On July 24, 2026 (Beijing Time), Nebius Group (NASDAQ: NBIS) closed at $187.77, dropping 15.02% in a single day and hitting an intraday low of $186.41. Trading volume reached 23.99 million shares, about 43% higher than the daily average of 16.62 million shares.
This sharp decline came despite a series of positive developments: On July 17, Nebius announced its first secured credit financing of approximately $775 million; on July 21, Nvidia disclosed a 9.3% beneficial ownership stake in Nebius; and on July 22, Baird initiated coverage of Nebius with an "Outperform" rating. However, these bullish catalysts failed to prevent a dramatic price reversal. Since the 52-week high of €261.00 (about $285) on June 22, Nebius has fallen roughly 37%.
The market’s outlook for AI computing demand remains optimistic. The combined AI capital expenditures of Amazon, Microsoft, Alphabet, and Meta are projected to reach $725–770 billion in 2026, up more than 77% from 2025. Yet, even as demand for compute power continues to surge, why are providers facing heavy sell-offs? The answer lies in a structural shift in how the market values AI cloud infrastructure companies.
Growth Narrative Remains Intact, but Market Focus Shifts to Profit Realization
Nebius’s growth story remains solid by the numbers. In Q1 2026, the company reported $399 million in revenue, up 684% year-over-year; AI cloud business revenue reached $389.7 million, up 841%. Adjusted EBITDA turned positive at $129.5 million, and cost of revenue dropped to 26% of sales. The company reaffirmed its full-year 2026 revenue guidance of $3–3.4 billion and set an end-of-year annualized run rate (ARR) target of $7–9 billion.
On the customer side, Nebius signed a five-year, $27 billion AI infrastructure agreement with Meta. As of Q1, remaining performance obligations stood at $33.59 billion, with a backlog exceeding $40 billion.
These figures paint a picture of robust demand, rapid revenue growth, and strong customer commitments. However, the market’s focus has shifted from "how much can revenue grow" to "when and at what scale will profits materialize."
Front-Loaded Capex: The Heavy Burden of $20–25 Billion
Nebius raised its 2026 capital expenditure guidance from $16–20 billion to $20–25 billion. The main driver: accelerated investment in capacity for 2027, with new capacity not expected to generate revenue until the first half of that year.
This means Nebius is under immense cash flow pressure even as revenue soars. The company’s debt has reached about $9.47 billion. While it holds roughly $9.3 billion in cash reserves, annual capex of $20–25 billion means current reserves cover less than half a year’s spending.
The model is clear: invest heavily today to build tomorrow’s compute capacity, then monetize that capacity through future revenue. But this approach hinges on two assumptions—continued growth in AI compute demand, and Nebius’s ability to convert capacity into profits at sufficient margins. The market is now stress-testing both.
GPU and Power Costs: Neocloud Model Faces Margin Squeeze
The business model for AI cloud infrastructure is not as simple as "buy GPUs, rent out compute, pocket the profits." Neocloud companies face ongoing margin pressure from three major cost lines: GPU procurement, data center construction, and electricity.
Tight GPU supply chains have driven up operating costs. In July 2026, Nvidia GPU server component costs fluctuated by as much as 40% in a single week, forcing Nebius and AWS to raise rental prices by up to 30%. On June 1, Nebius increased on-demand compute rental prices by about 30%. Such cost volatility is rare in traditional cloud computing but is becoming the norm for AI cloud infrastructure.
Electricity costs are another critical variable. Neocloud data centers rely on stable power, but surging global data center demand is driving up electricity procurement costs. Each additional 1 GW of compute capacity requires about $3.5 billion in investment. Nebius plans to build over 5 GW of compute capacity by the end of 2030—requiring more than $175 billion in long-term capital just for capacity buildout.
Deutsche Bank analysis suggests that if Meta adds 7 GW of compute in 2027, related AI infrastructure investment could theoretically hit $245 billion. These numbers highlight a reality: the cost of building AI infrastructure is rising far faster than revenue.
Double Pressure: Customer Concentration and Competitive Landscape
Customer concentration is another core risk. Nebius’s performance is highly dependent on a handful of major clients—Meta’s $27 billion deal is the most prominent. Some analysts estimate that a significant portion of future revenue growth will come from just two key customers, with Microsoft expected to reach full annualized run rate contribution from 2027, and Meta essentially reaching full run rate in 2026.
This concentration creates a dual risk: if a major client changes its procurement plans or builds its own compute, Nebius’s revenue takes a direct hit. At the same time, large customers’ bargaining power could squeeze Nebius’s profit margins.
Competitive dynamics add further pressure. Reports indicate Meta Platforms plans to sell excess AI compute through its own cloud business. If hyperscalers—Amazon AWS, Microsoft Azure, Google Cloud—internalize most of their AI compute needs, Neocloud companies may be left competing for overflow capacity at thinner margins. This is the core logic behind the bearish narrative now playing out in the market.
Valuation Compression: Classic High-Beta Stock Pullback
From a valuation perspective, Nebius’s pullback follows the classic pattern of high-beta stocks. Before the plunge, its trailing P/E ratio had climbed to 75.87—among the highest in the AI cloud sector. Its 30-day annualized volatility hit 119.30%—a level rare for regular stocks and more typical of options.
Sector-wide rotation amplified the stock’s drop. On July 24, investors broadly exited high-valuation AI infrastructure stocks. CoreWeave fell 7% to $75.15 the same day. The divergence between pure Neocloud companies and diversified cloud providers was stark—Cloudflare’s stock was flat and up 33% year-to-date—indicating selective selling, with the market punishing companies whose valuations are entirely tied to AI infrastructure buildout expectations.
The volatility index surged 12% to 18.7 on Thursday, its highest recent close. High-beta, high-P/E stocks typically bear the brunt of volatility-driven rotations, and the Neocloud sector fits this profile.
Valuation Reset or Tactical Pullback?
Nebius’s plunge raises a question that goes beyond a single stock: What’s changing in the valuation logic for AI cloud infrastructure?
The bullish thesis is that enterprise AI workloads will require exponential growth in compute, driving demand for specialized GPU clusters. In this framework, Nebius—positioned as a fully vertically integrated Neocloud provider with hardware deployment, platform orchestration, and model optimization—holds a unique market position.
But the bearish case is playing out in real time: hyperscalers may capture most AI compute demand, leaving Neocloud companies to compete for overflow at lower margins. The core disagreement isn’t about whether AI compute demand will grow, but about who will capture that growth and at what profitability.
Wall Street’s consensus price target for Nebius is $258.13, with nine "Buy" or "Strong Buy" ratings and only one "Sell." Baird set a $250 target, and Northland Capital Markets raised its target from $248 to $410 on July 21. Yet the gap between analyst consensus and market price underscores the uncertainty currently priced in.
On July 29, Nebius will release its Q2 2026 earnings report. The market expects a per-share loss of $0.61, but revenue is projected to surge 450% year-over-year to $577.66 million. This report will offer the first systematic test of the issues above: Is revenue growth enough to offset ballooning capex? Are margins holding up under cost pressures? Has customer concentration led to observable pricing power loss?
FAQ
Q1: What directly caused Nebius’s 15% stock plunge?
On July 24, Nebius closed at $187.77, down 15.02% for the day. Direct triggers included pre-earnings profit-taking, systematic selling of high-beta stocks as volatility spiked, and a market reassessment of the $20–25 billion capex guidance. Insider selling of over $140 million in stock over the past 90 days also heightened market concerns.
Q2: If AI compute demand is still growing, why is Nebius being sold off?
The core market debate isn’t about demand, but about profit distribution. Growing AI compute demand doesn’t automatically translate to high profits for Neocloud companies—volatile GPU procurement costs, data center investments, rising electricity prices, and customer bargaining power are all eroding margins. The market is repricing the question of "if and when compute providers can achieve sustainable profitability."
Q3: Is Nebius’s financial position healthy?
Nebius holds about $9.3 billion in cash, posted Q1 revenue of $399 million (up 684% YoY), and turned adjusted EBITDA positive at $129.5 million. However, it has $9.47 billion in debt, and 2026 capex guidance is a hefty $20–25 billion. Short-term liquidity is solid, but massive capex poses significant long-term cash flow pressure.
Q4: How do Neocloud companies differ from traditional cloud providers?
Neocloud companies focus on AI-specific compute infrastructure, with GPU clusters as core assets, providing specialized cloud services for AI training and inference. Compared to traditional cloud providers (AWS, Azure, GCP), Neoclouds are more specialized but have higher customer concentration, denser capex, and face the risk of hyperscalers internalizing their AI compute needs.
Q5: Does the long-term investment case for Nebius still hold?
The core long-term thesis—sustained growth in AI compute demand—remains intact. Nebius has over $40 billion in backlog, Nvidia’s 9.3% strategic stake, and full-stack vertical integration. But short-term valuation reset pressures are real, and the Q2 earnings on July 29 will be a key test of this thesis.
