Neocloud’s Ledger: Nebius’s 454% Revenue Surge, but the Structural Flaws Are in the Footnotes

The numbers are out, and they are explosive. Nebius (NBIS) posted a 454% year-over-year revenue increase to $582.3 million in its Q2 2025 earnings, with adjusted EBITDA swinging to a profit of $236.2 million. The stock responded with a 34.14% single-day surge—its largest since listing. IREN and Coreweave followed, rising 5% and 3% respectively in early US trading on August 13. The market is calling this a Neocloud sector breakout. But I’ve been through enough Solidity audits to know that the headline numbers are never the full story. The transaction log—in this case, the financial footnotes—tells a different tale. Let me set the context. Neocloud refers to a cluster of AI-focused cloud infrastructure providers that have emerged over the past two years, distinct from the Big Three (AWS, Azure, GCP). The three companies in focus here are IREN (formerly Iris Energy), a Bitcoin miner pivoting to AI GPU services; Nebius, the European AI cloud spun out of Yandex’s assets; and Coreweave, a pure-play GPU cloud provider that went public earlier this year at a ~$23 billion valuation. All three share a capital-intensive model: they build and operate massive GPU clusters, primarily using NVIDIA hardware, and lease compute power to AI startups and enterprises. The bull case is simple: AI training demand is insatiable, and these nimble players can scale faster than the hyperscalers. But the core of any analysis is verification. I have audited over 40 smart contracts in 2017, and I learned that reproducibility is the only currency of truth. So let’s verify Nebius’s numbers. The $582.3 million revenue figure is striking—but what is the base? The prior year Q2 2024 was $105.1 million, which itself was a recovery from post-split restructuring. The 454% growth is real, but the EBITDA margin of 40.6% is suspiciously high for a capital-intensive cloud business. In my 2020 DeFi stress-testing work, I modeled liquidity ratios for Compound and Aave, and I learned that high margins in capital-intensive businesses often come from one-time events or non-recurring contracts. Nebius’s adjusted net loss still stood at $33.2 million, though narrowed by 64%. The key question is: how much of the revenue is attributable to a single large customer or a contract that may not repeat? The earnings release does not disclose customer concentration, but the footnotes mention “a significant portion of revenue from a limited number of clients.” That is a red flag. I have seen this in NFT wash-trading patterns in 2021: when a few wallets accounted for 15% of floor price volume, the whole market looked healthy until the manipulation stopped. Nebius’s revenue may have a similar fragility. Now the contrarian angle. The market is treating the Nebius earnings as a sector-wide signal. IREN and Coreweave rose on the coattails of that news, but their own fundamentals are far less clear. IREN’s 5% gain is the strongest among the three, yet the company has not released any material update on its AI cloud contracts. Its HPC business, branded Aurora, is still in early stages. In my 2022 bear market rebalancing, I learned that correlation does not equal causation. The sector is moving as a block, but the individual risk profiles are distinct. IREN remains a Bitcoin miner with a GPU side-hustle; its electricity cost advantages are real, but the capital allocation between ASIC and GPU creates internal competition. Coreweave, meanwhile, is heavily leveraged—it borrowed against its GPU hardware to fund expansion. In a rising interest rate environment, that debt service becomes a structural flaw. The market is ignoring this because the narrative is bullish, but volatility is noise; structural flaws are signal. Another hidden risk is GPU supply concentration. All three companies depend on NVIDIA for chips. Any delay in Blackwell GPU deliveries or a shift in NVIDIA’s allocation strategy could cripple their expansion plans. I have seen this in DeFi: when a single oracle (like Chainlink) accounts for 80% of price feeds, a protocol is vulnerable. Here, the single point of failure is NVIDIA. The sector’s high growth is predicated on the assumption that NVIDIA will continue to prioritize these Neocloud providers over the hyperscalers. That assumption is unverified. Finally, the takeaway. The next week’s signal is not the stock price—it is the capital expenditure guidance. Neocloud companies must disclose their CapEx plans in the upcoming earnings calls. If Nebius guides for a CapEx figure that exceeds its EBITDA by a wide margin, the market will eventually price in dilution. I will be watching the 10-K filings for customer concentration and the debt-to-equity ratios. The bytecode lies; the transaction log does not. The earnings numbers are the transaction log, and they tell me that Nebius has passed the first test of viability, but the structural flaws—customer concentration, GPU dependency, and capital intensity—remain. Trust the hash, verify the execution path.