The data centre boom is running into serious real-world obstacles, from widespread local opposition and permitting delays to a scramble for electricity. Before those challenges even come into play, however, a more basic problem has emerged: a large share of the projects being announced may never get built.
Industry analysts estimate that billions of dollars are being pledged for AI infrastructure, yet the gap between announcements and completed facilities is widening. The mismatch threatens to slow the rollout of the high-performance computing capacity that AI developers say they need.
Why So Many Data Centres Are Announced
Tech companies and venture-backed developers have been racing to stake claims on land, secure financing and publicise plans for massive data-centre campuses. The hype is driven by three factors. First, generative AI models such as large language models require petabytes of storage and thousands of GPU cores, pushing demand for specialised infrastructure. Second, governments in North America, Europe and Asia have offered tax incentives, subsidies and fast-track permitting to attract the lucrative AI supply chain. Third, investors are eager to capitalise on the perceived inevitability of AI-driven growth, pouring capital into early-stage projects to lock in future revenue streams.
Local Opposition and Community Concerns
Communities where data centres are proposed often push back on several fronts. Residents worry about increased traffic from construction trucks, the visual impact of large metal buildings, and the potential for noise from cooling systems. Environmental groups raise alarms about the carbon footprint of data-centre operations, especially when the power mix relies on fossil fuels.
In the United States, several high-profile proposals in suburban Texas and the Pacific Northwest have been met with petitions, town-hall protests and legal challenges. Similar patterns are emerging in Europe, where local councils in Sweden and Finland have delayed approvals pending environmental impact assessments.
Permitting Delays and Regulatory Hurdles
Data-centre construction requires a suite of permits covering land use, building codes, fire safety, water usage and, most critically, electricity connections. In many jurisdictions, the permitting process can take 12 to 24 months, far longer than the rapid timelines companies advertise.
Regulators are increasingly scrutinising the energy efficiency of proposed facilities. New standards in the European Union, for example, demand that data-centre power usage effectiveness (PUE) meet stricter thresholds before a licence is granted. In the United States, the Federal Energy Regulatory Commission (FERC) is reviewing proposals that could strain regional transmission grids.
The Electricity Crunch
AI-focused data centres consume large amounts of power, often measured in tens of megawatts per site. The surge in demand is colliding with existing grid constraints, especially in regions that have already experienced capacity shortages.
- In the American Southwest, summer peak loads have approached 90% of available capacity, leaving little margin for new high-intensity users.
- European grids are grappling with the dual challenge of phasing out coal and integrating intermittent renewable sources, which can limit reliable supply for data-centre loads.
- Emerging markets in Southeast Asia are still expanding transmission infrastructure, making it difficult to guarantee the continuous power needed for AI workloads.
Utility companies are responding by raising connection fees, imposing demand-response obligations, or outright rejecting new high-consumption requests until upgrades are completed.
Financial Implications of Unbuilt Projects
When a data-centre project stalls or is cancelled, the financial fallout can be significant. Developers may have already secured land purchases, paid for preliminary engineering studies and begun site preparation. These sunk costs can run into tens of millions of dollars per site.
Investors who have committed capital based on announced projects may face write-downs. Venture-capital funds that allocated billions to AI infrastructure startups could see reduced returns if the underlying facilities never materialise.
Historical Context: Past Over-Promising in Tech Infrastructure
The current situation mirrors earlier tech booms. In the late 1990s, many dot-com companies announced massive server farms that never left the drawing board, leading to a wave of bankruptcies when the bubble burst. Similarly, the early 2000s saw a proliferation of "green-energy" data-centre projects that were later scrapped due to unrealistic power-availability assumptions.
These precedents highlight the importance of aligning public announcements with realistic timelines, regulatory pathways and power-supply guarantees.
What Regulators and Industry Players Are Doing
Governments are tightening oversight to prevent speculative announcements. In Canada, the federal ministry overseeing innovation has introduced a requirement that any AI-related data-centre seeking public funding must submit a detailed power-sourcing plan vetted by an independent auditor.
Industry consortia are also stepping in. The Global Data-Centre Alliance has released a best-practice guide that recommends developers conduct community impact assessments before publicising projects, and to engage with local utilities early in the planning stage.
Practical Takeaways for Stakeholders
For investors, the key lesson is to scrutinise the feasibility studies behind each announced project, not just the headline funding figures. For local governments, transparent permitting processes and clear communication with residents can mitigate opposition and speed approvals.
Companies planning new AI infrastructure should prioritize regions with surplus renewable generation or established grid capacity, and consider modular designs that can be scaled up as power becomes available.
What to Watch Moving Forward
Analysts will be monitoring several indicators to gauge whether announced data-centre projects are likely to proceed:
- Submission of formal electricity connection applications to regional utilities.
- Approval of zoning changes and building permits by local authorities.
- Public statements from major investors confirming continued financing beyond the announcement phase.
When these milestones are achieved, the probability of a project reaching construction increases markedly. Conversely, prolonged delays in any of these areas suggest the project may be abandoned.
In the meantime, the broader AI ecosystem will need to adapt to a potentially slower rollout of dedicated infrastructure, exploring alternatives such as shared cloud resources, edge-computing nodes and more efficient model architectures to reduce compute demand.