Better Cloud Investment: Amazon or Nebius Group?

Nebius Q2 revenue grew 454% year over year, versus AWS growth of 37%, but Nebius had a 30% negative operating margin while AWS posted a 39% operating margin. Amazon plans to spend about $220 billion on data centers this year and can fund expansion partly through its commerce business; Nebius is borrowing to grow and reported a payback period of just under two years for deals closed in Q2. The article favors a lower-risk portfolio allocation of 75% Amazon and 25% Nebius, while noting Nebius offers higher upside and risk.

Analysis

The key question is not whether AI compute demand is growing, but who can convert it into durable returns after power, hardware depreciation, and financing costs. Amazon’s advantage is funding flexibility and an established cloud customer base; its risk is that the scale of build-out pressures free cash flow before new capacity is absorbed. Watch AWS growth alongside segment operating income and Amazon’s cash-flow guidance—not capex alone.

Nebius’s reported deal payback period is promising but not equivalent to corporate cash payback: it may exclude shared infrastructure, financing costs, utilization ramp, or customer concentration. The negative operating margin makes execution sensitive to deployment delays and pricing. If capacity arrives faster than contracted demand, lower utilization can extend payback just as debt service continues. Conversely, constrained power or delayed hyperscaler capacity could support neocloud pricing and utilization. Microsoft Azure, Google Cloud, and Oracle are relevant competitive checks: their capacity decisions and pricing could limit Nebius’s ability to sustain attractive economics.

Near term, the article’s growth figures may favor NBIS sentiment, but they do not establish attractive per-share returns or a viable funding path. Over 1–3 months, verify debt maturities, liquidity, capex commitments, contracted backlog, customer concentration, and whether payback is measured on cash invested. Over 6–18 months, the decisive test is narrowing operating losses without repeated equity issuance or debt acceleration. The contrarian risk is treating AI demand as a guarantee of supplier returns: abundant demand can coexist with poor returns if capacity and capital costs outrun pricing power.

Key Decisions for Investors

• Prefer AMZN as the core exposure; do not infer that its data-center spend automatically translates into near-term free-cash-flow growth. Reassess on AWS growth, AWS operating income, and consolidated cash-flow guidance.
• Keep NBIS as a high-risk watchlist or limited satellite position, not a presumed 25% allocation. Before adding, verify the definition and realized performance of the sub-two-year deal payback claim, plus liquidity, debt maturities, committed capex, backlog conversion, and customer concentration.
• A conditional relative-value expression is long AMZN versus NBIS only if NBIS utilization or financing metrics disappoint while AWS economics remain resilient. Avoid initiating solely on the growth-rate comparison; the trade is falsified by sustained NBIS margin improvement and self-funding evidence alongside AWS deceleration or weaker returns on new capacity.
• Track power availability, GPU delivery timing, and cloud pricing over the next 1–3 months. Better-than-expected supply constraints could benefit NBIS near term; falling prices or delayed customer deployments would extend its cash burn and weaken the thesis.