When a London-founded startup closes the largest Series C in European history at a $14.6 billion valuation in under two years, the story is rarely only about the money. In Nscale’s case, the money is the prologue.
On March 9, 2026, Nscale Global Holdings raised $2 billion in a Series C led by Aker ASA and 8090 Industries, with participation from Nvidia, Dell, Citadel, Jane Street, Lenovo, Nokia, and Point72. The round valued the company at $14.6 billion, more than four times its reported September 2025 valuation of around $3.1 billion.
Josh Payne, founder and CEO of Nscale, said: “This is leading to the largest infrastructure buildout in human history. Nscale is leading this buildout. We are building this foundation that the market sits on, the engine of superintelligence.” Alongside the raise, the company appointed Sheryl Sandberg, Nick Clegg, and Susan Decker to its board.
Nscale’s anchor customer is Microsoft. Its largest single hyperscale deployment is in Barstow, Texas. Every chip it runs is American.
None of that stops Nscale from being Europe’s most credible private AI infrastructure bet. But it does make “sovereign” the word worth interrogating. The capital and the board signal ambition. Neither settles the question of what, exactly, Nscale is building.
What Nscale Actually Builds: Power, GPU Clusters and AI Compute at Hyperscale
Founded in 2024 by CEO Payne and spun out of Arkon Energy, an Australian crypto mining operation, Nscale describes itself as a vertically integrated AI infrastructure provider, a neocloud. That covers a company that controls the full stack: power access, data centre development, GPU superclusters (large clusters of AI chips for training or running AI models) and the software layer through which enterprises access that compute. The model places Nscale above a data centre landlord and below a traditional cloud giant like AWS, which is sometimes called a hyperscaler.
“We are building the AI-native infrastructure platform of tomorrow,” Payne said at the Series B. “AI is reshaping industries, economies, and national strategies – but it cannot happen without the physical backbone: the data centres, the GPUs and the software to orchestrate them.”
Nscale’s footprint spans Norway (an operational site at Glomfjord and planned campuses at Narvik and Kvandal), the UK (Loughton, Slough, and West London), Portugal (Sines Data Campus), Iceland (where the first GPU deployment phase at Verne’s campus is under way), and the United States, with facilities in Texas and North Carolina.
Goldman Sachs and JPMorgan have been engaged as the company weighs a potential IPO in the second half of 2026, according to Reuters.
Its relationship with Nvidia is structural, not incidental. Jensen Huang told a podcast cited by Fortune: “Neoclouds wouldn’t exist without Nvidia’s support. If we didn’t support Nscale, they wouldn’t be where they are today.”
Nscale’s Microsoft Dependence Complicates Its European AI Sovereignty Pitch
Nscale’s sovereignty argument deserves scrutiny, because the commercial picture complicates it.
Microsoft is Nscale’s primary anchor customer across every major region. In October 2025, Nscale announced a multi-year deal to deliver approximately 200,000 Nvidia GB300 GPUs to Microsoft across sites in Europe and the US. This deal was reportedly worth approximately $14 billion, a figure entirely separate from Nscale’s $14.6 billion valuation. Of those, around 104,000 are earmarked for a 240-megawatt campus in Barstow, Texas, leased from Ionic Digital, commencing Q3 2026. That deployment, the largest single GPU tranche, sits in the United States, not Europe.
On the European side, Nscale announced in May 2026 that it would deploy more than 66,000 Nvidia Rubin GPUs for Microsoft at Portugal’s SINES Data Campus, starting late 2027. This builds on the 12,600 Nvidia Blackwell Ultra GPUs already being deployed at the same site for Microsoft.
Jon Tinter, Microsoft’s President of Business Development and Ventures, said the partnership “helps ensure Microsoft customers have access to the advanced AI infrastructure they need as demand continues to grow across Europe.”
The OpenAI Deal That Quietly Became a Microsoft Contract
This brings us to OpenAI. Nscale was announced as the operator of Stargate Norway in July 2025. This is a 230-megawatt gigafactory near Narvik powered by hydropower, with OpenAI as the anchor off-taker under its “OpenAI for Countries” programme.
By April 2026, OpenAI had not concluded an agreement with Nscale, according to CNBC. Microsoft stepped in, renting 30,000 Nvidia Vera Rubin chips from the same site. OpenAI now accesses that capacity through Microsoft. The facility branded as Europe’s first sovereign AI gigafactory became, in commercial practice, another Microsoft infrastructure agreement.
It also raises the independence question. When OpenAI retreated from both its UK and Norway Stargate commitments, the replacement in each case was Microsoft, deepening an existing dependency rather than diversifying it.
Payne told Fortune: “You always have some negative commentators. But every dollar is backed by real revenues.”
Why a Texas Data Centre Isn’t European Sovereign Compute
And then there is the geographic tensions as well.
Sovereign AI is generally understood as infrastructure a region can access, govern, and rely on without full dependence on foreign providers. A 240-megawatt campus in Barstow, Texas, hosted at a site leased from a U.S. Bitcoin miner, filling capacity for an American tech giant, is U.S. compute, not European sovereign compute.
According to Fortune’s reporting, Nvidia guaranteed up to $860.3 million of Nscale’s obligations under that Texas lease in exchange for warrants, an arrangement that accelerated Nscale’s build-out, underscoring how structurally the company’s finances are bound to the US chipmaker, while also deepening its entanglement with the chipmaker’s interests.
CADA & EuroHPC: Europe’s State-Backed Answer to US Cloud Dominance
Nscale’s rise sits inside a broader reckoning. The policy context adds weight to Nscale’s European positioning. On June 3, 2026, the European Commission formally proposed the Cloud and AI Development Act (CADA), a legislative framework to reduce dependence on non-EU cloud providers, triple EU data-centre capacity within five to seven years, and set sovereignty standards for public-sector AI procurement. CADA is a proposal, not law; it requires trilogue before binding obligations apply. The draft notes EU-based providers’ share of the European cloud market has fallen from around 29% in 2017 to approximately 15% by 2022.
Europe’s public AI factory programme is also active. Running alongside private capital is a publicly funded model. For example, Poland’s Gaia AI Factory was inaugurated in Krakow in May 2026, a €70 million EuroHPC project with more than 1,000 GPU accelerators planned for 2027.
These facilities are modest in scale relative to Nscale. But they are state-funded, EU-controlled, and not dependent on a single commercial customer’s renewal decision. That is the policy model CADA is trying to create at scale.
Can European Startups and Enterprises Access Frontier AI Compute Without Going Through Big Tech?
The practical question for European AI companies is whether they can access frontier-grade compute without routing everything through AWS, Azure, or Google Cloud. In principle, Nscale’s European facilities, built for GPU-density workloads, aligned with GDPR and the EU AI Act, located close to renewable power, could provide that.
In practice, if most contracted capacity flows to hyperscaler anchor customers, Nscale functions as a commercial infrastructure provider with a European headquarters, not a structural alternative to US cloud dependency.
“There is increasing demand from enterprises and governments for operational AI,” Payne said in February, “and meeting that need requires infrastructure that is scalable, distributed, and ultimately sovereign.”
The next open question now is whether Nscale’s compute costs fall within range of what startups can afford, or whether those economics accrue primarily to Microsoft and other hyperscalers renting capacity in bulk.
The Trade-Offs Nscale Cannot Shake: Nvidia Chips, Microsoft Demand and IPO Pressure
The accumulation of capital, contracts, and board credibility is real. So are the tensions.
Nscale depends on Nvidia chips it does not manufacture. Its largest anchor customer is a US technology company, its highest-GPU-count near-term deployment is in Texas, and its founding OpenAI partnership dissolved without a commercial agreement. Nscale’s early accounts showed a $24 million loss. Since early 2026 it has added a $1.4 billion GPU-backed loan and a $790 million Norway financing facility. And a public listing, for which Goldman Sachs and JPMorgan are advising, reportedly targeting the second half of 2026, will reinforce pressure to concentrate revenue around its most reliable customers.
None of that makes Nscale’s infrastructure story false. The data centres are being built. The Microsoft contracts are real. The board is substantive. But sovereignty is not a headquarters registration. It is a claim about who controls what, under which law, for which customers, who audits the stack, who stays online when geopolitics turns hostile. Payne’s ambition, to become “the next great AI platform”, is one possible path. Whether the governance, customer independence, and operational continuity required to justify that label follow the capital is the question the next two years will answer.
Nscale Is Europe’s Best AI Infrastructure Bet, But Not Yet Sovereign
Nscale may be Europe’s most credible private-sector AI infrastructure challenger. It has capital, contracts, data centres, hardware partnerships, and a board with genuine operational depth. What it does not yet have is a sovereignty story that survives contact with its customer list, its chip supply agreements, its Texas footprint, and its IPO pressure.
But sovereign AI, the version that CADA and EuroHPC’s AI factories are pointing toward, means more than backed revenue. It means control: over data location, customer mix, chip supply, and operational independence when demand shifts.
Nscale is building the physical infrastructure that European AI needs. Whether it also builds the governance and commercial independence that justify the sovereignty label is the question the $2 billion just bought the right to answer at scale.
Author: Richardson Chinonyerem
See Also:
Is GPT-NL the Netherlands’ Best Shot at AI Independence?
Europe vs US AI Infrastructure: Who’s Winning?
The EU’s Sovereign AI Push: Claiming Tech Independence
Frequently Asked Questions
Nscale is a London-founded AI infrastructure company, established in 2024 and spun out of Arkon Energy. It describes itself as a vertically integrated “neocloud,” controlling the full stack: power access, data-centre development, GPU superclusters, and the software layer enterprises use to access that compute.
In March 2026 Nscale closed a $2 billion Series C at a $14.6 billion valuation, the largest Series C in European history. It was led by Aker ASA and 8090 Industries, with participation from Nvidia, Dell, Citadel, Jane Street, Lenovo, Nokia and Point72.
Nscale is Europe’s most credible private AI infrastructure challenger, but its sovereignty claim is contested. Its anchor customer is Microsoft, every chip it runs is American-made Nvidia hardware, and its largest single deployment sits in Texas. Genuine sovereign AI implies control over data, chip supply, customer mix and operational independence, which Nscale does not yet fully hold.
CADA is a legislative proposal the European Commission adopted on 3 June 2026 to reduce EU reliance on non-EU cloud providers, triple Europe’s data-centre capacity within five to seven years, and set sovereignty standards for public-sector AI procurement. It is a proposal, not yet law, and must pass trilogue before binding.
