37.5 億美元 “彈藥” 到位 Nebius 大舉擴張 AI 算力版圖

According to Zhitong Finance APP, in recent years, the European technology leader Nebius Group NV (NBIS.US), which has shifted its focus to providing cloud AI computing resources, has raised approximately $3.75 billion through the sale of convertible bonds and stocks. With the company's recent announcement of a massive contract closely related to cloud AI computing with American tech giant Microsoft (MSFT.US), it is seeking to purchase more land and AI computing resources.

The AI cloud service provider significantly raised its convertible bond issuance scale from the initially planned $2 billion to $2.75 billion on Thursday. The company also stated that it sold approximately $1 billion in new shares, priced at $92.50 per share.

Nebius conducted the sale of convertible bonds in two phases, raising about $1.375 billion in each phase, with one maturing in 2030 at a coupon rate of 1% and the other maturing in 2032 at a coupon rate of 2.75%. The conversion price for these bonds is $138.75 per share, which is a 50% premium over the pricing of this stock sale, highlighting the strong demand in the financial market for the company's equity and debt assets.

The Amsterdam-based company officially spun off from Russian internet giant Yandex last year, and its collaboration with Microsoft marks a significant shift in the company's fundamental expectations. Previously focused on providing AI infrastructure for startups and smaller businesses, it is now gradually acquiring the capability to provide large-scale AI training/inference computing resources.

As of Wednesday's close, Nebius's stock price has risen 237% year-to-date to $93.39 per share, with the company's stock price surging nearly 50% in a single day as of Tuesday's close.

Single-day surge of nearly 50% for the "AI computing myth"

On September 9 (Tuesday), Nebius's U.S. stock closed up approximately 49.4%, marking its largest single-day gain since listing on NASDAQ (at one point during the day, it rose about 53%), and also setting a new historical high price.

From an event-driven and fundamental perspective, the main reason for Nebius's stock price surge is the company's massive AI computing contract with Microsoft. Microsoft chose Nebius primarily due to its rapidly deliverable cross-regional capacity, its large AI server cluster's high compatibility with NVIDIA's high-performance AI GPU roadmap, vertically integrated AI cloud stack, and Microsoft's diversified supply chain needs. The financing increase further validates the certainty of execution and expansion, and positive feedback on performance and valuation is expected to continue in the short to medium term.

This week, Microsoft reached a multi-year agreement with Nebius worth nearly $20 billion, aimed at obtaining AI cloud computing capabilities from this tech company that spun off from Russian internet giant Yandex. In documents submitted to the U.S. Securities and Exchange Commission on Monday, the company stated that this large transaction is expected to bring Nebius between $17.4 billion and $19.4 billion in revenue by 2031.

The demand for AI computing driven by generative AI applications and AI agents at the inference end is described as "the stars and the sea," and is expected to drive the AI computing infrastructure market to continue showing exponential growth. The "AI inference system" is also considered by Jensen Huang to be the largest source of revenue for NVIDIA in the future Microsoft, a major shareholder of OpenAI, has long faced a severe shortage of AI cloud computing power due to OpenAI's enormous computing demands and the integration of generative AI and AI agents into core products like the Office suite. To address this, Microsoft is supplementing its resources through third parties, such as CoreWeave, known as the "favorite child of NVIDIA," and is now introducing Nebius to diversify supply and rapidly expand computing reserves.

Both Nebius and AI cloud computing leader CoreWeave belong to the "AI dedicated cloud" sector, focusing on faster delivery, optimized training/inference stacks, and flexible contracts to serve the needs of large model developers and enterprises with substantial AI workloads. Collaborating with both companies helps Microsoft alleviate supply bottlenecks.

Nebius is equipped with H100/H200 and the current state-of-the-art AI GPU product line—GB200 based on the Blackwell architecture and GB300 based on Blackwell Ultra (liquid cooling modifications are underway)—to meet different TCO demands for both training and inference, which is essential for Microsoft's multiple product lines (Azure/Copilot/GitHub, etc.).

Nebius is not merely a "managed data center" but an integrated AI cloud leader that encompasses hardware, networking, and scheduling stacks, emphasizing flexibility from single cards to "thousands of pre-optimized AI GPU clusters," thereby reducing integration costs and delivery uncertainties for Microsoft. The ability to quickly deliver regional capacity is also a key rationale for Microsoft's collaboration with Nebius, which has established a presence in both the EU and the US. Its large-scale AI server clusters include a self-built AI data center in Finland, large regional launches in France and the US, and a significant high-performance GPU capacity soon to be provided in Vineland, New Jersey, facilitating rapid access and network latency optimization for Microsoft in the northeastern US and even Western Europe