Nebius vs. CoreWeave: Which Is the Better Artificial Intelligence (AI) Infrastructure Stock to Buy Right Now

AI data-center demand is running ahead of supply, with Bank of America estimating a combined $2.3T backlog across Microsoft, Oracle, Amazon, and Google (end of Q2), supporting long-term growth for neocloud operators. CoreWeave revenue rose 112% YoY to $2.6B in Q2 while its loss per share increased 90% as it expanded infrastructure; Nebius revenue jumped 454% YoY to $582M and reduced adjusted net loss by 64% to $33.2M. Despite Nebius’ weaker stock performance vs CoreWeave’s stronger backlog visibility (CoreWeave revenue backlog: $104B, up 246% YoY), the article argues Nebius looks more attractive on profitability—though it trades at a higher sales multiple—implying a risk-adjusted preference depending on investor profile.

Analysis

The market is still underpricing the difference between booked demand and monetizable demand. In this sub-sector, the scarce resource is not customer interest; it is financed power, GPUs, and the ability to turn backlog into cash flow without repeated dilution. That makes the winners over the next 1-3 months less about who can announce the biggest pipeline and more about who can prove utilization, gross margin stability, and capex discipline on the next print.

The likely relative beneficiaries are the firms with the strongest balance sheets and the vendors that sell the picks-and-shovels into the buildout. MSFT, AMZN, GOOGL, ORCL and NVDA should absorb more of the economic rent than the neoclouds if capacity stays tight, because they can self-fund infrastructure and capture pricing power while smaller players absorb financing risk. The second-order loser is any model that depends on perpetual external capital to keep up with demand; that is where multiple compression usually starts when the market shifts from revenue growth to cash burn.

Contrarianly, the consensus is treating backlog as if it were deferred revenue rather than a contingent claim on future capacity. If power delivery, GPU allocation, or customer concentration slows conversion, the stocks can de-rate quickly even with strong top-line prints; that is the main 6-18 month falsifier. So the setup is not uniformly bullish: the right trade is to own the better-funded infrastructure layer and be selective on the asset-light growth stories that need flawless execution to justify premium sales multiples.

Key Decisions for Investors

• Prefer long MSFT / AMZN / GOOGL over NBIS and CRWV on a 6-12 month horizon; these names capture the same AI demand with lower refinancing and dilution risk. Risk/reward is better if the next cycle re-rates quality over headline growth.
• No fresh directional trade in NBIS or CRWV on this article alone; wait for the next earnings release to see backlog-to-revenue conversion, gross margin, and capex intensity before committing capital.
• If forced into a relative-value expression, consider short NBIS vs long CRWV only on valuation divergence, not fundamentals. This is a tactical mean-reversion trade with a tight stop if NBIS continues to prove operating leverage.
• Use NVDA as the cleaner long if you want exposure to the same capex wave; the thesis is that the supply-constrained middlemen will compete away economics while the component supplier keeps pricing power. Reassess if NVDA lead times normalize or AI capex guidance rolls over.
• Set an alert for any hint of equity issuance, debt refinancing, or customer concentration problems at NBIS/CRWV; that would be the first real signal that backlog quality is deteriorating and the trade should be reversed.