models training systems, code deployment and data collection, simulators, localization, and perception. Advances in autonomous vehicles drive the development of delivery robots, while delivery robots’ faster deployment and monetization support future car development. Autonomous Vehicles: Avride's autonomous vehicles are designed for use cases such as ride-hailing, logistics, e-commerce, and food/grocery delivery. The company has extensive experience, with over 47,000 passenger rides given and more than 22 million autonomous kilometers driven (as of Sept 2024). In March, Avride announced a partnership with Hyundai Motors to co-develop a full autonomous vehicle platform. Hyundai will handle vehicle manufacturing and assembly, while Avride focuses on software development and systems integration. This partnership provides a pathway to scale Avride's AV fleet, starting with 100 Ioniq vehicles in 2025. Avride and Hyundai plan to begin testing a robotaxi service in partnership with Uber by the end of this year in Dallas. Delivery Robots: Avride's delivery robots are deployed for restaurant deliveries, grocery deliveries, and small-scale logistics. They have delivered approximately 200,000 client orders. The robots have been successfully tested in diverse conditions, including city streets with heavy vehicle and foot traffic, operating at a max speed of 8 km/h and a range of 55 km. They feature in-house design and a factory partner in Taiwan for scalable production. In Q1, Avride scaled its food delivery operations across US university campuses and urban centers. As of January, over 100 robots are completing more than 1,200 deliveries per day at Ohio State University in partnership with Grubhub. Deliveries are also scaling in Jersey City (Uber Eats), and expanding in Austin and Dallas. International interest is growing, with Avride deploying an initial fleet in Tokyo through a partnership with Rakuten. Key Investment Highlights & Team: • Experience: Avride boasts one of the most experienced teams in the autonomous vehicles industry, comprising over 200 self-driving tech engineers and software developers with 7+ years of experience. • Cheap: The company offers a fast track to commercial deployment at a fraction of the costs. Extremely safe: Its technology has been tested in diverse road conditions with an outstanding safety record, backed by key partnerships with leading automakers and delivery platforms. • Market Opportunity: Avride targets scalable business models focused on large end markets. The TAM for robo taxis is projected at $46 billion by 2030, last mile delivery robots at $6 billion by 2030, and inter-city delivery at approximately $14 billion by 2030.

6) Financials (Revenue) Let’s talk revenue. $NBIS is still early in its journey, but it’s already putting up stunning growth. Last quarter, the Nebius group posted $55M in revenue, a massive 385% YoY increase. Looking ahead, they expect to hit $600M in FY2025, and they’re not slowing down. Management is targeting $3B in revenue in the medium term. That implies $NBIS expects to 5x its FY2025 revenue within just a few years, those are not rookie numbers haha

7) Financial (EBITDA + EBIT) At the moment, $NBIS is EBITDA negative, but it anticipates turning EBITDA positive during the second half of 2025, a key milestone for a company at this stage. Even more encouraging, Nebius is forecasting a healthy 20–30% EBIT margin in a few years, which would be an excellent result and well above typical expectations for a company in that phase.

8) Balance Sheet As of last quarter, $NBIS had $1.44B in cash on hand. Just over a month ago, the company also raised an additional $1B via convertible notes, further strengthening its position. With virtually no debt, $NBIS has the flexibility to aggressively invest in growth without balance sheet constraints.

9) Negative points and risks • Capex Intensive: To meet demand, $NBIS will need to invest heavily in new hardware (especially GPUs), data center expansion, and continued development of its cloud platform, developer tools, and AI studio. This could lead to further capital raises, potential dilution, or increased debt, depending on funding needs. • Execution Risk: Rapid scaling across multiple business units (cloud, data, edtech, and autonomous driving) increases operational complexity. Delivering on ambitious revenue and margin targets will require flawless execution. • Competitive Pressure: $NBIS is entering a highly competitive market dominated by hyperscalers (AWS, Azure, Google Cloud) and specialized AI cloud players. Differentiation and sustained pricing power could be a challenge.

10) “Extra” positive points • Founder-led: $NBIS is led by CEO Arkady Volozh, a serial entrepreneur with a remarkable track record. He has (co-)founded three major companies and served on the board of a startup that was later acquired by Facebook. 1. CompTek (1989) – Once the largest telecom company in Russia, focused on networking and internet technologies. 2. Yandex (2000) – Often called the “Google of Russia,” it reached a peak valuation of$30B and became one of Europe’s top tech success stories. 3. Getir (2015) – A European pioneering ultrafast delivery company that became a leader in the on-demand delivery space. 4. Face .com – Volozh was a board member of this facial recognition startup from 2010 until its acquisition by Facebook in 2012. Volozh isn’t just a founder, he’s a visionary operator with a history of building category-defining companies across different verticals. • ClickHouse: $NBIS holds a 28% stake in ClickHouse, a company specializing in online analytical processing (OLAP), ideal for businesses requiring fast, large-scale data insights. Just a month ago, ClickHouse raised $350M at a $6.35B valuation, valuing $NBIS’s stake at approximately $1.78B. If $NBIS ever needs to raise cash, it could sell part of this stake for a significant amount. That said, even if it never sells, I believe ClickHouse is simply a great long-term investment. • Partnership with $NVDA: As a result of the partnership with $NVDA, $NBIS will be one of the first AI clouds to offer the new NVIDIA Blackwell Ultra AI factory platform. Additionally, by EOY, $NBIS’ customers will have access to GB300 NVL72-powered instances, featuring 72 NVIDIA Blackwell Ultra GPUs for unmatched AI acceleration.

11) Valuation + long-term target At the moment, $NBIS trades at a FY2025 P/S ratio of 20. While that’s not exactly cheap, it’s also not steep when you consider how rapidly $NBIS is compounding revenue and expanding its footprint. Personally, I see fair value a bit above $60. Looking ahead, Goldman Sachs projects $2.8B in revenue for FY2027. Applying a conservative10x P/S multiple, that would imply a market cap of $28B, representing approximately 130% upside from today’s levels over the next 2.5 years.

12) Conclusion $NBIS isn’t just another cloud company, it’s much more. Nebius offers a comprehensive, full-stack AI platform, combining cutting-edge hardware, energy-efficient data centers, and purpose-built software that eliminates friction across the AI development lifecycle. This isn’t just infrastructure; it’s an integrated ecosystem tailored for machine learning and generative AI at scale. From custom GPU servers to an in-house cloud platform and developer-facing AI tools, Nebius provides everything an AI company needs to train, deploy, and iterate; fast, efficiently, and at scale. It’s this deep vertical integration that sets Nebius apart from traditional clouds and makes it a true enabler of the AI economy. With full-stack control, strategic NVIDIA partnerships, and a founder with a proven track record of building billion-dollar companies, $NBIS is positioned to be one of the most important infrastructure companies of the AI era. Add in fast-scaling subsidiaries like Toloka, TripleTen, and Avride, and you get exposure to multiple massive secular trends; from AI compute to data labeling, education, and autonomous mobility. Yes, there are risks, high capex, execution challenges, and strong competition, but the upside potential here is enormous. This isn’t a short-term trade. It’s a long-term conviction play. I believe $NBIS has the potential to become a distinguished cloud company in the future. Thanks for reading!