3 Cloud Leaders That Will Make a Fortune From AI

AWS and Google Cloud are growing rapidly, with reported growth rates of 37% and 82%, respectively, and each plans to spend at least $200 billion on data-center capex this year. Their Q2 backlogs stood at $496 billion for Amazon and $514 billion for Google Cloud. Nebius reported Q2 revenue growth of 454% year over year; analysts expect growth of 519% in Q3, 540% in Q4 and 268% in 2027, but the company is currently unprofitable, making execution and eventual profitability key risks.

Analysis

The key risk is not whether AI compute demand exists, but whether booked demand converts into durable, high-return utilization. Backlog can support visibility while still leaving timing, cancellation, pricing, and customer concentration unresolved; meanwhile, data-center spending commits capital before returns are proven. For Amazon and Alphabet, the second-order test is whether cloud growth translates into stronger incremental returns after power, depreciation, and custom-chip costs—not just revenue acceleration. Their diversified businesses provide a buffer, but capex intensity could still weigh on free cash flow and valuation if utilization disappoints.

Nebius is a higher-beta bet on capacity scarcity, with a more asymmetric downside: rapid expansion can magnify funding needs and depreciation exposure just as GPU capacity ages and larger providers compete on price or bundle compute with software. Power availability and customer concentration are material diligence gaps. The article’s analyst growth estimates are not evidence of profitable unit economics.

Near term, earnings commentary on capacity utilization, pricing, capex, and customer commitments is likely to matter more than headline growth. Over 1–3 months, watch guidance and financing disclosures; over 6–18 months, the differentiator is cash generation per unit of deployed capacity. Contrarian view: AI demand may remain strong while returns accrue disproportionately to customers or the most capital-efficient providers, rather than every infrastructure supplier.

Key Decisions for Investors

• Prefer AMZN and GOOG to NBIS for core exposure; consider a relative-value long in the hyperscalers versus a smaller NBIS position rather than treating all three as equivalent AI beneficiaries. Size NBIS for substantial drawdown risk; avoid initiating solely on revenue growth.
• Use upcoming results as entry catalysts: add only if cloud growth is accompanied by evidence of improving utilization or returns on incremental capital. Reduce exposure if capex rises while free-cash-flow conversion weakens or management pushes expected returns further out.
• Keep NBIS on a watchlist pending verification of cash runway, financing terms, customer concentration, capacity utilization, and economics after depreciation and power costs. These missing details determine whether growth is value-accretive or capital-intensive share loss.
• Thesis falsifiers: hyperscaler guidance indicating weaker cloud demand or deteriorating incremental returns; NBIS needing materially dilutive financing; or evidence of sustained price compression/underutilized capacity. Conversely, improving cash generation and firm, diversified customer commitments would support NBIS exposure.