Neocloud Q2 Results Show AI Revenue Growth Running Beside Bigger Power Bets
CoreWeave, Nebius and Cerebras all posted faster Q2 2026 revenue growth, but their results also showed larger losses, heavy capex and rising contracted power commitments as AI cloud demand drives capacity expansion.
AI cloud providers are turning revenue growth into larger power commitments rather than lower losses, with CoreWeave, Nebius and Cerebras all posting faster Q2 2026 sales while capacity spending kept pressure on earnings, Data Center Dynamics reported.
The pattern is clearest at CoreWeave.
Revenue rose to $2.575 billion from $1.212 billion a year earlier, but the company moved to a $49 million operating loss from $19 million in operating income and recorded a $626 million net loss, up from a $290 million loss in Q2 2025.
CoreWeave Adds 500MW In One Quarter
CoreWeave's adjusted EBITDA reached $1.51 billion, compared with $752 million a year earlier.
Capital expenditure in the quarter totaled $9.4 billion, and the full-year capex range is expected to reach $35 billion to $39 billion.
The buildout expanded active power by nearly 500MW during the three-month period, including 300MW in June, taking CoreWeave to 1.5GW across 51 active data centres.
Eight facilities joined the footprint in 2026, with live capacity expected to reach 1.85GW by year-end.
The company's contracted power stood at 4.2GW, and its 2030 target is 8GW.
Revenue backlog was approximately $104 billion as of June 30, before more than $25 billion of net new customer commitments in early Q3.
CoreWeave chief executive Michael Intrator framed the quarter as an inflection point as scale began to translate into operating leverage.
The company also pointed to continuing demand for older Nvidia A100 GPUs, including a contract that extends into 2029 at what executives called an attractive price.
Nebius Revenue Jumps 454 Percent
Nebius posted $582.3 million in quarterly revenue, a 454 percent increase from Q2 2025.
Almost all of the group's revenue came from its AI cloud business, where revenue rose 514 percent year on year to $575 million.
Adjusted profitability moved in the opposite direction from GAAP net income: adjusted EBITDA reached $236.2 million after a $21 million loss a year earlier, and adjusted net loss narrowed to $33.2 million from $91.5 million.
The GAAP bottom line moved from $584.4 million in net income to a $190.4 million net loss, with the prior-year figure affected by gains from revaluing equity securities.
The expansion bill grew alongside revenue.
Total operating costs and expenses rose from $216.3 million in Q2 2025 to $758.2 million in Q2 2026.
New deployment plans covered Vantage in Wales, Greenergy in Estonia and a second self-built facility at the company's campus in Mäntsälä, Finland.
A shareholder letter from Nebius founder and chief executive Arkady Volozh put four quarterly deals above $1 billion each at a yield of $20 million to $25 million per megawatt.
Another 1GW of contracted capacity came through during the quarter, and contracted power is expected to reach 5GW by the end of 2026.
Cerebras Cloud Revenue Nearly Quadruples
Cerebras delivered 74 percent year-on-year GAAP revenue growth to $180.1 million.
Cloud and other services revenue reached $126.0 million, up 281 percent year on year, while core revenue more than doubled to $210 million.
The chip developer's inference cloud business nearly quadrupled year on year, but the company posted a $450.4 million net loss, compared with $309.5 million in net income a year earlier.
On an adjusted EBITDA basis, the quarter remained negative at $53.1 million, deeper than the $38.3 million loss in Q2 2025.
Cerebras chief executive Andrew Feldman tied demand for fast inference to the company's push for more data centre capacity, expanded manufacturing and growth with customers and partners including OpenAI, AWS, AMD and CrowdStrike.
Live and contracted data centre capacity for delivery by the end of 2027 rose to more than 600MW, with the broader pipeline in the gigawatts.
For AI infrastructure buyers, the quarter shows the operating trade-off behind rapid neocloud expansion: revenue and backlog are rising, but contracted power, capex and losses remain the main proof points for whether demand can be converted into durable compute capacity.
