Why Power and Land, Not Chips, Are the New Bottleneck
That shift matters because it reframes the entire AI supply-chain conversation. For years, the story was about GPU scarcity. Now, according to Huang, the harder problem is physical: finding available power grids, zoning suitable land, and getting facilities operational fast enough to match demand. Nvidia itself illustrates the scale of that pressure — the company anticipates roughly 70% revenue growth year-over-year, yet customer demand is running above 100% growth. That gap is exactly where neocloud providers like CoreWeave and Nebius step in, absorbing infrastructure work that traditional cloud giants simply can’t scale quickly enough to match.
CoreWeave’s Growth and Financial Challenges
CoreWeave is growing at a pace few companies can match, but its balance sheet tells a more complicated story. The Nvidia-backed cloud provider is scaling revenue and backlog aggressively while also taking on substantial debt to fund the buildout — a tension that sits at the center of its investment case.
Revenue Backlog Hits $104 Billion
CoreWeave reported second-quarter revenue of $2.58 billion, a sharp jump from $1.21 billion a year earlier. More striking is its backlog, which swelled to $104 billion — a 246% annual increase that signals long-term contracted demand well beyond current capacity. Nvidia’s confidence in the company isn’t just rhetorical: in January, it invested $2 billion in CoreWeave at $87.20 per share, part of a broader partnership that includes plans to jointly develop more than 5 gigawatts of AI infrastructure facilities by the end of the decade.
Debt and Capex Risks Mount
The growth hasn’t come cheap. CoreWeave posted a $626 million net loss in the second quarter and was carrying roughly $33.8 billion in total debt as of June 30. On top of that, the company raised its 2026 capital expenditure guidance to a range of $35 billion to $39 billion — a figure that underscores just how capital-intensive building out AI data centers has become. CoreWeave shares have still climbed about 18% year-to-date in 2026, though that gain trails well behind its closest peer.
Nebius’s Rapid Expansion and Profitability Gains
Nebius is winning the growth-versus-profitability comparison right now, posting numbers that stand out even in a sector full of triple-digit growth rates. Its second-quarter revenue jumped 454% year-over-year to $582 million, while its AI-cloud adjusted EBITDA margin reached 50% — a level that suggests the company is scaling more efficiently than many of its rivals. Contracted power capacity guidance also topped 4 gigawatts, putting Nebius on a similar infrastructure trajectory to CoreWeave, just with a leaner cost structure so far.
Meta and Microsoft Deals Add Scale
Nvidia backed Nebius with a $2 billion investment in March through a pre-funded warrant agreement covering 21.1 million shares. According to SEC disclosures, Nebius also holds a potential five-year, $27 billion infrastructure agreement with Meta Platforms, along with GPU-cloud commitments from Microsoft worth up to $17.4 billion through 2031. Those contracts, if fully realized, would place Nebius among the more heavily contracted independent AI infrastructure providers in the market. During the same quarter, Nebius cut its adjusted net loss by 64% to $33.2 million, and its stock has surged 146% in 2026 — a performance that has made it one of the standout names in the neocloud trade.
Neocloud AI Infrastructure Market Outlook
The broader neocloud infrastructure sector generated $25 billion in revenue during 2025 and is projected to expand to $400 billion by 2031, according to Synergy Research Group — a 58% compound annual growth rate that would make it one of the fastest-growing corners of enterprise technology. Separately, Bank of America research found that the combined backlog from Microsoft, Oracle, Amazon, and Google reached $2.3 trillion by the end of the second quarter, a figure that puts the scale of AI infrastructure demand in perspective even against the smaller neocloud players.
What emerges from CoreWeave’s and Nebius’s diverging financial profiles is less a rivalry than a division of strengths. CoreWeave offers more backlog transparency and sheer scale, while Nebius shows faster growth and better near-term profitability metrics. For Nvidia, backing both isn’t a hedge so much as an acknowledgment that no single neocloud provider can absorb the infrastructure gap alone. As demand for compute keeps outpacing what traditional hyperscalers can build, the companies willing — and financially able — to secure land, power, and shell space first are likely to keep capturing outsized attention from both Nvidia and Wall Street.
FAQ
Why did Nvidia CEO Jensen Huang highlight CoreWeave and Nebius?
He identified them as key AI cloud infrastructure providers addressing critical challenges like securing land, power, and data center facilities that traditional cloud service providers have already exhausted.
What is the main bottleneck to AI infrastructure expansion according to Nvidia?
The main bottleneck is securing physical infrastructure such as land and power capacity, not semiconductor availability, as Nvidia anticipates demand growth exceeding its own 70% revenue growth forecast.
How is CoreWeave performing financially in Q2 2026?
CoreWeave reported $2.58 billion in Q2 revenue with a $626 million net loss and carries $33.8 billion in debt, alongside a revenue backlog of $104 billion.
What major agreements does Nebius have with leading tech companies?
Nebius has a potential $27 billion infrastructure agreement with Meta Platforms and a $17.4 billion GPU-cloud deal with Microsoft extending through 2031.
Article produced with the assistance of artificial intelligence and reviewed by the editorial team.
