Nebius vs. CoreWeave: Two Neoclouds, Two Reporting Frameworks

CoreWeave (CRWV) is the larger, leveraged, colocation-heavy neocloud market leader; Nebius (NBIS) is the smaller, owned-and-operated net-cash challenger growing six times faster off a much smaller base. The two now trade within 6% of each other on market cap despite a 5:1 revenue gap. The comparison below pulls every figure via the Daloopa MCP, which connects directly to Daloopa’s normalized financial database — financial statement data hyperlinked back to its original filing, and stock price history used to build the valuation section. This is particularly useful here because NBIS and CRWV define non-GAAP profitability differently, and reconciling the two requires auditability at the line level. The analysis includes a side-by-side reconciliation of both adjusted EBITDA and adjusted operating income for that reason.

Q1 2026 scorecard

How each company defines EBITDA

Both companies report “adjusted EBITDA” but the two definitions are not interchangeable. Two differences matter: stock-based compensation treatment and the equity-revaluation carve-out.

CoreWeave adds back 100% of stock-based compensation ($153M in Q1 2026, ~7.4% of revenue). Nebius adds back only “certain SBC expense” ($35M in Q1 2026), explicitly leaving the rest in the EBITDA calculation. On a like-for-like basis, CoreWeave’s headline EBITDA margin is overstated relative to Nebius’s by 4 to 5 points of revenue, attributable to the SBC add-back alone.

On a fully consistent definition (treating SBC as a real expense for both), the gap between Nebius’s 32% group EBITDA margin and CoreWeave’s 56% would narrow by several points.

Adjusted operating income: where the reporting differs

CoreWeave reports three non-GAAP profitability metrics: adjusted EBITDA, adjusted operating income, and adjusted net income. Nebius reports two: adjusted EBITDA and adjusted net loss. There is no “adjusted operating income” line in Nebius’s quarterly press releases or financial supplements. The non-GAAP bridge goes directly from operating loss to EBITDA territory.

The difference matters most for capital-intensive operators like these. Adjusted EBITDA strips out depreciation and amortization. For CoreWeave, that is $1,147M in Q1 2026, or 55% of revenue. Adjusted operating income includes D&A. For a business depreciating $36B of P&E, that is the difference between a 56% margin and a 1% margin.

CoreWeave’s adjusted OI bridge

CoreWeave reconciles from GAAP operating income (loss) to adjusted operating income with three add-backs: all SBC, acquisition-related costs and amortization of acquired intangibles. D&A stays in the cost base.

CoreWeave adjusted OI: the bear case in one chart

The decline is almost entirely a depreciation story. In 2024, when the fleet was smaller, D&A was a manageable drag, and the adjusted OI margin was 19%. FY 2025 D&A increased to $2,454M (48% of revenue); Q1 2026 D&A alone was $1,147M, more than half of revenue. Adjusted EBITDA still looks healthy because depreciation is excluded. Adjusted OI puts the business near breakeven on a fully loaded basis.

Nebius: no reported adjusted OI but constructible

Nebius doesn’t disclose adjusted OI, so a comparable figure must be built from the reconciliation components. The above-the-operating-line adjustments are partial SBC and one-off restructuring, with acquisition costs and intangibles amortization aren’t separately broken out in this bridge.

Two caveats: Nebius adds back only “certain” SBC ($35M) versus CoreWeave’s full $153M add-back, making Nebius’s constructed figure the more conservative of the two. Nebius’s D&A in Q1 2026 was $212M, 53% of revenue, matching CoreWeave’s structural depreciation drag. The constructed ~(21%) margin is wider than CoreWeave’s +1%, mostly because Nebius is still in pre-revenue ramp on its biggest contracts (Microsoft, Vineland, NJ, and Meta), driven by contract timing rather than structural cost differences.

Owned vs. leased facilities

The structural choice that drives most of the divergence in margin and balance sheet profile.

CoreWeave is almost entirely a colocation operator. Management confirmed on the Q1 2026 call that its 1+ GW of active capacity at quarter-end was sourced “entirely from long-term leases with data center partners,” with no single landlord above 17% of capacity. The owned P&E on the balance sheet is GPUs and racks inside leased buildings: $26.6B of technology equipment plus $9.6B of construction in progress. Finance-lease P&E is just $433M, ~1% of net P&E. The first self-built site (Kenilworth NJ) is targeted for late 2026.

Nebius is the inverse. Management disclosed in Q1 2026 thatowned data centers account for more than 75% of contracted capacity, with only ~25% from colocation. The flagship owned sites are Mantsala (Finland), Vineland NJ (the custom Microsoft facility, opening late 2025), a 1.2 GW Kansas City campus (power delivery starts H2 2026) and a second 1.2 GW greenfield in Pennsylvania announced May 2026. Operating lease ROU assets on the balance sheet are $1,266M, which are small relative to $7.1B of owned P&E. Nebius is targeting >4 GW of contracted power by year-end 2026 (vs. CoreWeave’s 3.5+ GW target).

The trade-off is visible in the margin trajectories. Colocation lets CoreWeave deploy faster with less upfront capital intensity. That is how it scaled to $5B of revenue from a standing start. Owned facilities are slower and more capital-hungry up front, but eliminate landlord margin over the contract life. That is part of why Nebius’s Core AI margin is now accelerating while CoreWeave’s adjusted operating margin has compressed from 19% in 2024 to 1% in Q1 2026.

Customer concentration

Both businesses have customer concentration exposure, but are moving in opposite directions. CoreWeave discloses it explicitly: Customer A (Microsoft) was 62% of revenue in 2024, peaked at 72% in Q1 2025, and fell to 45% in Q1 2026 as OpenAI, Meta, and Anthropic contract revenue ramped. Customer B (likely OpenAI) rose to 20%. The top two are still around 65% of revenue. Of the $99.4B total contracted backlog, Meta is ~40% and OpenAI ~25% per press disclosures.

Nebius does not disclose by customer in the same format, but the $17B-$19B Microsoft agreement plus the $27B Meta agreement together represent >$46B of committed long-term revenue against a customer base that was AI-startup-led as recently as 2024. Management has been explicit that diversifying the tail (Revolut, 1X Technologies, enterprises in pharma, manufacturing, life sciences) is strategic, with the qualified pipeline growing 3.5x quarter over quarter in Q1 2026.

Valuation: a twelve-month repricing in three acts

Over the trailing twelve months, the market has repriced both names in opposite directions. The narrative breaks into three distinct phases, each tied to a specific catalyst visible in the price history pulled from Daloopa’s MCP.

Act 1 (Mar-Jun 2025): CoreWeave is the only AI infrastructure trade

CoreWeave went public on March 28, 2025, pricing at $40 and closing the first day at $40.00. The stock then ran nearly uninterrupted to an intraday high of $187.00 and a closing high of $183.58 on June 20, 2025, a +359% return in under three months. The market was treating CRWV as the only viable public way to express the AI compute trade.

Meanwhile, Nebius saw more modest movement through the same period. On CRWV’s IPO day (March 28, 2025), NBIS closed at $22.31. By June 20, NBIS was at $47.97, up +115%, but the absolute share price was less than a third of CRWV’s. The market cap gap was wider still, given CRWV’s larger share count.

Act 2 (Jul-Aug 2025): CRWV gives back gains; NBIS slowly climbs

CoreWeave’s IPO lock-up expiration and questions about customer concentration triggered a multi-week selloff. CRWV closed at $86.59 on May 19th, $163.06 on June 30th, then $99.97 on August 15 and $121.08 on August 7 after Q2 earnings. Through August, the stock oscillated in a $90-$130 range as the market digested the debt load and the implications of 72% concentration in a single customer.

Nebius climbed steadily but with smaller absolute moves. NBIS went from $37.56 in mid-May to $65.31 on the Q2 print (August 7), reaching $71.62 by mid-August. Revenue was still under $110M a quarter, and the market was waiting for proof that the business could land a hyperscaler-scale contract.

Act 3 (Sep 2025 onward): the Microsoft deal repriced the entire neocloud trade

On September 9th, 2025, Nebius announced its $17.4B-$19.4B multi-year capacity agreement with Microsoft. NBIS gapped from $64.06 on September 8th to a close of $95.72 on September 9th, a +49% single day move on 88.4 million shares of volume, four times the average daily volume. CRWV moved +7% the same day, from $93.55 to $100.22. That was the moment the market repriced what Nebius could be: an owned-and-operated alternative at hyperscaler scale.

The two stocks then traveled inverse paths through the back half of 2025 and early 2026. NBIS continued to climb through September (peak around $112), gave back some ground on lighter Q3 revenue and the $3B Meta deal on November 12th, 2025 (NBIS fell from $109.95 on November 10 to $85.43 on November 12th), then consolidated in the $80-$110 range through April 2026. CRWV traded sideways to down: $97 in late February after Q4 earnings, $74.41 on March 9th, around the time of the $21B Meta-CoreWeave deal, and $77.47 at March 31st, 2026.

The inflection came on May 13th, 2026, when Nebius reported Q1 2026 with 684% YoY revenue growth, a $27B Meta contract, a $2B NVIDIA equity investment, and Core AI EBITDA margin doubling to 45%. NBIS closed at $207.27 (up +16% from the prior day’s $179.11) and hit an intraday all-time high of $233.73 on May 14. CRWV reported Q1 2026 earnings the prior week and closed at $128.84 on May 7th, drifting to $107.30 by May 15th as the market focused on the widening adjusted operating income/debt picture rather than the headline $99B backlog.

Detailed price history at key inflection points

Returns and the crossover

The summary statistics make the divergence stark:

The valuation crossover happened in late March/early April 2026. CoreWeave traded down post Q4 2025 earnings and the Meta $21B announcement (which the market interpreted as backlog-good, debt-bad) while Nebius held a $90-$110 range that gave it a market cap in the high $20s of billions, then ran sharply on the May 13th Q1 2026 print. From the IPO benchmark, NBIS has returned more than five times what CRWV has. The two now sit within 6% of each other on market cap despite a 5:1 revenue gap.

Bottom line

The two companies looked similar in 2024; the financial profiles have diverged significantly since. CoreWeave is the faster-scaling, higher-revenue operator, but leveraged and colocation-dependent — EBITDA margins are at 56% but the adjusted operating margin has moved from 19% to 1% as depreciation ramps. Nebius is smaller but growing six times faster, with a net cash balance sheet, owned facilities driving structural margin upside, Core AI EBITDA margin expanding to 40%, and a non-GAAP framework that does not yet include adjusted OI.

The valuation gap has closed with the market pricing NBIS at 17x revenue versus 6.5x for CRWV, reflecting the difference in balance sheet composition and margin trajectory between the two companies. Nebius’s multiple depends on converting 4 GW of contracted power into the $7B-$9B exit ARR it has guided to. CoreWeave’s multiple depends on contractually locked-in backlog converting fast enough to deleverage and expand adjusted operating margin before the next refinancing window