Power Access Is Becoming an Allocation Decision

Power Access Is Becoming an Allocation Decision

For data-centre developers, a grid connection is no longer only a technical study followed by a date. It is increasingly an allocation decision. Grid operators and governments are starting to decide which loads are credible, which can operate flexibly, which create wider economic value and which should lose their place when scarce capacity is blocked.

Two decisions published on 28 October make that shift unusually clear. In California, the CAISO board approved a large-load proposal that includes a flexible interim connection while related network upgrades are completed. In the United Kingdom, the government’s new Data Centre Covenant offers faster treatment for qualifying strategic projects, including the reservation and reallocation of grid capacity, possible transfer of pre-2030 connection contracts and tightly controlled temporary generation.

The two frameworks are not equivalent. CAISO is responding to a federal regulatory process and must preserve reliability and cost allocation. The UK is creating an explicitly strategic gateway through government. Yet they point in the same direction: a project will increasingly need to prove more than demand and land control. It will need to prove readiness, flexibility, public value and an ability to carry the risks that used to sit implicitly with the network.

My conclusion is practical. Developers should stop treating a connection offer as a passive project asset. It is becoming a performance contract. The decisive question is not only “How many megawatts have been requested?” It is “What operating, financial and delivery commitments justify allocating scarce system capacity to this project?”

California is turning flexibility into an interim connection product

The California ISO board decision advances new procedures for large loads such as data centres and manufacturing plants. CAISO plans to file the proposal with the Federal Energy Regulatory Commission by 16 November 2026 in response to FERC’s June show-cause order.

The most important mechanism is the proposed flexible interim load interconnection. A large load could connect before associated network upgrades are complete, subject to enforceable flexibility determined by grid conditions and constraints. This is not an unconditional acceleration. It exchanges part of the developer’s right to consume continuously for an earlier, conditional right to energise.

That distinction matters economically. A conventional connection delay leaves a completed building, customer pipeline and equipment plan waiting for power. An interim flexible connection can bring part of the asset into service earlier, but it transfers new obligations to the operator. The campus must know which workloads can move in time or location, how backup systems interact with grid instructions, what curtailment does to customer commitments and which party bears the resulting revenue risk.

The proposal also seeks better coordination among utilities, technical studies and operational requirements. That should improve system visibility, but it also raises the standard of evidence required from a developer. A vague long-term load forecast becomes less useful when the system operator needs to know how the load will behave during constrained hours.

In my view, flexible connection is most valuable when four conditions hold. The initial energisation has real commercial value. The flexible portion can be controlled without breaching customer or resilience commitments. The network-upgrade path remains credible. The contractual allocation of curtailment risk is explicit. If any of those conditions is missing, an earlier connection date can create the appearance of progress without producing dependable capacity.

The UK is moving from queue order to strategic selection

The UK Data Centre Covenant goes further in political terms. Projects that meet strategic tests may receive support from a new ministerial control centre. The government says it can reserve future capacity, reallocate capacity released when projects leave the queue, identify locations with capacity before 2030 and facilitate transfers of pre-2030 connection contracts to strategically important data centres.

Temporary on-site generation may be considered for exceptionally strategic projects when other measures cannot secure capacity in time. That option is described as tightly controlled, not a general route around the grid. The same covenant expects projects to address national security, resilience, energy-system costs, local benefits, water and energy efficiency, UK jobs, supply chains and AI hardware.

The scale of the queue explains the intervention. The government’s strategic-demand consultation refers to around 140 data centres in the transmission queue representing approximately 50 GW. An earlier Ofgem consultation said total demand connection applications had risen from 41 GW to 125 GW in under a year, with data centres accounting for at least 80 GW of the increase. These figures cover different points and definitions, so they should not be combined. Both nevertheless show a queue too large to be treated as a neutral forecast of future load.

The policy response is therefore selective. A commitment fee and progress milestones are intended to remove speculative demand. Strategic alignment can then move qualifying projects faster. This replaces one imperfect signal, queue position, with a more complex judgement about credibility and national value.

That may improve allocation, but it creates a new diligence question. The government has not yet published all the detailed criteria for a strategic data centre. Until those criteria are operational, political support should be treated as a pathway, not bankable capacity. Investors should ask what evidence triggers priority, who can revoke it, how competing strategic projects are ranked and whether a transfer or reservation survives changes in project scope.

The same constraint is producing two different instruments

PowerlandMap’s review of the two announcements separates their legal stage and their operating mechanism. This matters because a policy offer, a board-approved proposal and an operational connection are not interchangeable.

The common mechanism is risk transfer. Grid systems cannot build against every requested megawatt as if it were certain. They are responding by asking developers to post stronger evidence, accept operational constraints, pay commitment costs or demonstrate strategic value. Capacity is still scarce, but the right to use it is becoming conditional.

What changes in project economics

This shift changes the value of flexibility. A data centre that can move non-urgent training workloads, stage commissioning or contract for controllable demand may reach revenue earlier than a rigid campus waiting for its full firm connection. Flexibility is no longer only an energy-market optimisation. It can become a schedule asset.

But the value is not automatic. Earlier energisation must be compared with the costs of workload orchestration, backup operation, customer concessions and periods of constrained output. A developer should model at least three states: the firm capacity available at first energisation, the conditional capacity available under the interim arrangement and the final capacity after network upgrades. Revenue should be attached only to the capacity that customers can actually use under each state.

Temporary generation creates a similar trade-off. It can bridge a schedule gap, but fuel supply, emissions, permits, noise, local opposition and the transition to grid supply all affect the economics. A temporary plant should not be valued as if it creates permanent firm grid capacity. Nor should the announced campus load be converted into delivered IT megawatts without the electrical design, PUE boundary and commissioning evidence.

The parties best positioned to benefit are projects with committed capital, credible customers, modular commissioning and operational control. Campuses whose economics require immediate, continuous use of the full requested load may find the new frameworks less helpful. The same is true for projects built around speculative queue positions rather than a deliverable site and customer plan.

Strategic allocation can solve one problem and create another

The strongest argument for the new approach is straightforward. First-come treatment rewards early applications, not necessarily the best projects. It can encourage oversized or duplicated requests and force networks to plan against demand that may never arrive. Fees, milestones and flexibility can release blocked capacity and reduce cost shifting.

The counter-argument is equally serious. Strategic selection can become opaque. If criteria are unstable, developers may optimise for policy language rather than system value. A project with strong political sponsorship could displace a quieter but more advanced scheme. Flexible arrangements can also hide risk if customers assume firm service while the grid retains broad curtailment rights.

The answer is not to return to a purely chronological queue. It is to make allocation rules measurable. Readiness should be tied to evidence such as land control, permits, financing, equipment orders, customer commitments and a dated construction programme. Flexibility should specify duration, notice, frequency and recovery conditions. Public value should be assessed against disclosed criteria. Reallocated capacity should carry an audit trail.

For investors, the next decisive documents will not be another press release. In California, it will be the FERC filing and the enforceable tariff language, followed by the first projects using the interim route. In the UK, it will be the detailed definition of strategic data centres, the commitment-fee and queue-management rules, and evidence that reserved or transferred capacity reaches viable projects.

The project screen I would use now

I would add four questions to every powered-land and data-centre review.

First, what exactly is firm at each stage? Separate the initial connection, any conditional interim capacity and the final upgraded position.

Second, what flexibility is technically and commercially available? The answer must cover workloads, notice periods, backup systems and customer contracts, not only a theoretical demand-response claim.

Third, what evidence supports priority? Record the applicable policy, decision-maker, criteria, expiry conditions and next formal milestone.

Fourth, who pays if the project or the network slips? Connection charges, temporary generation, stranded works and customer compensation should not remain implicit.

Power scarcity is not disappearing. What is changing is the institution that decides how scarce capacity is used. The winning projects will not simply be those that applied first or announced the largest number. They will be those that can prove delivery, accept transparent conditions and convert a conditional allocation into reliable service.

*Illustration: original PowerlandMap editorial artwork. It is conceptual and does not depict a specific grid or data-centre site.*

*Sources reviewed 29 October 2026: California ISO large-load decision and stakeholder materials, UK Data Centre Covenant and strategic-demand consultation, Ofgem demand-connections consultation. Secondary editorial benchmarking included Data Center Dynamics and Data Center Knowledge coverage of grid queues and large-load flexibility. All interpretations are PowerlandMap analysis.*