Meet the 2 Newcomers Challenging the Cloud Computing Titans in Artificial Intelligence (AI)

AWS grew revenue 28% in Q1 and generated 59% of Amazon’s operating income, while Azure grew 40% and Google Cloud 63%, underscoring strong profitability and demand among the large cloud players. CoreWeave and Nebius are growing much faster on a percentage basis — 112% and 684% year over year in Q1 — but remain unprofitable as they reinvest heavily in capacity. The piece argues the legacy cloud leaders remain strong businesses, while the neocloud names offer higher-risk, higher-upside AI exposure.

Analysis

The market is implicitly re-rating AI infrastructure from a pure growth story into a capacity-constrained utilities-like market. That favors the scaled incumbents first because they control distribution, procurement, and financing, but the second-order winner is Nvidia-like ecosystem leverage: every incremental dollar of AI cloud spend still requires high-end accelerators, networking, and power-intensive racks, so the capex cycle stays sticky even if pricing pressure emerges. The key distinction is durability of margins: the large clouds can absorb lower unit economics because they monetize the same customer through broader software and enterprise relationships, while the neoclouds are forced to convert growth into fixed-asset intensity almost immediately.

CoreWeave and Nebius likely remain narrative stocks until they prove operating leverage on a trailing-twelve-month basis. The risk is not simply “no profits”; it is that the market may eventually penalize them for front-loading depreciation, power contracts, and lease obligations just as AI demand normalizes from burst to steady state. If hyperscalers decide to reserve more internal capacity or bundle AI compute into larger enterprise contracts, the neoclouds could see growth decelerate faster than consensus expects because their client concentration leaves them exposed to a handful of large procurement cycles.

The contrarian view is that the current enthusiasm may be underestimating the scarcity value of near-term GPU-ready capacity. In the next 6-12 months, any delay in new datacenter energization, grid interconnects, or chip deliveries can keep pricing irrationally strong, which supports the high-growth names longer than fundamentals would justify. But over a 12-24 month horizon, the incumbents’ balance sheets and cross-subsidy power likely compress the addressable profit pool for the smaller players.

For portfolio construction, this is a relative-value, not an outright sector-long, setup: the upside is in ownership of the infrastructure toll collectors, while the downside sits in capacity financiers. The cleanest expression is to own the cash-generative platforms and selectively fade the most expensive growth optionality where expectations are already pricing a near-perfect execution path.

Key Decisions for Investors

• Long AMZN vs short CRWV on a 3-6 month horizon: favor the company with embedded cloud cash flow and broader enterprise monetization over the pure-play capex story; target a 15-20% relative spread if AI spend remains strong but financing costs stay elevated.
• Long MSFT as a quality AI infrastructure compounder over 6-12 months: use any pullback to add because Azure benefits from enterprise bundling and has lower earnings volatility; risk/reward favors low-20s forward earnings expansion rather than multiple compression.
• Selective long GOOGL on 12-month horizon, but hedge with a small short CRWV/NBIS basket: the TPU angle supports margin resilience, while the hedge offsets any broad AI enthusiasm that favors the smaller names on momentum.
• Avoid outright long NBIS/CRWV ahead of earnings unless using calls: these remain high-beta execution stories; if participating, use limited-risk option structures with 2-3 month tenor to capture sentiment without balance-sheet downside.
• Pair trade META long against NBIS short if you want AI exposure with less financing risk: META has internal demand and optionality, while NBIS is more exposed to a narrow set of external contracts; look for the spread to work over 6-9 months.