Nebius reported Q1 2026 revenue of $399 million, up 684 percent year-over-year, and net income from continuing operations of $621.2 million against a loss of $104.3 million a year earlier. The headline figures are arithmetically correct and analytically misleading. Reported net income is dominated by a $780.6 million non-cash gain from revaluation of investments in equity securities — almost certainly the mark-up of the ClickHouse stake following its latest funding round. Strip it out, along with stock-based compensation and one-off restructuring charges, and Nebius posted an adjusted net loss of $100.3 million, modestly worse than the $83.6 million loss a year ago. The operating line tells the same story: a $128 million loss from operations, only marginally better than the $120 million operating loss of Q1 2025 despite revenue growing nearly eightfold. Nebius is not yet profitable in any operational sense, and the bottom-line print should be read as a financing event, not an earnings event.
The figure that matters in this release is buried in the cash flow statement. Deferred revenue rose by $3.2 billion in the quarter, lifting the balance sheet total from $1.6 billion at year-end to $4.8 billion. This is customer prepayment — almost certainly the Microsoft and Meta commitments announced earlier this year beginning to flow through as cash, with multi-year compute capacity reservations recognized as a liability on the balance sheet until delivered. It is the same accounting signature CoreWeave produced when it disclosed its anchor cloud commitments. It is also what allowed Nebius to report $2.26 billion of operating cash flow against an operating loss; the cash came from customers paying in advance for capacity that will be delivered, recognized as revenue, and depreciated against over the contract life. Investors should read deferred revenue as the contract book — and the contract book grew by an order of magnitude inside ninety days.
Capacity Is Being Built With Other People’s Money
Nebius spent $2.47 billion on property, equipment, and intangible assets in the quarter, and another $170 million on acquisitions. The financing came from three distinct sources: $4.3 billion in convertible notes issued in the quarter, $2.0 billion in prefunded warrants, and the $3.2 billion deferred revenue inflow noted above. Total debt rose from $4.1 billion at year-end to $8.4 billion. Cash on hand more than doubled from $3.7 billion to $9.3 billion. The capital stack is being rebuilt at scale to finance a buildout whose duration extends well beyond the current cash position, and the company is signaling — through the simultaneous use of convertibles, warrants, and prepaid customer commitments — that it does not intend to rely on any single funding channel.
The Pennsylvania announcement disclosed alongside the print — 1.2 GW of power and land secured for a new owned AI factory — is the operational counterpart to the financing. Combined with the Missouri gigawatt-class site already under construction in Independence, Nebius is committing to multi-gigawatt owned infrastructure in the United States. Owned, not leased, matters here: it reflects a decision to take depreciation onto the Nebius balance sheet rather than rent capacity from a third-party colocation provider. That is the hyperscaler model, and the company is now financing it as a hyperscaler does — with customer prepayments locking in revenue, convertibles financing the build, and equity-linked instruments diluting at the top of the capital stack. The 2026 capex guide of $16 to $20 billion implies that the $2.5 billion spent in Q1 is the floor, not the run-rate.
Margin Trajectory Is the Bull Case
The operating leverage is real and measurable. Cost of revenues fell from 49 percent of revenue to 26 percent year-over-year. Product development collapsed from 72 percent to 17 percent. Sales, general, and administrative dropped from 120 percent to 36 percent. These are extreme percentage moves, partly an artifact of the small revenue base in the prior-year quarter, but the absolute spend lines are growing far slower than revenue. Adjusted EBITDA flipped from a $54 million loss to $130 million in profit. The path to GAAP operating profitability runs through depreciation and amortization, which rose to $212 million in the quarter on a property and equipment base of $7.1 billion, and which will continue to climb materially as the Missouri and Pennsylvania facilities come online over the next eighteen to thirty-six months. The question is not whether Nebius can produce EBITDA — that has been answered — but whether revenue per dollar of depreciating capital can compound faster than the depreciation itself.
The diluted share count is now 309 million against 258 million basic, reflecting the conversion overhang from the convertible notes. The treasury holds another 68 million Class A shares, partially offsetting future issuance pressure. For investors, the question is whether the deferred revenue book grows fast enough to keep filling the next gigawatt before the convertibles need to be settled, and before the dilution shows up in fully diluted per-share economics. The mathematics of the trade is straightforward: every additional gigawatt of contracted capacity at hyperscaler-grade unit economics is worth more than the dilution required to fund it. Every gigawatt of speculative capacity is not.
Stock Trajectory: Where the Tape Goes From Here
Nebius enters the print trading near $176, up roughly 120 percent year-to-date and around 600 percent over the trailing twelve months, on a market capitalization of approximately $45 billion. The street is sharply bifurcated. Bank of America raised its target to $205 from $175 on May 11, ahead of the print, citing the Microsoft anchor commitment and the Meta option ladder. DA Davidson sits at $200. The consensus target across roughly a dozen analysts is approximately $170 to $178 — meaning the market has already absorbed much of the buy-side upside on a twelve-month view, with Wolfe Research the conspicuous outlier on Peer Perform citing execution risk at current scale. The trailing P/E of 364x is meaningless on a company still operating at a loss; the relevant valuation framework is forward EV / ARR, where Nebius trades at roughly five to six times the midpoint of its $7 to $9 billion exit-2026 ARR target. CoreWeave trades in the same range. Microsoft, by contrast, trades at roughly 11x revenue — the multiple gap is the embedded reward for becoming, or failing to become, a structural infrastructure provider rather than a cyclical compute reseller.
The base case from here is a continuation of the current pattern: deferred revenue compounds quarter-over-quarter as Microsoft and Meta capacity is delivered, ARR ramps toward the $7 to $9 billion exit target, and the multiple compresses modestly as revenue catches up to the share price. That points to a $200 to $215 twelve-month target — broadly consistent with where BofA and DA Davidson already sit. The bull case requires a Q2 print that shows the deferred revenue line accelerating rather than normalizing, plus visible progress on the Missouri site and Pennsylvania permitting, plus a third anchor customer disclosure. That scenario supports $240 to $260 by mid-2027, particularly if Nebius can begin to demonstrate revenue per megawatt that approximates CoreWeave’s published unit economics. The bear case is more interesting. It does not require Nebius to miss its ARR target; it requires the neocloud cohort to derate as a group, which would happen if a hyperscaler announces in-sourced capacity at scale, if Nvidia allocation tightens in favor of the three U.S. cloud incumbents, or if a single neocloud peer prints a working-capital event that the market reads as systemic. In that scenario Nebius compresses to three times forward ARR, which is roughly $120 to $140 — the low end of analyst targets, and not far from where Wolfe is implicitly positioned.
The asymmetry from here is no longer the asymmetry that produced the 600 percent twelve-month return. The easy money has been made. What remains is a credible multi-quarter path toward $215 at the midpoint, with a tail outcome on either side determined less by Nebius’s execution than by the trajectory of the neocloud cohort as a whole. The single most useful data point for resetting that distribution will be the Q2 deferred revenue print in August. If it grows again, the bull case stays intact. If it stalls, the bear case is no longer a tail.
The Position
Nebius is now visibly inside the second tier of AI infrastructure providers, behind the three U.S. hyperscalers and roughly alongside CoreWeave and a small set of European entrants. The Q1 print confirms commercial traction at hyperscaler-customer scale and a capital structure willing to fund the buildout. It does not confirm unit economics, and it does not yet demonstrate that the company can convert its contracted backlog into recurring free cash flow at hyperscaler-comparable margins. The non-cash mark-up on equity investments should be ignored entirely. The customer prepayment is the signal. Watch the deferred revenue line each quarter; that is where the contract book will reveal whether the Pennsylvania and Missouri capacity has already been sold, or still needs to be.