$10,000 in Nebius When It Returned to Trading in 2024 Is Worth About $105,000 Now

Nebius has secured more than $40 billion in customer commitments, including a Microsoft contract worth $17.4 billion through 2031 (up to $19.4 billion) and a Meta agreement that could reach $27 billion. Q2 revenue surged 454% year over year to $582.3 million and adjusted EBITDA swung to a $236.2 million profit, or a 41% margin. However, management plans $20 billion-$25 billion of annual capex against just $3.0 billion-$3.4 billion of guided revenue, leaving the company reliant on customer prepayments and likely additional debt or equity financing while major capacity ramps toward 2027.

Analysis

NBIS is transitioning from an equity story priced on contracted AI capacity to a financing-and-execution story. The key sensitivity is not headline backlog but conversion of commitments into enforceable take-or-pay revenue, customer prepayments, and deployed GPU utilization; a modest delay in power delivery, construction, or customer acceptance can leave depreciation and interest expense accruing ahead of revenue. Its reported adjusted margin is therefore less informative than future EBIT/FCF after GPU depreciation, network costs, and the cost of repeated external capital.

The likely second-order beneficiary is NVDA: accelerated NBIS buildout expands GPU demand and creates another well-capitalized buyer, but this is economically a lower-quality demand vector than hyperscaler capex because it relies on NBIS retaining access to equity and convert markets. MSFT and META gain capacity diversification and negotiating leverage versus CoreWeave and the major cloud platforms; they are not meaningfully dependent on NBIS operationally. Competitive pressure should be greatest on other GPU-cloud operators with similarly leveraged construction pipelines, where customer concentration and refinancing needs are comparable.

Over the next 1-3 months, NBIS can remain momentum-driven if quarterly deployment milestones and prepayment receipts validate the build schedule. Over 6-18 months, the central risk is dilution: capex materially above internally generated cash flow means each incremental capacity tranche may require equity issuance, converts, asset-backed debt, or customer-funded structures. The contrarian view is that the market may be underpricing the value of contracted scarce power/GPU capacity, but it is also likely extrapolating an adjusted EBITDA margin that excludes the economic cost of rapidly obsolescing hardware.

Key Decisions for Investors

• Do not chase NBIS spot strength; maintain a watch-only stance until the next filing quantifies take-or-pay terms, cancellation rights, customer prepayments, and debt/convert maturity profile. Upgrade only if deployed-capacity revenue and operating cash flow scale faster than capex without a new equity raise.
• For an AI infrastructure allocation, prefer long NVDA over NBIS for the next 6-12 months: NVDA captures NBIS procurement while avoiding NBIS's construction, utilization, and financing risk. Falsifier: evidence of broad GPU order deferrals or a material reduction in NBIS procurement guidance.
• Consider a tactical NBIS short versus long NVDA after a sharp NBIS rally or ahead of financing announcements, sized small given squeeze risk. The thesis is multiple compression from dilution or commissioning delays rather than a collapse in AI demand; cover on disclosed customer-funded financing that materially reduces equity issuance risk.
• Monitor NBIS quarterly for revenue recognized from its two largest customers, capex per incremental revenue dollar, and operating cash flow. A sustained improvement in those three metrics for two quarters would invalidate the bearish financing thesis and support reassessing a long position.