Should CoreWeave and Nebius Group Investors Be Worried About Circular Financing? Here's What the Numbers Say

CoreWeave (CRWV) and Nebius (NBIS) both rely on Nvidia backstops/backstop-like structures to finance AI data-center buildouts amid “circular” vendor financing concerns. CoreWeave scaled to a $10B/year run rate but carries about $35B of debt and faces 2026 capex of $35B–$39B, while Nebius grew sharply (up 454% YoY to $582M last quarter) and raised $5.75B via convertible notes alongside >$20B planned 2026 capex after $8B spent in the first half. With AI demand commitments potentially less binding if growth slows—shares down 41% (CoreWeave) and 30% (Nebius) from highs—the news is a cautious signal about financing fragility rather than an immediate default risk.

Analysis

The market is starting to separate AI capex as a growth story from AI capex as a financing story. That matters because the first phase rewards revenue growth, while the second phase is about who owns the balance sheet when utilization normalizes. The neocloud model is effectively a levered call option on sustained GPU scarcity; Nvidia’s support lowers default probability, but it does not lower equity dilution risk or the risk that returns on incremental data-center build-out compress faster than expected.

Near term, this is more about multiple pressure than fundamental collapse. CRWV and NBIS can still post strong topline growth, but the market will increasingly demand proof that booked demand converts into cash flow before absorbing another wave of capex. If AI demand pauses for even one budget cycle, the reflexive funding loop weakens first in equity, then in converts/credit; that is the right sequence to watch over the next 1-3 months. Longer term, the structural winners are the platforms with self-funded optionality (MSFT, META) and the chip supplier with the best papered-in demand, but even NVDA can see a quality-of-growth discount if investors conclude it is financing its own end markets.

The contrarian view is that bankruptcy risk is not the main issue; liquidity support makes these names survivable. The bigger risk is that survivable does not mean investable: if capacity is being built ahead of durable end-demand, incremental returns on capital can fall sharply while reported revenue keeps rising. The thesis is falsified if utilization stays tight through the next two quarters and financing spreads remain benign despite continued capex acceleration.

Key Decisions for Investors

• Short CRWV vs long NVDA on any 8-12% relief rally in CRWV; catalyst window 1-3 months. Thesis: NVDA retains demand, while CRWV carries the leverage and dilution risk if utilization or financing terms wobble.
• Prefer MSFT and META over NBIS/CRWV as a cleaner AI capacity exposure for the next 6-12 months. These names can delay or self-fund capex, so they have less financing convexity if AI demand cools.
• Buy CRWV/NBIS put spreads 2-4 months out if implied vol remains elevated but equity keeps drifting lower. Defined-risk way to express that the market is underpricing dilution and refinancing risk, not just business risk.
• Watch NVDA on any sign that backstop/circular-financing optics start to weigh on gross-margin multiple. If the stock underperforms semis on strong results, reduce long exposure because the market may begin assigning a lower quality premium to incremental AI demand.