Morgan Stanley views AI cloud newcomer Nebius (NBIS.US) as "neutral": Customer demand far exceeds supply, and a shortage of computing power is hindering growth.

Morgan Stanley recently released a research report on Nebius Group (NBIS.US) AI cloud business, combining the content of the company's 2026 industry summit to comprehensively analyze its business transformation, customer layout, product strategy, financial outlook, valuation, and potential risks, giving the stock a "neutral" rating, and judging the industry as attractive overall, with a target price of $144. The report stated that the 2026 Nebius industry summit further confirmed the company's overall AI cloud development strategy: in addition to large-scale cloud service providers, various enterprises and AI startups have generated real purchasing demand; the company has built a complete software technology stack, with the ability to increase revenue per megawatt of computing power and improve profit margins, while also being able to reasonably allocate computing resources to meet the massive market demand. The report stated that the industry summit clearly demonstrates the business forms and application scenarios of enterprise customers and AI startups, and fully presents the company's complete AI full technology stack capabilities. The summit intuitively reflects the real market demand, and also confirms the broad market opportunities for traditional enterprises and native AI enterprises. The company's software-centric AI cloud strategy path is becoming increasingly clear. Previously, Nebius's development focus was on signing large cooperation orders to complete financing and business expansion. In the next stage, the company will focus on delivering computing resources to more enterprise customers, driving up revenue per megawatt of computing power and profit margins, and building long-term differentiated competitive advantages based on its self-developed AI cloud software stack. This summit also provides stronger evidence: Nebius's customer base is no longer limited to large-scale cloud service providers, and the company is gradually transitioning to high-value AI cloud service providers, moving away from a position solely as a raw computing power provider. Dispelling market doubts This summit is enough to dispel the views of market bears and skeptics: Nebius is not simply a GPU computing power intermediary. The company accurately positions itself between traditional large cloud providers and emerging cloud service providers: compared to traditional cloud providers, the product is native to AI scenarios and has higher response efficiency; compared to small GPU specialty service providers, the service has stronger stability and full-stack technical capabilities. The summit sends a clear signal: for enterprises implementing AI at the production level, relying solely on raw chips is far from enough, and stable operation, technical support, resource scheduling, inference performance optimization, cost control, and a series of supporting capabilities are also needed. Customer cases are an important positive The disclosed cooperation customers in this summit are a major highlight. Nebius's current customer base has expanded significantly, covering AI labs, AI product companies, traditional large enterprises, and digital native enterprises. Typical cooperation cases include: Recraft: with the help of Nebius services, model training speed has increased by 6 times; Brave: relies on the platform to process over 16 million AI content summaries daily; Shopify: uses its computing power for model pre-training and fine-tuning work; Mastercard: uses Nebius's intelligent search function; Higgsfield: lists Nebius as its preferred infrastructure partner. The rich customer cases also address the most concerning issue for investors: as the business expands, Nebius has the ability to sell computing power to a diversified customer base, rather than relying solely on cooperation orders with large cloud service providers. Optimization of customer structure, benefits profit margin improvement The management reiterated that the large orders with large-scale cloud service providers are mainly used to support the overall AI cloud infrastructure construction; and the company prefers to directly cooperate with end customers. Compared to the AI cloud business directly facing enterprises and native AI enterprises, the profit margin is significantly higher than that of pure GPU computing power large orders; if added with value-added functions such as software services, inference computing, intelligent routing, intelligent agents, the profit space will further expand. Strong market demand, computing power supply becomes a major bottleneck Many customers have expressed that market demand is very strong, but the lack of computing power resources is the biggest constraint. Shopify: GPU computing power demand has reached a turning point, traditional large cloud service providers either have insufficient computing power reserves or have priced too high, so the company started seeking alternative solutions; Rhoda: highly recognizes the platform's stability, low latency, technical support, and resource availability, but still struggles to obtain sufficient computing power; Higgsfield: acknowledges the platform's outstanding inference performance, but its own business demand growth rate has far exceeded the existing computing power supply. The feedback from multiple customers all points to the same fact: market demand is much greater than computing power supply. Product roadmap focuses on production-level AI and intelligent agents Nebius's product planning is increasingly focused on AI businesses and intelligent agent applications that are ready for production. The company emphasizes that supporting stable operation of AI in a production environment requires a full set of capabilities: high-performance computing power, storage, resource scheduling, serverless AI, inference computing, intelligent routing, real-time scenario adaptation, and intelligent agent tools. The current industry situation is that most companies can build basic intelligent agents, but stabilizing and cost-effectively maintaining these applications is extremely challenging. This also confirms the core logic of Nebius's software-driven development. Once the company embeds more platform functions into computing infrastructure, it will be able to achieve higher value commercialization, while further enhancing customer stickiness. Customer viewpoint summary Customers choose Nebius mainly for its operational performance, service stability, delivery speed, professional technical support, low latency, and flexible and efficient service response capabilities. Shopify: can achieve more efficient technical integration with Nebius compared to traditional large cloud providers; Rhoda: the platform architecture is streamlined while also having complete resource scheduling, storage, and hosting database capabilities; Higgsfield: considers Nebius as the top partner in terms of inference performance. The existing pain points in the industry are the scarcity of computing power resources, difficulty in pricing and cost control, complex model routing logic, and difficulty in operating AI systems in a production environment. And these pain points align with Nebius's current focus on developing inference computing, resource scheduling, intelligent routing, and intelligent agent core product directions. Product and platform dynamics The company has now built a complete AI cloud technology stack, including NVIDIA Corporation GPU servers, network equipment, AI-specific storage, accelerated computing, resource orchestration, serverless AI, enterprise operations platform (MLOps), inference services, and intelligent agent tools represented by Nebius Echo. The core of this product update is not a single new product, but a strategic shift in its entirety; Nebius is extending from a basic infrastructure service provider to a full-service AI production workflow service provider. Financially, Morgan Stanley predicts that the company's revenue will experience explosive growth, with revenue reaching $5.3 billion in 2025 and an estimated $17.929 billion in 2028; profitability continues to improve, operating profit margins and adjusted EBITDA profit margins are increasing year by year, losses are narrowing and gradually turning into profits, and the main revenue comes from North America. In terms of valuation and investment rating, the report gives the stock a "neutral" rating, judging the industry as attractive overall. The company's closing price on June 9th was $211.69, with a target price of $144 from Morgan Stanley. The report divides into three scenarios: in an optimistic scenario, the target price is $400, corresponding to large-scale deployment of computing power and exceeding expectations in software monetization; in a base scenario, the target price is $130 to $300, performance steadily improving with the delivery of computing power and customer expansion; in a pessimistic scenario, the target price is $70, dragged down by declining demand, increasing competition, and deployment falling short of expectations. 65% of market institutions give a buy rating, and 35% give a neutral rating.