When the CEO of Nvidia publicly tells you to pick up the pace, it is probably not a casual suggestion.
At Nvidia’s GTC event in Berlin on October 21-22, 2026, Jensen Huang urged UK-based AI infrastructure company Nscale to “work around the clock” to fix delays and equipment issues at its data centers. The timing is awkward, since Nscale is preparing for a NYSE IPO at a targeted $30 billion valuation.
A fast climb meets the physical world
Nscale’s rise has been unusually quick, even by AI-boom standards. The company was founded in 2024 as a spinout of Arkon Energy, and it secured £500 million from Nvidia in 2025.
In the two years since launch, Nscale has built up over $100 billion in contracted revenue, according to the research findings. It has also signed partnerships with Microsoft and Anthropic, two of the most closely watched names in artificial intelligence.
The clearest example of that gap is in Loughton, Essex. Microsoft has withdrawn from a £2 billion UK project there with Nscale after delays in power grid connections pushed the site’s availability into the 2030s.
That project would have housed 23,040 Nvidia GB300 GPUs.
Losing a client of Microsoft’s stature on a project of that size is the kind of development prospective IPO investors tend to read twice.
Expansion continues elsewhere
The Essex setback has not stopped Nscale from pushing ahead in other markets. The company is under contract for approximately 200,000 GPUs across sites in the US and Europe.
It is also reportedly moving forward with plans for a major data center campus in West Virginia, known as Monarch. Nscale is in advanced talks with Anthropic for additional capacity deals at that site, though talks are not signed contracts.
Part of the company’s strategy relies on Nvidia’s DSX modular AI factory technology, with sites planned across multiple countries using that approach.
Why Huang’s words carry weight
Nvidia is not a neutral observer here. It is an investor in Nscale, a technology supplier to Nscale, and arguably the single biggest beneficiary when companies like Nscale turn GPU orders into running capacity.
It also highlights a broader industry problem that the research findings point to directly: concerns over power availability and deployment timelines. Those factors are increasingly seen as central to whether the next generation of AI infrastructure arrives on schedule.
What this means for the IPO
For anyone weighing Nscale’s planned listing, the central question is simple to state and hard to answer. Can a company with over $100 billion in contracted revenue convert those commitments into operating data centers on time?
The research findings describe the $30 billion IPO ambition as currently under scrutiny because of execution challenges and investor skepticism. Huang’s remarks are unlikely to make that scrutiny any lighter.
There is also a dependency question. Nscale leans heavily on Nvidia for chips, funding and now modular data center design.
The Anthropic relationship is another thread worth watching. Advanced talks over West Virginia capacity could help offset the Microsoft withdrawal in Essex, but until any deal is finalized, it remains a possibility rather than booked business.
Investors will likely focus on a few practical signals ahead of the listing. Those include progress on grid connections, whether the West Virginia campus advances as planned, and how many of the approximately 200,000 contracted GPUs are actually up and running.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.