Nebius Keeps Getting My Money. There Is One Reason and It Is Not Hype.

Nebius reported $37.49 billion of remaining performance obligations and roughly $40 billion of total contracted backlog, versus a market capitalization of about $49.49 billion, supporting substantial multiyear revenue visibility. Q2 revenue rose 454.04% year over year to $582.3 million, while adjusted EBITDA improved to $236 million from a $21 million loss and AI Cloud achieved a 50% adjusted EBITDA margin. Customer prepayments, which covered about 70% of Q2 deals, are expected to provide more than $9 billion in 2026 funding and cover 50%-60% of related capex, reducing external financing needs; however, customer concentration, $8.5 billion of convertible debt carrying value, and quarterly GAAP EPS of -$0.68 remain material risks.

Analysis

NBIS is transitioning from an equity-funded GPU buildout toward a customer-financed infrastructure model, which can materially improve FCF conversion if prepayments are tied to enforceable capacity commitments rather than cancellable reservations. The key valuation question is not the backlog headline but conversion: a high share of short-duration, premium-priced capacity would lift near-term revenue and EBITDA but creates a larger repricing cliff as supply normalizes. Verify contract duration, termination rights, cash-versus-letter-of-credit collateral, and whether the largest customer commitments are take-or-pay before capitalizing the full order book.

Customer concentration creates a non-linear downside: a reduction in utilization or expansion pace by any one major buyer would simultaneously impair revenue, reduce prepaid-capex funding, and force NBIS back into dilutive equity or expensive debt markets. The gap between convertible carrying value and fair value implies substantial embedded dilution sensitivity if the equity remains strong; upside in the stock may therefore accrue less efficiently than headline EBITDA growth suggests. Over the next 1-3 months, evidence that new customer prepayments fund incremental sites without increased net debt would support multiple expansion; over 6-18 months, the decisive issue is whether returns on deployed megawatts remain above the cost of capital after power, depreciation, and GPU-refresh costs.

The consensus likely overweights scarcity pricing and underweights customer vertical integration. MSFT has strategic reasons to secure external capacity quickly, but its long-run incentive is to internalize supply where economics permit; NBIS should be valued as a capacity merchant, not as a software-like recurring-revenue platform. A sustained narrowing of AI-cloud capacity premiums, slower NVIDIA accelerator delivery, power interconnection delays, or a major customer's capex revision would be the clearest falsifiers.

Key Decisions for Investors

• Do not add outright NBIS solely on reported contracted backlog; require next-quarter disclosure showing cash prepayments rising in line with deployed capacity and no deterioration in customer concentration. Treat failure to convert backlog into revenue/run-rate growth within two reporting periods as a thesis break.
• For high-beta AI-infrastructure exposure, prefer a 3-6 month long NBIS / short MSFT dollar-neutral pair only after NBIS demonstrates funded capacity additions; this isolates merchant-capacity execution from broad AI-capex sentiment. Exit if NBIS announces incremental equity financing or if MSFT signals reduced third-party GPU-cloud procurement.
• Use defined-risk exposure rather than common equity if NBIS rallies materially ahead of earnings: buy 6-9 month call spreads financed with higher-strike calls, targeting upside from proof of backlog conversion while limiting losses from dilution or a single-customer reset.
• Monitor credit as the early-warning indicator: widening spreads or higher funding costs on NBIS asset-backed facilities would signal that lenders are discounting collateral quality or contract durability before equity estimates adjust. Avoid adding if financing terms move materially above the stated secured-funding benchmark.