Stargate UK fails due to electricity costs? OpenAI halts British AI mega-project and flees to Norway
Xpert Pre-Release
Available in 27 languages 📢
Prefer Xpert.Digital on GoogleⓘPublished on: April 11, 2026 / Updated on: April 11, 2026 – Author: Konrad Wolfenstein
Stargate UK fails due to electricity costs? OpenAI halts British AI mega-project and flees to Norway – Image: Xpert.Digital
Too expensive, too little power: OpenAI's 150 billion plan and its dramatic fall
8,000 Nvidia chips without a home: The drama surrounding Europe's largest AI center
In September 2025, the UK celebrated its ultimate breakthrough: the mega-project "Stargate UK" was poised to catapult the country to the global forefront of artificial intelligence with the help of OpenAI and thousands of state-of-the-art Nvidia chips. But just a few months later, a harsh reality check followed. Due to exorbitantly high industrial electricity prices and massive regulatory hurdles in copyright law, OpenAI put its plans on hold. This surprising withdrawal is not only a painful slap in the face for the British government, but also reveals a structural problem that threatens large parts of Europe. While the US and countries like Norway attract billions in investment with cheap energy, pragmatic policies, and gigantic capacities, the old continent risks falling behind in the global AI arms race. An analysis of shattered dreams, escalating infrastructure costs, and the crucial question: Who will truly win the AI race of the decade?
Related to this:
- Comparison of electricity grid expansion: USA, China, EU, Japan, South Korea and Germany at a glance
Stargate UK: When superpower ambitions meet reality checks
In early April 2026, the British government was caught completely off guard. OpenAI, the company behind ChatGPT and a flagship of the global AI revolution, announced it was putting its showcase project on hold in the UK. What had been celebrated as a historic milestone for the British economy in September 2025 now appears to be a symbol of the structural deficiencies that Britain and much of Europe face in the global AI race. The name of the halted project: Stargate UK—a term that promised much and now brought considerable disillusionment.
The dream of a British AI flagship
The announcement of Stargate UK in September 2025 coincided with a period of euphoric optimism. Just a few months earlier, in January 2025, Prime Minister Keir Starmer had presented his AI Opportunities Action Plan, which aimed to catapult the UK to the forefront of artificial intelligence. Fifty measures, all approved by the government, were designed to make the country the preferred investment location for global AI companies. The initiators emphasized that the announcements made so far had already mobilized £25 billion in investment for new data centers. The Stargate announcement, presented as part of a broader £150 billion investment package, was accompanied by a visit from former US President Donald Trump to London in September 2025.
The specific plan was ambitious: OpenAI, together with its partners Nvidia and the British data center specialist Nscale, intended to install up to 8,000 high-performance GPUs, either from the H100 series or the newer Blackwell generation, in the UK. One of the proposed locations was the so-called Cobalt Park in Northumberland in northeast England—a government-designated AI growth zone intended to revitalize economically weaker regions through digital infrastructure. Nscale, the company with former British Deputy Prime Minister Nick Clegg on its advisory board, had been explicitly highlighted by the government as a prime example of how the UK could become a leader in AI infrastructure.
The fall from AI heaven
But the euphoria was short-lived. In April 2026, OpenAI spokespeople told CNBC and Reuters that the company considered the project not feasible for the time being. While they saw great potential for the UK's AI future, they would only move forward when regulatory frameworks and energy costs allowed for sustainable infrastructure investments. A seemingly diplomatic formulation—but one whose directness felt like a slap in the face to the Starmer government.
Two main reasons underlie the decision. First, energy costs in the UK are the highest in Europe for large industrial consumers—by a wide margin. Second, the British government had recently abandoned its controversial plan to allow AI companies largely unrestricted use of copyrighted content for model training, following massive opposition from the creative industries. For OpenAI and similar companies that rely on virtually unlimited training data, this means regulatory uncertainty for an indefinite period.
Related to this:
Electricity as a strategic weapon: Why Britain is losing structurally
The energy price issue is central to understanding the entire debate. For years, the UK has held the dubious distinction of having the highest industrial electricity prices in Europe. Recent government data shows that large industrial consumers in the UK paid 25.33 pence per kilowatt-hour in the first half of 2025—roughly 125 percent above the median of the fourteen EU countries compared. The comparison with individual member states is even more striking: very large industrial consumers in the UK pay 22.39 pence per kWh, while Finland—the cheapest EU country—charges only 4.37 pence. This means that British industrial companies pay more than five times what their Finnish counterparts have to pay.
Historically, this wasn't always the case. As recently as 2008, UK industrial electricity prices were lower than Germany's and only slightly above the EU-14 average. The dramatic deterioration of its competitive position is the result of a structural problem: the UK generates a significantly higher proportion of its electricity from natural gas—around 35 percent in 2023, compared to 16 percent in Germany, 6 percent in France, and 23 percent in Spain. UK industrial electricity prices of 25.4 pence/kWh in the same year were considerably higher than 15.6 pence in Germany, 17.6 pence in France, and 13.3 pence in Spain. Since AI data centers are among the most energy-intensive facilities in modern industry, this disadvantage is magnified astronomically.
To make matters worse, data centers in the UK are not currently recognized as Energy Intensive Industries (EIIs). This has significant consequences: unlike steel or chemical companies, which can be exempted or relieved of network charges and policy costs, data centers bear the full burden of all transmission, network, and policy costs. Without this classification reform, escalating operating costs will permanently deter technology-oriented investors.
The government recognized the problem and in November 2025 presented a package of measures for AI growth zones, promising electricity price reductions of up to £24 per MWh for sites in Scotland, £16 in Cumbria and £14 in the northeast of England — but not until April 2027 at the earliest. For OpenAI, this was obviously too little and too late.
Related to this:
The copyright debacle: Creative industries versus AI industry
The second major obstacle is legal in nature and deserves separate consideration, as it illustrates the structural tension that runs through the entire AI regulation debate in Europe. In December 2024, the British government presented a consultation draft that would have allowed AI companies to use copyrighted works for model training—provided rights holders did not actively object (a so-called opt-out model). OpenAI had publicly championed this approach, arguing before the British House of Lords that without access to copyrighted material, training competitive AI models would be simply impossible.
But the UK's creative industries—a sector that claims to contribute more than £120 billion annually to the economy—responded with a coordinated counter-campaign. Under the banner of "Make it Fair," the industry mobilized broad public support, with front-page and homepage takeovers of national media outlets on the final day of the consultation period in February 2025. This was accompanied by prominent voices from the music industry and the literary world. The message was clear: de facto legalization of scraping creative content without compensation was nothing less than state-sanctioned intellectual property theft.
The government caved. In March 2026, the Department for Culture announced it was shelving the original plan and launching new rounds of consultations. An agreement has thus been postponed indefinitely—likely until at least 2027. What was intended as a pragmatic solution to make the UK an attractive training location for AI models has become a political minefield. The result: maximum legal uncertainty for all involved and a competitive disadvantage that directly impacts the AI industry's decisions.
8,000 chips without a home: The immediate operational disaster
Specifically, the Stargate UK pause means that planned resources worth hundreds of millions of dollars are currently unavailable. The original plan called for leasing up to 8,000 of the latest generation H100 or Blackwell chips as early as the first quarter of 2026. While H100 rental prices have fallen by 64 to 75 percent since the end of 2024 to around $2 per hour, these reduced costs add up to a significant sum when considering the scale of tens of thousands of chips over several months.
In parallel, the data center project by Microsoft and Nscale in Essex suffered a similarly unfortunate fate. Planned as the UK's largest AI data center in Loughton, the project was postponed from 2026 to at least 2027—officially to allow time for Nvidia's latest generation of chips (Vera Rubin NVL72), but also because Loughton Town Council raised fundamental concerns about the capacity of the local power grid and the impact on household electricity prices. The project—with a planned capacity of 50 megawatts—was reclassified as 50 percent larger than originally announced, requiring a completely new planning application.
The situation is telling: Two of the UK's most ambitious AI infrastructure projects are failing not due to a lack of capital or technology partners, but due to fundamental site conditions — excessive energy costs, overloaded power grids, lack of building permits and regulatory uncertainty.
When the global AI boom becomes geographically selective
The withdrawal from the UK doesn't mean OpenAI is slowing down its global ambitions—quite the opposite. The Stargate program in the US has become a monument to industrial policy will: In Abilene, Texas, the first operational data center opened in September 2025, equipped with Oracle Cloud Infrastructure and extensive Nvidia hardware. Five more sites in Texas, New Mexico, and Ohio are under construction, bringing Stargate's planned capacity to nearly 7 gigawatts—with investments exceeding $400 billion over the next three years. The ultimate goal: 10 gigawatts, $500 billion.
The US offers what the UK denies: cheap energy in sparsely populated regions, fast-track permitting processes thanks to deregulation-friendly policies under President Trump, and a regulatory climate that doesn't expose AI companies to copyright disputes. Europe, on the other hand, is retreating to the few locations that offer genuine competitive advantages in a more orderly but also more complex market.
The prime example is Norway. In July 2025, OpenAI announced Stargate Norway, marking its first foray into Europe—strategically, and not coincidentally, in a country with abundant hydropower, minimal local electricity costs, and a cool climate that significantly reduces the energy-intensive cooling of data centers. The data center in Narvik, in northern Norway, is slated to be equipped with 100,000 Nvidia GPUs by the end of 2026 and, with a capacity of 230 megawatts, will be among the largest in Europe. Partners are once again Nscale and the Norwegian energy company Aker, which, together with Nscale, is investing approximately one billion US dollars each in the first project phase. Key features include 100 percent renewable energy, closed-loop direct cooling systems, and the utilization of GPU waste heat for regional low-emission companies. The UK could have offered something similar—but it didn't.
Our global industry and economic expertise in business development, sales and marketing
Our global industry and economic expertise in business development, sales and marketing - Image: Xpert.Digital
Industry focus areas: B2B, digitalization (from AI to XR), mechanical engineering, logistics, renewable energies and industry
More information here:
A thematic hub offering insights and expertise:
- Knowledge platform covering global and regional economies, innovation and industry-specific trends
- A collection of analyses, insights, and background information from our key areas of focus
- A place for expertise and information on current developments in business and technology
- A hub for companies seeking information on markets, digitalization, and industry innovations
Norway as an AI magnet: hydropower, cooling climate, locational advantages
Nvidia: The only inalienable asset in the AI universe
In this globally redistributed competition for data centers, there is precisely one player that structurally benefits from fragmentation: Nvidia. Whether the data center is located in Texas, Norway, Abu Dhabi, or—whenever conditions allow again—in the UK, Nvidia chips are the indispensable foundation of all large-scale AI infrastructure. According to current estimates, the H100 and Blackwell GPUs control more than 80 percent of the market share for AI accelerators in the data center. The data center segment recorded revenue growth of 75 percent year-over-year and sequential growth of 22 percent in early 2026.
The projections are staggering: Cumulative sales of Blackwell and Rubin chips are forecast to reach one trillion US dollars by 2027. Hyperscalers like Microsoft, Google, and Amazon exceed 200 billion US dollars in their combined annual AI infrastructure capacity. Lease prices for H100 chips may have fallen by 64 to 75 percent—a sign of supply growth—but the underlying demand for computing power is not declining; it is simply shifting to locations with better economic parameters.
For investors, the consequence is clear: data center locations are interchangeable—Texas can be replaced by Norway, Essex by Narvik. Chips, however, are not. To understand Nvidia's business, one must understand that it is essentially an infrastructure monopoly: not politically mandated, but created through sheer technological superiority, cemented by the CUDA ecosystem, and rendered virtually unassailable by the capital intensity of alternatives.
Related to this:
The Oracle Dilemma: When Ambition Becomes a Debt Trap
At the other end of the risk spectrum is Oracle—and the example teaches us much about the dangers associated with the current AI infrastructure boom. As the central infrastructure partner for the Stargate program in the US, Oracle has budgeted $50.64 billion in capital expenditures for fiscal year 2026—an increase of nearly 139 percent year over year. This massive expenditure has pushed the company's free cash flow into negative territory, with projections of minus $23.28 billion for the coming year.
To finance these investments, Oracle issued $30 billion in bonds and convertible notes in February 2026, with plans to raise another $50 billion by the end of the year. The credit markets reacted accordingly: The cost of hedging Oracle's debt—measured by credit default swaps—rose to levels last seen during the 2008 financial crisis. The structural risk is obvious: AI data centers are fixed-cost monsters. Once built, the meter keeps running—regardless of whether customers actually use the resources as desired. Depreciation doesn't wait, and interest payments are unaffected by underutilization.
When multiple providers enter the market simultaneously with high levels of debt financing and optimistic utilization assumptions, and demand then grows more slowly than expected—as is common in the enterprise sector—a price war ensues that can be existential for capital-intensive infrastructure providers. For Oracle, this means that despite the strategic prestige of the Stargate project, the stock remains subject to significant fundamental risk.
Europe between the claim to sovereignty and the reality of its economic position
The parallel development of Stargate UK and other European AI infrastructure projects reveals a worrying pattern. Europe has embraced AI sovereignty as a political imperative—the European Commission has announced a Data Centre Energy Efficiency Package for the first quarter of 2026, aiming for carbon-neutral data centers by 2030. Furthermore, the AI Act mandates transparency regarding the energy consumption of general-purpose AI models.
But the structural contradiction remains: In cities like Dublin, Amsterdam, and Frankfurt, permitting bottlenecks and grid congestion are already blocking a fifth of all planned computing capacity. European data centers currently account for around 4 percent of the EU's electricity consumption, with forecasts reaching 150 TWh by 2026—a significant share of the total consumption of some smaller member states. By 2035, European data centers could consume more than 230 TWh annually—double the current level.
Globally, the scale is even more impressive: Worldwide, data centers' electricity consumption reached approximately 415 TWh in 2024, representing about 1.5 percent of total global electricity consumption. In the US alone, data centers consume more than 4 percent of the nation's electricity production. The International Energy Agency (IEA) predicts that data centers will consume around 80 percent more energy by 2026 than in 2022. If Nvidia ships 1.5 million AI servers by 2027, these machines alone could consume between 85 and 134 TWh of electricity annually—a cross-section of the total annual electricity consumption of several European countries.
Against this backdrop, Norway's attractiveness appears not as a coincidence, but as the result of clear geographical and energy policy advantages. Northern Norway, with its access to inexpensive hydropower, cool climate, and low local demand, offers precisely the combination of low operating costs and green energy that the AI industry is increasingly prioritizing. The UK could theoretically develop similar advantages in northern Scotland or other wind power regions—however, the political framework currently prevents this.
What's at stake: The strategic dimension for the United Kingdom
The economic implications of the Stargate UK hiatus cannot be reduced to lost investment volume. It concerns whether Britain can forge a new industrial identity in the global technological arena after Brexit. AI infrastructure is to this decade what steel and coal were to the 19th century: the foundation for economic power, productivity, and geopolitical influence.
Starmer had declared that Britain should be an AI producer, not an AI consumer—an ambitious goal that is structurally unattainable without massive domestic computing capacity. The AI Opportunities Action Plan, which comprises 50 measures and envisions multiplying government-controlled computing capacity by 2030, is hardly feasible without private co-financing from global technology leaders like OpenAI. The government has laid the groundwork with a £1 billion investment of its own—but given the scale required, this is at best a drop in the ocean.
The missed opportunity is real and measurable. In September 2025, as part of the UK's AI strategy, Nscale announced a total investment of US$2.5 billion in UK data center infrastructure over three years. Now, the flagship project is on hold, the Essex counterpart has been postponed to 2027, and Stargate's actual European push is landing in Norway, not Northumberland.
Related to this:
Lessons for Europe: Structural competitiveness as the fate of AI
What lessons can be learned for Europe from the Stargate UK disaster? First, political rhetoric alone does not attract gigawatt investments. Companies like OpenAI, which think and operate in capital cycles of decades, need reliable framework conditions—stable energy prices, clear regulations, and streamlined approval processes. Where these are lacking, capital flows away, regardless of all commitments to national AI strategies.
Secondly, the copyright conflict between the creative industries and the AI industry is not unique to Britain, but a structural European problem. The EU-wide AI regulation, the AI Act, introduced transparency obligations for training data without resolving the fundamental licensing issues. Germany, France, and other major creative nations face the same conflicting objectives as the UK. Those who wait too long risk ensuring that the training centers for global AI models remain permanently located across the Atlantic or in third countries.
Thirdly, energy and grid infrastructure is the underestimated bottleneck of the AI decade. The debate about AI focuses too much on algorithms and model architectures, and too little on the simple but crucial question: Where does the electricity come from? Norway answers with hydropower, the USA with untapped desert regions and a political will to deregulate. Great Britain and large parts of continental Europe struggle with outdated grids, high grid fees, and a regulatory framework that cannot keep pace with the speed of the AI industry.
The IEA expects data centers to require around 80 percent more energy by 2026 than in 2022, with the AI sector alone contributing 20 percentage points to the overall growth in data center electricity consumption. A single planned hyperscale data center in Northumberland could theoretically require 1.1 gigawatts—equivalent to the electricity consumption of one million homes and about one-third the output of the Hinkley Point C nuclear power plant. Building just three such facilities would be equivalent to constructing an additional nuclear power plant—solely for the operation of AI infrastructure.
Who will win the location race of the AI decade?
The stage is set. The Stargate program in the US will continue to gain momentum—with a planned total capacity of 10 gigawatts and an investment volume of 500 billion US dollars, it is the most ambitious industrial project of its kind since the development of the automotive industry in the 20th century. Sites in Texas benefit from cheap electricity, ample space, and a political climate that facilitates rapid infrastructure decisions. In Europe, Norway is becoming an AI powerhouse: a small country with a huge locational advantage—an abundance of clean, affordable energy.
For the UK, the window of opportunity is not yet completely closed. The government has recognized the need for action, and the announced electricity price subsidies for AI growth zones—valid from April 2027—are a step in the right direction. Crucially, the copyright issue will depend on whether it can be resolved in a way that is acceptable to all parties and whether the network infrastructure in potentially attractive regions of the north and Scotland is expanded in a timely manner. If this succeeds, Stargate UK could return to the agenda at a later date.
If this fails, the UK risks becoming a permanent footnote on the AI infrastructure map—while Nvidia chips train in Abilene, Narvik, and Abu Dhabi, and Britain's most ambitious AI strategist, Keir Starmer, continues to talk about a Britain that produces AI instead of merely consuming it. The difference between vision and reality is easy to identify in this case: it's energy prices and legal certainty. Both are politically determined—and both can be changed politically. The only question is whether quickly enough.
Related to this:
Your global marketing and business development partner
☑️ Our business language is English or German
☑️ NEW: Correspondence in your native language!
Konrad Wolfenstein
I and my team are happy to be available to you as your personal advisor.
You can contact me by filling out the contact form here or simply call me at +49 7348 4088 965. My email address is : [email protected]
I'm looking forward to our joint project.
☑️ SME support in strategy, consulting, planning and implementation
☑️ Creation or realignment of the digital strategy and digitization
☑️ Expansion and optimization of international sales processes
☑️ Global & Digital B2B trading platforms
☑️ Pioneer Business Development / Marketing / PR / Trade Fairs
🎯🎯🎯 Data-driven B2B industry hub as a quasi-in-house solution
The quasi-in-house solution: How Xpert.Digital closes operational gaps in B2B marketing and sales – Smart Content-Driven Business - Image: Xpert.Digital
Xpert.Digital is a data-driven B2B industry hub led by Konrad Wolfenstein . The company acts as an external, quasi-in-house solution for industrial partners, closing operational gaps in marketing, content, and sales – without requiring additional resources on the client side.
More information here:
