The Phantom Data-Centre Pipeline

Permit arbitrage, queue inflation, and why “planned MW” is not the same as “buildable MW”

From Demand Forecasts to Delivery Risk

In the previous article, we showed that AI‑driven electricity demand is uncertain. Efficiency gains, hardware cycles, and changing usage patterns mean demand does not follow a single straight line.

The same is now true on the supply side. More so now with the SPACEX and XAI merge and the reality of data centers in space closer.

The data‑centre pipeline dominating electricity forecasts is not a construction schedule. It is a set of possibilities. Developers submit multiple applications across regions and substations, knowing that only a small share will ever be built. As one senior industry participant recently put it in a private discussion, “we expect roughly a 25% hit rate on data‑centre connection requests.”

The result is a phantom pipeline: headline megawatts that appear firm in aggregate but compress when tested for deliverability. This is not a moral failure by developers. It is a predictable response to congestion, slow permitting, and uncertain grid access. Filing multiple applications preserves optionality in a congested, slow‑moving system. The risk arises when these applications are interpreted as firm demand rather than intent.

Why Project Counts and Capacity are Systematically Overstated

Across multiple markets, the same structural behaviour is emerging. A single developer or hyperscale tenant may submit multiple interconnection and planning applications across regions, utilities, or substations, with the intention of building only where the first viable connection clears. The rest are withdrawn later.

From the developer’s perspective, this is rational. Grid access has become the scarcest input to AI infrastructure. Filing multiple “shots on goal” preserves optionality in an environment where queues are long, information is imperfect, and timing has real economic value.

From the system’s perspective, however, these requests are often read as firm demand. Regulators and independent analysts have warned that duplicate applications can inflate official load forecasts and, if treated unadjusted, encourage network investment that may ultimately not be required. This risk—especially when upgrades are built for projects that do not materialise—has been highlighted by the California Public Advocates Office and others (publicadvocates.cpuc.ca.gov).

In Great Britain, the issue became explicit as the connections queue grew to levels that far exceeded plausible, deliverable capacity. NESO’s connections-reform programme aims to restore signal quality by prioritising projects with credible deliverability evidence and deprioritising unviable requests. The objective is not to suppress demand, but to make queue signals useful for planning and investment (National Energy System Operator (NESO) (opens in a new tab)).

This is the key point: “planned MW” is not the same as “buildable MW.” Treating them as equivalent introduces systemic risk—not because developers behave irrationally, but because planning frameworks often interpret intent as commitment.

Scarcity changes behaviour - and valuation

Not every project that fails to secure a grid connection disappears. In some cases, developers change approach instead of walking away. They look to supply power themselves through on‑site generation, co‑located renewables with storage, private wires, or hybrid solutions that reduce reliance on the public grid.

This matters because these projects no longer appear in connection queues or official demand forecasts. From the outside, it can look like demand has eased. In reality, it has shifted off‑grid. The risk has not gone away, it has moved. Instead of depending on system access, these projects carry higher capital costs, fuel and operating risks, and greater exposure to policy and regulatory change.

When projects self‑provision, they stop being a planning signal and become a balance‑sheet decision. Treating grid‑connected and self‑powered capacity as equivalent would blur where risk sits and can make the public pipeline appear clearer than it is.

Importantly, these responses do not remove the underlying scarcity. When access is uncertain and signals are noisy, rational developers hedge earlier and more widely, applying in more locations and keeping options open until something clears. What begins as an adaptation to congestion can become a system‑wide feedback loop.

The vicious cycle of queue inflation

Once permit and connection arbitrage becomes widespread, a self-reinforcing cycle takes hold:

1. Speculative requests proliferate as developers hedge against scarcity.
2. Forecast demand inflates, because system planners treat requests as real load.
3. Network build pressure increases, attracting political attention and public concern about costs.
4. Rules tighten, through gating, moratoria, or discretionary approvals.
5. Developers respond with even more arbitrage, filing earlier, wider, and less mature applications.
6. Signal quality deteriorates further.

The outcome is contradictory. Systems can appear overwhelmed by demand on paper, while genuinely viable projects struggle to progress in practice. Capital may be allocated against options rather than assets, while deliverable capacity is delayed. The challenge is not behaviour, it is interpretation.

Making queue inflation visible

Evidence of speculative or “phantom” data-centre demand is now visible across multiple power systems (see details in Table 1). While the drivers differ by jurisdiction, the underlying pattern is consistent: connection queues increasingly reflect option-taking behaviour rather than firm construction pipelines.

• Swelling queues: Developers in the US are filing large numbers of interconnection requests, many of which utilities acknowledge will never materialize. This creates “phantom data centres” that complicate peak demand forecasts and investment planning. (Rextag Corporation (opens in a new tab))
• Market responses: US utilities serving major data-centre hubs are tightening deposit and tariff requirements to weed out low-commitment filings. (IPPC (opens in a new tab))
• Queue reform: In Great Britain, Ofgem and NESO are rapidly reforming the connections process to prioritise projects with demonstrable readiness, reducing previously bloated queues dominated by speculative applications. (NESO (opens in a new tab))
• Regulatory acknowledgement: The Office of Gas and Electricity Markets has detailed how the demand queue now vastly exceeds forecast need and is consulting on interventions to manage this growth. (Ofgem (opens in a new tab))

The result is a growing recognition within planning and regulatory communities that connection queues are not demand commitments; they are signals influenced by speculative behaviour that must be screened and weighted if forecasts and investment decisions are to be reliable.

Table 1 — Comparative Risk Landscape

What this Means for Different Stakeholders

Infrastructure Investors

Connection scarcity is creating a clearer value divide in the data-centre world. For existing or confirmed data centres with secured grid access, scarcity raises valuation. These assets benefit from reduced competitive supply and greater confidence in revenue delivery, which supports tighter discount rates and higher long-term value.

Market behaviour supports this shift. Utilities with established infrastructure are beginning to realise value through data-centre related asset sales. For example, RWE recently reported a €225 million gain on the sale of a UK data-centre project, boosting earnings and lifting its share price to multi-year highs. Analysts linked the move to strong hyperscaler demand for sites with ready power access. (CNA (opens in a new tab))

In contrast, the lack of certain queues around leads to fallout. A recent note from analysts at TD Cowen indicates that Microsoft has walked away from planned data centre projects totalling around 2 gigawatts of capacity in the U.S. and Europe, citing oversupply relative to demand expectations and evolving infrastructure strategy. (Data Center Dynamics (opens in a new tab))

For infrastructure investors, the implication is straightforward: firm, deliverable assets with grid access are being re-rated upward, while unconfirmed or speculative projects are increasingly de-rated. In a constrained network environment, distinguishing between buildable capacity and paper options is essential to avoid overstating portfolio resilience.

Governments

Governments increasingly frame data centres as engines of economic growth and digital sovereignty. However, growth ambitions based on headline pipeline megawatts risk overstating what can realistically be delivered. Connection queues and planning interest are not the same as buildable capacity, particularly once constraints in electricity networks, water availability, and local infrastructure quality are taken into account.

This matters for designated high-growth regions. Many of the areas identified as priority data-centre clusters face weak transmission capacity, limited water resources, or long reinforcement timelines. Without parallel investment in energy and water assets, stated AI-led growth targets may prove difficult to realise. The risk is not a lack of demand, but a mismatch between policy ambition and physical system readiness.

For investors, this creates geographic differentiation. Regions where governments align industrial strategy with credible investment in power, water, and permitting capacity are more likely to attract durable capital. Others may see interest stall as delivery risk becomes clearer.

Utilities and System Operators

Building infrastructure in the UK is expensive, slow, and capital-intensive. Misreading data-centre demand risks crowding out other essential investments or locking capital into upgrades sized for projects that never connect. This heightens the importance of signal quality in planning decisions.

NESO plays a central role in addressing this. By reforming connection queues and prioritising (NESO (opens in a new tab)) demonstrably viable projects, it improves capital sequencing and reduces the risk of speculative overbuild. But delivery also depends on coordination across the system. Transmission owners, distribution networks, and independent DNOs must align incentives and timelines.

In some cases, IDNOs offering flexible, multi-connection solutions may be better placed to unlock viable data-centre sites quickly. However, without system-level coordination, this can shift costs rather than reduce them. For utilities and investors alike, disciplined planning - treating the pipeline as submissions for optionality, not commitments - is essential to deploying capital efficiently.

Pricing reality, not optionality

AI is reshaping electricity demand, but the near‑term risk is not that growth disappears. It is that intent is mistaken for certainty.

For investors, regulators, and system planners, the discipline is the same: separate buildable capacity from speculative scale. In constrained grids, that distinction increasingly determines value.

For investors, regulators, and system planners, the discipline is the same: separate buildable capacity from speculative scale.

At Vallorii, this is central to how we approach infrastructure valuation. Expected outcomes matter more than headlines. Signal quality matters more than size.