Artificial intelligence (AI) infrastructure startup Nscale is hurtling headfirst into the global AI boom, but severe grid constraints and construction setbacks are raising questions about the company’s ambitious roadmap ahead of a $35 billion IPO. listing.
The UK-based developer is preparing for a public listing after raising approximately $3.36 billion in pre-IPO funding. Despite backing from NVIDIA Corp., which provided $1 billion in financing to secure demand for its own processors, Nscale’s foundational projects are confronting mounting physical and structural operational hurdles.
Just north of the Arctic Circle in Glomfjord, Norway, Nscale’s flagship European site is lagging six months behind schedule.
Buildout of the planned 30-megawatt (MW) data center has been hampered by cooling system leaks, regulatory hurdles, general construction delays, and an on-site fire. Currently, only 6.4 MW of capacity is operational.
To accelerate work on the fjord’s edge, developers recently towed in the Kalmar, a three-story floating hotel featuring 220 bedrooms to house construction crews arriving.
Simultaneously, in Loughton, Essex, Nscale’s headline domestic initiative — promoted as the UK’s largest AI supercomputer — faces severe energy bottlenecks.
Built in tandem with NVIDIA and meant to launch by 2027, the site requires 90 MW of power, an amount equivalent to the energy consumption of roughly 315,000 homes. However, the site was operating as a scaffolding yard as recently as March, and connection delays have forced network operators to address the site’s immediate viability.
The power deadlock in Essex underscores a broader crisis facing the UK energy grid.
Regulators at Ofgem revealed in June that 315 data centers were stuck in connection queues, representing a combined demand of 73 gigawatts (GW). The backlog exceeds the entire national power capacity of 45 GW by 62%. With grid capacity strained, around 100 data center developments across the country are reportedly considering constructing their own gas-fired power stations to bypass National Grid bottlenecks.
Nscale’s strategy reflects a wider shift among major AI hyperscalers and research labs, which increasingly rent infrastructure from third parties to convert long-term capital risks into flexible operating expenses.
However, financial analysts note that this model relies heavily on debt and complex balance-sheet support mechanisms. Rather than eliminating the risk of overcapacity, the burden has largely transferred to smaller, highly leveraged neocloud providers like Nscale.
As Nscale moves toward its $35 billion public debut, prospective investors face a central question: whether the company’s reported $103 billion in contracts represents a long-term revenue annuity or merely a temporary windfall driven by current computing shortages.
With assets built largely on credit, Nscale’s long-term valuation remains closely tied to the endurance of the ongoing AI expansion cycle.
