Nscale raised $2 billion in its Series C round in March 2026, the largest funding round in European technology history for an AI infrastructure company, at a $14.6 billion valuation. The round was led by Aker ASA and 8090 Industries, with participation from NVIDIA, Dell, Nokia, Citadel, and Jane Street. Two years ago, Nscale was a cryptocurrency mining operation called Arkon Energy. Now it operates data centers across the UK, Norway, Portugal, Iceland, and the United States, and its CEO has said the company plans to go public in 2026.

The round is not an isolated event. It sits inside a broader capital cycle that is reshaping who owns and operates the physical layer of AI. A new class of company, the neocloud, is building vertically integrated GPU factories optimized for AI training and inference, and doing it at a scale that increasingly resembles the hyperscaler buildouts of the 2010s. The difference: these platforms are purpose-built for accelerated compute from day one, with no legacy general-purpose services to maintain.

The Q1 2026 Numbers

CoreWeave, the sector’s largest public company, reported Q1 2026 revenue of $2.08 billion on May 7, up 112% year-over-year and 32% sequentially, according to CNBC. Revenue backlog reached $99.4 billion, up roughly 50% quarter-over-quarter and approximately 4x year-over-year. The company now has 10 clients committed to spending at least $1 billion each, up from a period where 62% of revenue came from Microsoft alone.

The backlog number is the one that matters most. Weighted average contract lengths for new capacity sit at approximately five years. CoreWeave expects 36% of its backlog to convert within 24 months and 75% within four years. That is not speculative demand. It is contracted revenue from Meta, Anthropic, OpenAI, and others who are locking in GPU capacity years in advance because they cannot build it themselves fast enough.

CoreWeave ended Q1 with 3.5 gigawatts of contracted power and surpassed 1 GW of active capacity, per its earnings release. Full-year 2026 revenue guidance stands at $12 to $13 billion, with an annualized run-rate exceeding $30 billion targeted by end of 2027.

The cost of this growth is significant. Net loss widened to $740 million in Q1, up from $315 million a year earlier. Capital expenditure guidance was revised upward to $31 to $35 billion for 2026. CoreWeave carries nearly $25 billion in debt. “We have reached hyperscale,” CEO Mike Intrator told CNBC. The stock dropped 10% after hours on the lighter-than-expected Q2 revenue guidance of $2.45 to $2.6 billion versus analyst consensus of $2.69 billion.

Europe’s Compute Gap

Nscale’s thesis is geographic: Europe needs its own AI compute layer, and the company that builds it first will own a structural position in the continent’s AI economy.

The urgency became concrete in April when OpenAI paused its Stargate UK data centre project, citing electricity prices four times higher than in the United States and unresolved copyright regulation. If the largest AI company in the world cannot make the economics work for building in the UK, someone else has to supply the capacity that European customers and regulators demand stay on European soil.

Nscale is filling that gap. On May 6, the company announced €695 million ($812 million) in new infrastructure investment in Portugal, expanding its partnership with Microsoft at the Start Campus data center site in Sines. The deal will deliver more than 66,000 NVIDIA Rubin GPUs to a facility with 1.2 gigawatts of permitted capacity. Microsoft committed $10 billion to the build-out in October 2025. Nscale supplies the GPU infrastructure; Microsoft provides the customer relationship.

The crypto-to-AI transition is central to this story. Nscale spun out of Arkon Energy in early 2024. Crypto mining companies already owned the two assets AI infrastructure requires: access to large-scale electrical capacity and experience operating GPU-dense computing environments. Morgan Stanley reclassified the entire crypto mining sector as an energy infrastructure play for the AI economy in February 2026. Nscale executed the transition faster than any competitor, raising $3.1 billion across three rounds in less than eighteen months.

Nscale is not alone in Europe. Nebius Group, the AI infrastructure spinout from Yandex, delivered 830% year-over-year growth in its core AI cloud segment in Q4 2025, with annualized run-rate revenue at $1.2 billion by year-end, according to Converge Digest. Analyst consensus anticipates approximately $375 million in Q1 2026 revenue, with results due May 13. Nebius operates across European and global markets as a sovereign AI cloud provider.

The Contract Structure Shift

What distinguishes the neocloud cycle from earlier infrastructure booms is the contract structure. Unlike pay-as-you-go cloud instances, these are five-to-seven-year commitments with upfront prepayments, project-level EBITDA margins above 80%, and customer concentration among the most capital-rich AI companies on earth.

IREN’s $9.7 billion multi-year contract with Microsoft, announced in November 2025, is the template. The agreement covers phased deployment of NVIDIA GB300 GPUs across 200 MW at the Childress campus, with a five-year average term and 20% upfront prepayment. Converge Digest reports an expected contribution of roughly $1.94 billion in annualized recurring revenue at an estimated 85% project EBITDA margin once fully online. IREN is targeting $3.4 billion in AI Cloud annualized run-rate revenue by end of 2026.

Lambda, another neocloud operator, has a 24 MW AI Factory in Kansas City coming online in early 2026, with the entire initial deployment dedicated to a single customer under a multi-year agreement, per Converge Digest. Partnerships with EdgeConneX add more than 30 MW of additional capacity in Chicago and Atlanta.

The pattern is consistent: long-term contracts, hyperscaler anchor tenants, dedicated facilities, and margins that justify the enormous capital outlay.

When Competitors Become Customers

The most striking deal of the cycle came on May 6, when SpaceXAI signed an agreement to lease Colossus 1, its Memphis-based supercomputer, to Anthropic. The cluster contains over 220,000 NVIDIA H100, H200, and GB200 GPUs across approximately 300 MW of capacity. xAI’s announcement confirmed the deal provides SpaceXAI with approximately $6 billion in annual revenue, while Anthropic gains immediate high-performance capacity for Claude Pro and Claude Max subscribers.

Elon Musk, who previously referred to Anthropic as “Misanthropic,” posted on X that he spent time with senior Anthropic team members and “was impressed” with their safety work before approving the lease. SpaceXAI migrated its own training workloads to Colossus 2. The deal also includes Anthropic’s expressed interest in developing orbital AI compute capacity leveraging SpaceX’s launch infrastructure.

This transaction captures the core dynamic of the current moment: compute scarcity is so severe that companies that were ideological rivals six months ago are now signing multi-billion-dollar infrastructure contracts. The neocloud sector exists because frontier AI labs cannot build data centers fast enough to meet their own demand, and they are willing to pay premium rates on five-year commitments to anyone who can deliver GPUs, power, and cooling at scale.

Power as Strategic Moat

The common thread across every neocloud operator is power. Not compute architecture, not software orchestration, not developer experience. Electrical power, measured in megawatts and gigawatts, is the binding constraint.

CoreWeave’s contracted power stands at 3.5 GW with a target of 8 GW by 2030. SpaceXAI’s Colossus complex in Memphis draws approximately 2 GW across three facilities and 555,000 GPUs. Nscale’s Portugal site at Start Campus is permitted for 1.2 GW. IREN is targeting 5 GW globally through its NVIDIA partnership, according to Converge Digest.

Site selection has become a strategic function. Nscale chose Sines, Portugal for its deep-water port, cheaper energy, and fewer grid constraints than the UK. IREN’s Childress campus in Texas offers similar advantages. Lambda’s Kansas City facility prioritizes power availability over proximity to tech hubs.

The neocloud operators who can secure long-term power purchase agreements, build in jurisdictions with permissive grid access, and deploy liquid-cooled GPU clusters at densities that traditional data centers cannot support will define the infrastructure layer of AI for the next decade.

Scale and Risk

The numbers are enormous and so are the risks. CoreWeave’s $25 billion debt load, rising operating expenses (technology and infrastructure costs up 127% in Q1), and concentration among a handful of AI lab customers create real fragility. If any major customer renegotiates or delays, the cascading effect on a company with 80%+ margins but massive fixed costs could be severe.

Nscale’s rapid ascent, from crypto miner to $14.6 billion valuation in 24 months, carries execution risk. Delivering 66,000 Rubin GPUs to a Portuguese data center while simultaneously building out facilities in the UK, Norway, Iceland, and the US requires flawless supply chain management and sustained access to NVIDIA’s most in-demand chips.

The sector is also betting that AI infrastructure demand is structural rather than cyclical. CoreWeave’s $99.4 billion backlog and the five-year average contract length suggest customers agree. But enterprise AI adoption timelines have historically slipped, and the capital deployed today against 2027-2030 demand projections leaves little room for a demand correction.

The Industrial Phase

The AI infrastructure buildout has graduated from pilot projects to industrial program. Individual clusters are measured in hundreds of thousands of GPUs. Power capacity is measured in gigawatts. Revenue backlogs are measured in tens of billions. Nscale’s $2 billion round is a single data point in a cycle where CoreWeave, IREN, Nebius, Lambda, and SpaceXAI are collectively deploying more capital into purpose-built AI factories than the early cloud infrastructure cycle ever required.

For European operators specifically, the window is narrow. If they can build sovereign compute capacity before hyperscalers lock up the demand, they own a structural position that is difficult to displace. If they cannot, Europe remains a customer of American infrastructure companies, paying the margin and accepting the data residency compromises that come with it. Nscale is betting $3.1 billion that they can build faster.