Nebius Raised Its Year-End Power Target From More Than 3 Gigawatts to 5 in 6 Months

Nebius raised its contracted power target to 5 GW (from >3 GW in February and >4 GW in May) alongside 2Q revenue up 454% YoY to $582.3M. The company’s AI cloud deals are priced at $20M–$25M annual contract value per megawatt, up from about $12M per MW on its prior 2026 base, and it reported adjusted EBITDA margin of 50% in the quarter. However, capex remains heavy ($5.7B P&E in 2Q; $20B–$25B full-year expected) and the firm is still loss-making (operating loss of $176M), though guidance was reaffirmed for $3.0B–$3.4B revenue and a $7B–$9B year-end run-rate target.

Analysis

The important signal is not that demand exists; it is that Nebius is repeatedly discovering scarce power faster than it can turn it into revenue. That makes the equity less a normal operating-company valuation and more a financed execution option: the upside is real if the build is flawless, but the path is dominated by time-to-energization, dilution, and the cost of carrying idle assets. In the next 1-3 months, the stock should trade on confidence in bookings and financing, not on earnings quality.

Second-order winners are the GPU and infrastructure suppliers that monetize every incremental watt regardless of whether the operator’s ROIC works. NVDA is the cleaner expression of the capex cycle because it captures demand without taking construction, interest, or customer-concentration risk; electrical gear, cooling, and power-management names should also benefit, though the article does not name them. The loser set is any higher-cost AI cloud competitor that needs to match aggressive pricing or capex intensity to stay relevant, because that tends to compress future returns on capital across the space.

The contrarian point is that the market may be underpricing how long the cash conversion lag can persist. If deployment slips even modestly, the model becomes a heavier balance-sheet story than a growth story, and that usually hits multiples before it hits reported revenue. Falsifiers are concrete: energized capacity ramping faster than planned, sustained prepayment coverage above expectations, or a credible path to self-funded expansion by the 2027 window.

Key Decisions for Investors

• Long NVDA vs. short NBIS as a 1-3 month pair trade: own the demand beneficiary, fade the capital-intensive executor. Best entry is on any NBIS strength after financing or contract headlines; thesis breaks if NBIS demonstrates materially faster-than-planned capacity deployment and narrows the cash burn gap.
• Buy 3-6 month NBIS put spreads on post-rally strength rather than shorting outright. The risk/reward is attractive if the market starts discounting 2027 revenue as if it were near-term cash flow; cover if management raises the 2027 energization cadence above 1 GW/year or if guided capex meaningfully steps down.
• Prefer semis/AI infrastructure exposure through NVDA over a direct NBIS long until there is evidence of sustained free-cash-flow inflection. NBIS can work as a trading vehicle, but not yet as a core position; the thesis needs proof of deployment, not just contract value.
• Set an alert for any revision to capex, debt issuance, or prepayment coverage. A downgrade in financing terms or a gap between signed commitments and actual energization is the clearest catalyst for multiple compression over the next 1-2 quarters.