마이크로소프트, 네비우스와 190억 달러 AI 계약 체결: GPU 강자 엔비디아 크게 수혜

Microsoft (MSFT) has taken a decisive step in the artificial intelligence (AI) infrastructure race, signing a five-year agreement with Nebius, an AI-focused infrastructure provider, valued between $17.4 billion and $19.4 billion. As one of the largest deals of its kind, the contract underscores the rapid expansion of AI computing demand and positions Nvidia as a primary beneficiary.

The agreement will supply Microsoft with advanced GPU-powered computing infrastructure, with deliveries beginning in late 2025 and continuing for at least five years. It also includes an expansion option that could increase the total contract value from $17.4 billion to $19.4 billion. The deal reflects the surging demand for computing power driven by AI models and applications, which require vast quantities of high-performance GPUs. By partnering with Nebius, Microsoft secures the capacity needed to compete with Amazon and Google in the cloud and AI markets. The announcement triggered an immediate market reaction: Nebius shares surged more than 40%, marking their highest level since the company’s founding. The rally underscores investor confidence in Nebius’ role within the expanding AI ecosystem.

Source: TradingView

Nebius is a relatively new entrant in the global AI infrastructure market. Headquartered in Amsterdam, the company spun out of Yandex’s international operations and operates as a “neocloud” provider, delivering GPU-focused infrastructure specifically optimized for AI workloads. The Microsoft partnership has rapidly elevated Nebius to a leading position in the sector. Beyond guaranteeing billions in revenue, the deal significantly strengthens Nebius’ credibility with prospective clients and partners. To support this growth, Nebius plans to raise an additional $3 billion through convertible notes and equity offerings. The capital will fund data center expansion and infrastructure upgrades designed to fulfill Microsoft’s contractual requirements. This expansion is critical amid accelerating AI adoption and intensifying competition among the world’s largest technology firms for reliable GPU access. With Microsoft as an anchor client, Nebius has proven its capability to deliver infrastructure at scale.

Source: Grand View Research

While Microsoft and Nebius executed the agreement, Nvidia stands out as a primary beneficiary. Nebius’ infrastructure depends heavily on Nvidia GPUs, which remain the industry standard for training and deploying AI models. The contract translates into billions in new product demand for Nvidia, as Nebius must substantially scale its GPU procurement to meet delivery schedules. Furthermore, Nvidia already holds a stake in Nebius, aligning its financial interests directly with the company’s operational success. This development reinforces Nvidia’s dominance in the AI semiconductor market. Despite competition from AMD and Intel, Nvidia remains the preferred supplier for large-scale AI infrastructure. The Microsoft–Nebius agreement further confirms that Nvidia’s hardware will remain foundational to AI advancement for years to come.

RELATED: NVIDIA (NVDA Stock) Hits $4 Trillion, Igniting ESG Investment Momentum Across the Semiconductor Sector

A key strategic advantage of this arrangement lies in Microsoft’s capital allocation approach. Instead of financing and constructing all necessary data centers in-house, Microsoft is leveraging specialized providers like Nebius. This model enables faster infrastructure scaling while minimizing upfront capital expenditure. By outsourcing significant capital expenses to Nebius, Microsoft mitigates financial risk while securing essential GPU capacity. Nebius will assume responsibility for financing, construction, and ongoing management of the new facilities. In return, Microsoft secures guaranteed access to the resources required to scale AI services like Azure OpenAI, bypassing the multi-year timelines typically associated with in-house data center development.

RELEVANT: Microsoft Buys 3.5 Million Carbon Credits to Offset AI’s Soaring Emissions

The Microsoft–Nebius agreement signals a broader structural shift in AI infrastructure development. Rather than relying exclusively on internal resources, major technology firms are increasingly partnering with specialized infrastructure providers. This model yields several strategic advantages. First, accelerated scaling: strategic partnerships enable firms like Microsoft to rapidly expand AI capacity, circumventing the lengthy timelines associated with greenfield construction. Second, capital efficiency: outsourcing large-scale infrastructure projects conserves financial resources, allowing companies to redirect capital toward software development, application innovation, and customer-facing services. Third, market validation: multi-billion-dollar contracts establish the viability of emerging players like Nebius, providing the financial foundation required for sustained expansion.

The Microsoft–Nebius contract aligns with rapid expansion across the AI infrastructure sector. Global expenditure on AI semiconductors and cloud capacity is projected to surpass $200 billion by 2030, rising from approximately $45 billion in 2024. Some forecasts suggest the market could reach $400 billion by the end of the decade.

Source:

GPU demand is anticipated to compound at an annual rate of 25–30%, fueled by widespread generative AI adoption. Industry analysts project that “neocloud” operators like Nebius may secure up to 15% of AI infrastructure contracts by 2030. With an existing market share exceeding 80%, Nvidia is positioned to maintain its dominance as global GPU demand accelerates.

SEE MORE: Tesla’s AI5 Chip Challenges NVIDIA’s Dominance in AI Hardware Innovation

While the $19 billion Microsoft–Nebius agreement highlights rapid AI commercialization, it simultaneously intensifies scrutiny regarding environmental sustainability. Training and operating AI models require substantial computational resources, translating directly into high energy consumption. Recent research indicates that training a single large language model (LLM) can generate over 500 metric tons of CO₂ emissions—roughly equivalent to the lifetime output of several passenger vehicles. The following analysis compares the energy consumption and carbon footprint of leading LLMs currently in deployment. Data centers supporting AI workloads currently consume approximately 1.5% of global electricity, a figure projected to climb to 4% by 2030 as AI adoption accelerates. A significant portion of this load stems from GPUs, which draw substantially more power than conventional processors. Consequently, companies like Microsoft face mounting pressure to reconcile infrastructure expansion with environmental stewardship. Addressing these challenges requires three primary strategies: investing in renewable energy sourcing, improving data center operational efficiency, and exploring low-carbon infrastructure alternatives. Without intervention, unchecked AI infrastructure growth risks becoming a dominant source of carbon emissions within the technology sector. Achieving sustainable scaling will depend on major breakthroughs in energy efficiency.

INTERESTING READ: ChatGPT, Gemini, and DeepSeek Are on an AI Race – But at What Climate Cost? A Comparison

The $19 billion Microsoft–Nebius agreement represents a pivotal moment in the AI infrastructure landscape. It provides Microsoft with the scalable capacity required to maintain competitiveness in a rapidly evolving market. Simultaneously, it elevates Nebius from an emerging operator to a global cloud infrastructure leader. Finally, it reaffirms Nvidia’s position as the foundational layer of AI computing, with direct financial upside from surging GPU demand. As AI demand continues to accelerate, similar strategic partnerships will become increasingly common. While the Microsoft–Nebius transaction ranks among the largest to date, it is unlikely to be the final word. The AI infrastructure race is only beginning, and success will depend on the coordinated efforts of specialized infrastructure providers, semiconductor manufacturers, and cloud hyperscalers.

FURTHER READING: Study Shows How AI Can Cut Over 5 Billion Tons of Carbon Emissions in 3 Key Sectors