Nscale’s $35 Billion IPO Is Wobbling Because Its Data Centers Aren’t Ready
Nscale’s Narvik, Norway site, built with Aker and backed by Nvidia and Point72, runs at 6.4 of a planned 30 megawatts, while net losses hit $1.02 billion in six months.
loomberg reported on October 15 that Nscale’s flagship Norway data center is six months behind schedule, with only 6.4 of its planned 30 megawatts actually running. For a company trying to go public on the strength of AI infrastructure it hasn’t finished building, that gap is the whole story.
Nscale filed to go public on the New York Stock Exchange last month under the ticker NSCL, targeting a valuation of roughly $35 billion. According to Bloomberg’s October 15 report, the UK-based GPU cloud provider’s marquee project, a data center in Narvik, Norway, built with Norwegian energy group Aker, has been hit by cooling system leaks, local permitting fights, construction delays and a fire. Workers are being housed on a floating hotel just to keep the build moving. Only 6.4 megawatts of the planned 30-megawatt facility is actually live.
That’s not a footnote. Nscale’s entire pitch to investors rests on the idea that it can turn signed contracts into running GPUs fast enough to justify a $35 billion price tag. Microsoft and Anthropic are its anchor customers, with multi-year deals together worth close to $88 billion, and the company says it has about $51 billion in total contracted revenue. But a contract is a promise, not a server rack, and Narvik shows how far the promise can drift from the concrete.
The numbers underneath the growth story are just as jarring. Nscale’s revenue for the six months ended June 30 came in at $140.6 million, up sharply from $10.4 million a year earlier. Net losses over the same period jumped to $1.02 billion, from $369 million the year before. Goldman Sachs and JPMorgan are reportedly working the deal, and Nscale is said to be separately raising up to $3.5 billion in pre-IPO financing. None of that changes the fact that the company is burning through roughly a billion dollars while its biggest site runs at a fifth of capacity.
Nscale isn’t the first neocloud to learn that Wall Street has started pricing AI infrastructure risk with a sharper pencil. CoreWeave’s five-year credit-default-swap spread hit roughly 881 basis points in December 2025 before easing to about 452 basis points more recently, a move that at its peak implied close to a 50% chance of default within five years. CoreWeave’s total debt has climbed to $35.6 billion, and its 2031 unsecured bonds yield in the low teens. Interest expense alone ate more than a quarter of its revenue last quarter. That’s the pattern now.
Get maximum flexibility with our wide selection of GPUs. These Cloud GPUs are optimised for a variety of cloud environments, including the Managed Kubernetes Service, AI, and machine learning solutions.
Transparent and competitive prices
Retain control over your budget with transparent, predictable pricing. Get accurate budget estimates for your AI projects with a competitive and flexible pricing.
Sustainability and efficiency
Achieve your environmental goals with our sustainable development policies. Our Cloud GPU instances are built with energy-efficient technologies that limit your carbon footprint.
Data security and compliance
Our commitment to compliance ensures the safety and privacy of your data. Rest assured, your sensitive data is protected in accordance with regulations in force.
SPECIFICATIONS
Key featuresGPU1–4 GPUs per instance. Public networkUp to 25 Gbps included.High-performance storageNVMe storage on most instances.Private networkUp to 25 Gbps included.AutomationManaged via the OVHcloud Control Panel, API or command line.Secure and PrivateCertifications: ISO27001, SOC, Health data.
Use case examples
Generative AI inference for chatbots
Get the computational power you need to quickly train advanced models. Cloud GPUs enable the parallel processing of complex calculations, which significantly reduces training time for deep neural networks, and facilitates efficient analysis of datasets. This capability enables faster innovation and the utilisation of AI solutions in sectors such as healthcare, finance, and technology. Supercharge your chatbot training with NVIDIA H100 GPUs, built to drive the growth of complex language models, enabling faster and more sophisticated responses.
Deep learning applied to computer vision
Edit and process your videos or streaming files, and then livestream your content. Cloud GPUs speed up video encoding, decoding and processing, for more extensive manipulation and efficient distribution on various platforms. The use of this technology is vital for content creators, broadcasters, and streaming services. Leverage high-performance GPUs for computer vision to efficiently process workloads. GPUs are designed to accurately interpret real-time monitoring data.
Next-generation graphics rendering
By using cloud-based applications for graphics-intensive tasks, you can eliminate the need for costly hardware. These workstations offer first-rate performance for graphic design, 3D modelling and simulation, and also support remote mobility and collaboration. Access graphics-intensive resources anytime, anywhere. Take your 3D projects to the next level with NVIDIA L40S GPUs, designed for creators seeking fast iteration and graphic rendering.
n a landmark partnership, Stargate Norway plans to deliver renewable-powered, sovereign AI infrastructure, marking OpenAI’s first gigafactory initiative in Europe
Oslo, Norway – 31 July 2025 – Nscale Global Holdings Ltd. (“Nscale”), Aker ASA (“Aker”) and OpenAI today announced the launch of Stargate Norway: an advanced AI infrastructure project in Narvik, Northern Norway. Powered entirely by renewable energy, the facility will target to deliver 100,000 NVIDIA GPUs by the end of 2026, with ambitions to scale significantly in the years ahead. Stargate Norway will be designed to provide secure, scalable, and energy-efficient infrastructure for sovereign AI workloads across Europe, fully aligned with European regulatory frameworks. The initiative is led by global AI infrastructure provider Nscale, in partnership with Aker and OpenAI.
Stargate Norway is planned to deliver 230MW of capacity, with ambitions to expand by an additional 290MW – positioning it among the most significant AI infrastructure investments in Europe. The project marks OpenAI’s first data centre initiative on the continent and the inaugural European site under its “OpenAI for Countries” programme.
The facility will be located in Kvandal, just outside Narvik – a region characterised by abundant hydropower, low local electricity demand, and limited transmission capacity. These conditions result in power prices well below the European average, complemented by a cool climate and established industrial infrastructure, making Narvik an ideal location for large-scale data operations.
Stargate Norway is projected to generate significant activity and a long-term operational presence in the Narvik region. The project will also collaborate with local and regional academic institutions to advance AI research and development, supporting the growth of regional expertise and innovation in Northern Norway.
“Announcing Stargate Norway and delivering one of the first European AI Gigafactory to market is a strategic milestone for the region and boosts its role in the global AI landscape,” said Josh Payne, CEO of Nscale. “Sovereign, scalable and sustainable infrastructure is now essential to remain competitive. We’re proud to partner with Aker, drawing on the team’s extensive experience and impressive track record, to deliver a new generation of AI economic and productivity growth in Europe. Together, we are combining OpenAI’s leading models with Nscale’s vertically integrated AI cloud, to provide the sustainable infrastructure necessary to support public benefit, build industrial resilience and enable long-term regional innovation.”
The site will be designed and built by Nscale, leveraging its experience in delivering sovereign-grade AI infrastructure across Europe. It will be owned by a 50/50 joint venture between Nscale and Aker, to be formed through wholly owned subsidiaries of Aker following the planned merger with Aker Horizons. Completion of the joint venture is subject to agreed-upon closing conditions.
In parallel, Nscale and Aker will begin preparatory steps and mobilise resources for construction activities. Both parties have committed approximately USD 1 billion to the initial 20MW phase of the project, including upwards of USD 250 million in equity on a 100% basis, where each partner will contribute equally. The Kvandal site alone has the potential to scale tenfold in future phases as customer contracts are secured. OpenAI welcomes the opportunity to be an initial offtaker with the option to scale over time under its “OpenAI for Countries” programme.
The UK’s AI Growth Zones and What’s Actually Being Built NVIDIA, Nscale, CoreWeave, and Microsoft are building the UK’s next generation of AI infrastructure: 120,000 Blackwell Ultra GPUs, backed by up to £11 billion in investment, targeted to be operational by the end of 2026, according to NVIDIA’s own announcement. That makes it the largest single AI infrastructure rollout in the country’s history.
“The United Kingdom is building the infrastructure for the AI industrial revolution, advancing science, transforming industries and creating new economic opportunities,” said Jensen Huang, NVIDIA’s CEO, in that announcement. Sam Altman added: “Stargate U.K. builds on this foundation to help accelerate scientific breakthroughs, improve productivity and drive economic growth.”
The Five Designated AI Growth Zones The UK government has designated five AI Growth Zones so far, sites picked for access to at least 500MW of power and streamlined planning support:
Zone Location Designated Anchor Partner Culham Oxfordshire April 2025 UK Atomic Energy Authority campus Cobalt Park North East England September 2025 Nscale, part of Stargate UK North Wales Anglesey and Trawsfynydd November 2025 Former nuclear site redevelopment South Wales Bridgend (former Ford plant) November 2025 Site conversion for AI data centers North Lanarkshire Airdrie, Scotland January 2026 DataVita and CoreWeave Collectively, the first cohort has secured £28.2 billion in private investment and is projected to create more than 15,000 jobs, per a tracker of the program.
The North Lanarkshire zone alone, the newest of the five, is expected to attract $11.2 billion in private investment and generate over 3,400 jobs, according to
Retired, living in the Scottish Borders after living most of my life in cities in England. I can now indulge my interest in all aspects of living close to nature in a wild landscape. I live on what was once the Iapetus Ocean which took millions of years to travel from the Southern Hemisphere to here in the Northern Hemisphere. That set me thinking and questioning and seeking answers.
In 1998 I co-wrote Millennium Countdown (US)/ A Business Guide to the Year 2000 (UK) see https://www.abebooks.co.uk/products/isbn/9780749427917
