Nebius (NBIS.US) delivered explosive growth in the second quarter of 2026, with its AI Cloud revenue surging 514% year-over-year, a key metric that has prompted a fresh look from Wall Street. Despite the impressive results, Morgan Stanley (MS.US) did not raise its price target after the earnings release, maintaining an Equal-weight rating and a $144 target. This seemingly cautious stance reflects deeper concerns about the company's ability to deliver on its infrastructure commitments.

From a financial perspective, Nebius's growth is almost entirely driven by its AI Cloud business. Total revenue for the quarter reached $582.3 million, up 454% year-over-year, with AI Cloud revenue accounting for approximately $575 million, or about 99% of total revenue. A more critical metric is annualized recurring revenue (ARR), which climbed to $3 billion by the end of the quarter, a roughly 56% increase from $1.92 billion at the end of the first quarter. The quality of orders also improved significantly, as Nebius signed four landmark deals in the second quarter, with contract durations ranging from 1 to 3 years and an average transaction size exceeding $1 billion. The annualized contract value of new signings exceeded $20 million per MW, while some short-term contracts in the third quarter reached over $40 million per MW. The total value of new contracts was about four times that of the previous quarter, and the total value of new customer contracts increased more than ninefold. Capital recovery efficiency has improved substantially, with the expected payback period shortening to about 1 year and 10 months, much faster than the previous 2 to 3 years. These changes indicate that while most new contracts will not go live until later in 2026, limiting their contribution to this year's revenue, the high unit prices and shorter cycles effectively ease the pressure on upfront capital expenditures, laying a solid foundation for future growth.

The adjustment of capacity expansion targets is another major focus. Nebius raised its target for contracted power capacity by the end of 2026 from over 4 GW to 5 GW and plans to begin deploying more than 1 GW of active power annually starting in 2027. This ambitious plan implies active power will surge from 170 MW in 2025 to 780 MW in 2026, 1,680 MW in 2027, and 2,620 MW in 2028. Based on this, Morgan Stanley raised its long-term financial forecasts, projecting FY27 revenue of $10.926 billion and FY28 revenue of $19.059 billion, with adjusted EBITDA estimates increased to $6.294 billion and $12.108 billion, respectively. However, contracted power capacity does not equate to immediate computing power. The process from securing power to generating revenue involves multiple complex steps, including site approvals, power grid connections, GPU procurement, and construction. Any delays in these steps could impact the final revenue realization.

High capital expenditures are a necessary cost to support this expansion. According to data compiled by Woofun AI, Morgan Stanley expects Nebius's FY27 capital expenditures to reach $36.98 billion and FY28 capital expenditures to reach $41.066 billion, up 19.0% and 13.5% from previous forecasts, respectively. Second-quarter data already reflects this capital intensity: adjusted EBITDA was $236.2 million with a 40.6% margin, operating cash flow was $2.2461 billion, but cash spent on purchasing property, equipment, and intangible assets was $5.657 billion. Despite robust operating cash flow, construction spending far exceeded operating earnings. Notably, pricing for Blackwell computing power showed a premium, with the first capacity auction price 15% higher than the previous highest price, and short-term capacity transactions reaching annualized contract values of over $40 million per MW. This reflects the current tight supply-demand dynamics in the AI computing market, where customers are willing to pay a higher price for immediately available computing power. However, the sustainability of this premium remains uncertain, and high capital expenditures pose a significant challenge to financing capabilities.

Potential risks are primarily concentrated in customer concentration and competitive landscape. While large contracts improve revenue visibility, they also make the company more dependent on the deployment timelines and renewal decisions of a few clients. Competitors like CoreWeave, Microsoft (MSFT.US), Amazon (AMZN.US), and Google (GOOGL.US) are all accelerating their AI infrastructure investments, with Microsoft, Amazon, and Google holding significant advantages in funding, customer ecosystems, and GPU procurement. Taking the planned 300 MW Vineland, New Jersey project as an example, although management is optimistic about approval progress, the project still requires public hearings, local approvals, power grid connections, and construction. This capacity cannot be considered delivered until it is fully operational. Additionally, Nebius needs to supplement its capital through asset-backed financing, corporate debt, or equity-linked instruments. The cost of financing will directly impact shareholder returns and dilution pressure. If GPU supply becomes more abundant, short-term price premiums could be compressed, affecting profitability.

The rationale behind Morgan Stanley's decision to maintain its Equal-weight rating and $144 price target lies in the view that the market has already priced in strong demand expectations, while the greater uncertainty stems from capacity delivery capabilities. Its scenario valuations vary widely: a bull case target of $400, a base case valuation range of $130 to $300, and a bear case target of $70. In the bull case, Nebius would need to bring online over 5 GW of power capacity by 2030, expand its customer base, and improve profit margins through software services. In the bear case, balanced GPU supply and demand, increased competition, and delivery delays would lead to lower premiums. The key variables determining the final outcome are financing costs, the pace of approvals and construction, and the competitive landscape. Nebius's second-quarter results demonstrate its ability to secure large contracts at high prices, but achieving the 5 GW target depends not only on order acquisition but also on timely construction and conversion into cash flow, which is the core risk not fully priced into the current valuation.