Is Nebius Profitable? Positive Adjusted EBITDA, Heavy Capex, and AI Cloud Economics
Nebius has reached strongly positive adjusted EBITDA in its AI cloud, but adjusted net income remains negative and expansion requires billions in capital spending.
Nebius has crossed into positive adjusted EBITDA while remaining negative on adjusted net income
Nebius offers one of the clearest examples of why AI infrastructure profitability requires multiple measures. In Q2 2026 the group reported $582.3 million of revenue and $236.2 million of adjusted EBITDA, while adjusted net income remained a $33.2 million loss and net income from continuing operations was a $190.4 million loss for the quarter.[1] The company had therefore reached strong pre-interest, pre-depreciation operating performance without reaching a clean recurring bottom-line profit. Calling Nebius simply profitable or unprofitable would hide the real transition underway.
The cloud operation is stronger than the consolidated bottom line
Nebius said its core AI-cloud business generated roughly $285.7 million of adjusted EBITDA in Q2, with a margin near 50%.[2] That is meaningful evidence that deployed capacity can generate attractive operating cash economics before depreciation and financing.
Revenue growth has been extraordinarily fast
Q2 group revenue grew 454% year over year, and first-half revenue grew more than fivefold.[1] Rapid growth can create operating leverage because engineering and corporate costs are spread across more customer consumption. It can also force the company to build faster than internal cash generation supports. Nebius is doing both: improving adjusted EBITDA while committing enormous amounts of capital to additional GPU and data-center capacity.
Operating leverage and capital intensity can coexist
A cloud business may become profitable at the EBITDA layer even while free cash flow remains deeply negative because every new dollar of future revenue requires hardware and facilities today.
The capex number is larger than the revenue number by design
Nebius reported about $5.7 billion of capital expenditures in Q2 2026, primarily for GPUs and data-center expansion.[3] That figure dwarfed quarterly revenue because the company is installing capacity intended to support future contracts. This is the defining feature of neocloud economics: the income statement can improve at the same time cash is pouring into long-lived assets. Investors must therefore evaluate both operating margin and the return expected on each new generation of capacity.
Capex is not automatically a loss
Buying a GPU does not immediately flow through the income statement as an expense. The asset is capitalized and depreciated over time, which is why EBITDA can look very strong during an aggressive buildout.
Customer prepayments are becoming an important financing tool
Nebius said it expected more than $9 billion of customer prepayments during 2026 and described a financing framework in which upfront payments, debt, and contracted cash flows help fund expansion.[3] Prepayments reduce the amount of corporate equity needed to build capacity and partially transfer financing risk to customers. They also improve visibility because buyers have committed capital before all service is delivered.
Prepayments can make infrastructure self-funding
When a customer finances a meaningful share of construction or hardware, the provider’s return on invested equity can improve dramatically—assuming it executes the contract successfully.
Adjusted EBITDA should not be confused with free cash flow
Nebius’s Q2 adjusted EBITDA was strongly positive, but the company was simultaneously spending billions on infrastructure.[2] EBITDA excludes depreciation and does not subtract capital expenditures, so it is best understood as a measure of operating performance before the full cost of asset ownership. In a capital-light software company that distinction may be modest; in an AI cloud it can be enormous.
GAAP net income is unusually noisy for Nebius
Nebius reported positive net income over the first six months of 2026 even though Q2 continuing operations produced a loss, partly because investment and other non-operating items can move GAAP results.[1] For that reason, adjusted net loss and cloud adjusted EBITDA may provide a clearer view of operating progress than headline net income alone. Profitability analysis must look through one-time gains and portfolio effects.
Nebius is profitable at the AI-cloud EBITDA layer, not yet at the recurring net-income layer
As of September 2026, the strongest precise statement is that Nebius’s AI-cloud operation is generating positive adjusted EBITDA with a very high reported margin, while the group still reports an adjusted net loss.[3] That places it further along the profitability curve than many neocloud peers. It has demonstrated that deployed capacity can generate operating earnings; the remaining test is whether those earnings ultimately cover depreciation, financing, corporate costs, and the next wave of capacity.
Why Nebius matters to AI profitability history
Nebius makes the economics of the AI infrastructure boom unusually visible because it is public and publishes both operating metrics and capital requirements. The company moved from negative adjusted EBITDA in 2025 to hundreds of millions of positive adjusted EBITDA in 2026 while revenue multiplied.[4] At the same time, its investment program remains enormous. Nebius therefore illustrates the central paradox of AI infrastructure: a cloud can have attractive unit economics and still consume huge amounts of cash because growth requires physical assets. Nebius continues to publish its quarterly materials through an investor hub that makes this capital-intensive transition unusually observable.[5] Its path will help define when the market should call an AI cloud truly profitable.
Works Cited
- 01
- 02
- 03FT Markets — Nebius Q2 2026 Financial Results markets.ft.markitdigital.com
- 04
- 05Nebius Investor Hub nebius.com
CodeHistory is a living archive. Citations document the evidence used for this edition; later evidence may refine the account.
Submit a research lead