Gigawatts and gridlock: current legal risks for AI data hubs

June 2026 | SPOTLIGHT | SECTOR ANALYSIS

Financier Worldwide Magazine

June 2026 Issue

The accelerating adoption of artificial intelligence (AI) has supercharged demand for data hubs and computing capacity. This explosion in demand has triggered a worldwide race to build, backed by some of the largest capital commitments ever seen.

Hyperscalers including Google, Microsoft, Amazon and Meta are expected to collectively spend more than $650bn in 2026 alone on expanding AI capacity, in what is otherwise estimated to be a $3 trillion AI global infrastructure arms race through to 2030.

The speed and scale of this build-out has created frictions with grids, planning systems, local communities, investors and regulators, leading to delays, some outright suspension of construction laws and even recent drone strikes. Together, this is rewriting laws and norms, many seemingly only just set in place.

Delays, bans and regulation in the US

The US remains the dominant force in global hub development, accounting for the largest share of operational capacity worldwide. According to the International Energy Agency, electricity consumption for all data hubs is projected to exceed 1000TWh by the end of 2026, equivalent to Japan’s entire annual electricity usage.

In the US alone, total hub energy demand is estimated to almost double from 80GW in 2025 to 150GW by 2028, largely driven by AI data workloads. Investment at the federal level has been framed largely as a matter of national competitiveness that has developed in varying directions.

The original Stargate programme, announced in January 2025, described a $500bn commitment to AI infrastructure by OpenAI, Oracle and SoftBank, yet this joint venture effectively stalled amid disputes over control, financing and ownership. Partners have retreated into separate bilateral arrangements, with plans to expand the flagship Abilene, Texas campus subsequently abandoned, even as Texas remains a key location for hyperscalers and data hubs.

Other shifts include Google’s commitment to a more modular approach, including a planned Minnesota hub powered by dedicated solar, wind and battery storage infrastructure in partnership with utility Xcel Energy, a data and power integration model increasingly favoured by hyperscalers seeking to sidestep grid congestion.

That same grid congestion is now a defining feature of the US market. Capacity auction prices in the PJM Interconnection region covering much of the Mid-Atlantic and Midwest rose from about $29 to about $270 per MW‑day in the 2025/26 auction, with the Independent Market Monitor attributing a substantial portion to hub demand.

In response, a significant industry move toward ‘behind the meter’ power generation is underway. A growing share of planned new hub capacity in the US now incorporates onsite or hybrid generation, with some of the largest campuses designed to operate largely independent of the public grid, often on low-emissions energy technology of the kind deployed by Google in Minnesota. On a large and similar scale, a 2800-acre Texas development is being built by an independent hub developer for Anthropic, powered by onsite gas turbine generation.

Despite hyperscalers planning over $650bn in AI spending in 2026, close to half of all planned US hub builds this year are projected to be delayed or cancelled. Known existing constraints include electrical hardware: transformers, switchgear and batteries are in critical short supply.

US manufacturing capacity for this equipment remains insufficient after a decade of reshoring initiatives, and China, still a significant global producer of power-delivery infrastructure, is simultaneously subject to significant tariff pressure and global demand.

In the end, these delays are costly in terms of lost revenue for every month of delay, leading to a variety of disputes and potentially corporate market value mismatches if delay is excessive or a significant site is cancelled.

In terms of regulation, subject to its governor’s approval or veto, Maine became, in April 2026, the first US state to pass a law prohibiting new hub projects consuming 20MW or more until late 2027.

And it is not alone. More than 200 bills addressing hub issues were introduced across the US in 2025, rising to over 300 across 30 states in just the first six weeks of 2026, spanning energy pricing, water usage, environmental standards, zoning and national security. Data Center Watch estimates that $156bn worth of projects were blocked or delayed by local opposition in 2025.

At the federal level, a bill introduced by a senator from the Republican majority proposes to let hub operators build off-grid generation systems free of federal utility regulation to avoid costs being passed to residential consumers.

The broader state regulatory patchwork is itself also under legal attack. Challenges already afoot include a civil rights lawsuit filed in January 2026 by a California hub developer against city officials’ alleging coordinated administrative obstruction, and a federal Justice Department task force established specifically to challenge state AI laws.

Roadblocks in the UK and Europe

The UK ranks as the world’s third-largest hub market. In January 2025, Keir Starmer, the UK prime minister, declared an ambition to make the UK an AI superpower, designating data hubs as “critical national infrastructure” on a par with energy and water systems, and launching “AI ‘growth zones’” intended to fast-track planning and grid access.

The strain on that ambition is already visible. In April 2026, OpenAI indefinitely halted its Stargate UK project citing high energy costs and regulatory uncertainty, though having just completed its own record $122bn fundraising. The underlying tension is structural: the national energy regulator has disclosed that projected peak demand from around 140 planned new hubs would reach 50GW, exceeding Britain’s current entire peak demand of roughly 45GW.

These issues are crystallising in real time. In January 2026, England’s High Court granted permission for judicial review in an early case specifically targeting a UK hyperscale hub on environmental grounds. This came after the government conceded a serious logical error in approving a 90MW facility near London without requiring an environmental impact assessment, having relied on developer mitigation commitments that were not yet legally secured.

The case signals that non-governmental organisation litigation against hub approvals is now an active strategy, and also grid connection disputes are already materialising as developers compete for constrained connection capacity and utilities seek to recover upgrade costs from large-load customers.

Europe’s regulatory environment is considerably more complex than either the US or UK, and its hub capacity significantly lags behind both. The majority of European data is hosted on US-owned servers, and the European Commission’s ‘State of the Digital Decade 2025’ report acknowledged that investment in sovereign cloud infrastructure falls well short of European Union (EU) policy goals. The Cloud and AI Development Act, expected later in 2026, aims to triple EU processing capacity within five to seven years, but permitting timelines and grid constraints continue in the meantime.

Separately, Europe’s data protection laws and evolving interpretation around cross-border data transfers creates ongoing risk. A November 2025 proposal to reform the definition of personal data offers only limited relief, and the invalidation of the EU-US Privacy Shield remains a reminder of problematic transatlantic data arrangements.

Further disputes ahead in AI finance, construction and asset security

Construction disputes are among the most immediate, as data hub projects are technically complex, multi-party undertakings involving developers, contractors, equipment suppliers, utilities and planning authorities.

The rapid evolution of hardware has transformed cooling, power and structural requirements, meaning that projects face a growing risk of specification drift: equipment specified at the outset may be technically obsolete before construction is complete, prompting scope changes, suspension of works or outright termination.

Supply chain disruption, as illustrated by the US electrical equipment shortage, adds a further layer of delay risk that existing construction contracts may not have been drafted to address. Research estimates that a 60MW AI hub loses approximately $14.2m in revenue for every month of delay, and that a six-month slip can cut project returns by nearly half, in many institutional models pushing projects below investment thresholds.

Major investors have also reportedly declined hub debt transactions due to insufficient insurance cover. Insurers are responding by turning to catastrophe bonds and special purpose vehicles to bring in alternative capital, with catastrophe bonds expected to yield above comparable government bonds and thereby increase project cost and risk, including existing projects proceeding on dated financial and contractual assumptions.

In terms of project financing and hyperscalers, Moody’s warned in February 2026 that US generally accepted accounting principles rules may allow companies involved in the AI business to keep billions of dollars in potential hub liabilities off their balance sheets.

Through short-term leases via special purpose vehicles, residual value guarantees may not appear as liabilities for those companies unless renewal or payment is deemed ‘reasonably certain’ or ‘probable’. Moody’s indicated it will conduct its own probability assessments and make quantitative debt adjustments where it believes reported lease liabilities are understated, a development with direct implications for credit ratings and market disclosures.

At the very serious end, beginning in March 2026, Iranian drone strikes affected two facilities in the UAE and caused proximate damage to a third in Bahrain, which was reported as the first deliberate military attacks on hyperscaler infrastructure. This has forced insurers to reprice attack risk and placed force majeure provisions under fresh scrutiny.

Sovereign risk is emerging as a parallel concern: as hubs become strategic national assets, the risk of host governments restricting operator access, mandating data localisation or effectively expropriating infrastructure through regulation is no longer theoretical.

Investors are increasingly looking to bilateral investment treaties and international arbitration as the primary protective mechanisms, though coverage and redress may be uneven in jurisdictions where build-out is fastest, especially without the right advice and forward planning.

The $3 trillion race to build AI infrastructure is consequently moving faster than legal and regulatory frameworks, resulting in strain and disputes. For businesses, investors and their advisers, all such legal exposure demands a sophisticated and tailored response.

Adam McWilliams and Elinor Sutton are partners and Julia Ochelska is a legal assistant at Quinn Emanuel Urquhart & Sullivan LLP. Mr McWilliams can be contacted on +44 (0)20 7653 2052 or by email: adammcwilliams@quinnemanuel.com. Ms Sutton can be contacted on +1 (212) 849 7325 or by email: elinorsutton@quinnemanuel.com.

Adam McWilliams, Elinor Sutton and Julia Ochelska

Quinn Emanuel Urquhart & Sullivan LLP