When Does a Grid Connection Become Deliverable?

Power can be allocated on paper and still fail to arrive when a data centre needs it. That distinction is becoming one of the most expensive blind spots in AI infrastructure underwriting.

The issue is visible in the United Kingdom today. Nscale’s official Loughton announcement described a 50 MW IT deployment, scalable to 90 MW, with a targeted first-quarter 2027 opening. On 4 October 2026, TechRadar reported that full supply may not be available until the early or mid-2030s. Euronews also reported the mismatch between the intended delivery programme and the network timetable.

This is not an argument that the project will not happen. It is a reminder that “capacity”, “connection” and “deliverable power” are different facts. We already separate those concepts in our note on why data centre capacity is not one number. Grid due diligence should apply the same discipline.

Start with the question the investment committee actually needs answered

A grid connection is deliverable when the contractual, technical, permitting and construction conditions can credibly support the required energisation date and load profile.

That definition is deliberately stricter than “the utility has acknowledged the request”. It is also stricter than “a connection offer exists”. Offers can contain dependencies, reinforcement works, staged capacity, security payments, long-stop dates and assumptions that materially change the investment case.

The practical question is therefore not:

How many megawatts are associated with the site?

It is:

Which megawatts can be energised, under what conditions, by which date, and what evidence supports that conclusion?

This framing connects directly to development-readiness milestones. A project can be advanced on land, design and permitting while remaining early-stage on power.

An evidence ladder for grid claims

I use six levels of evidence. They are not a universal legal taxonomy, but they provide a common language across markets.

1. Network proximity

A substation, overhead line or cable route is near the site. This is useful for screening, but it proves neither spare capacity nor a viable connection configuration.

At this level, the site is a candidate. It is not powered land.

2. Utility engagement

The sponsor has made an enquiry, submitted an application or entered a formal study process. The network operator may have provided an indicative point of connection or technical parameters.

This is evidence of activity, not delivery. In France, for example, RTE explains the staged process for connecting an installation to the public transmission network. Each stage matters because technical studies, proposals and agreements do not carry the same certainty.

3. Connection offer

A formal offer identifies capacity, connection point, works, milestones and commercial conditions. This is a material step, but the offer must be read rather than merely cited.

The key questions are:

• Is capacity firm or interruptible?
• Is it available in one tranche or several?
• Which reinforcement works sit on the critical path?
• Which dates are binding, indicative or conditional?
• What security, reservation or milestone payments are required?
• Can the programme survive a change in the proposed IT load?

A data room should preserve the full document trail, not a slide that restates the headline MW.

4. Executed agreement and maintained milestones

Once the connection agreement is signed, the diligence focus shifts to compliance. Has the project paid required deposits? Has it submitted technical data on time? Are land rights secured for the substation and cable route? Are planning and environmental approvals aligned with the utility programme?

This is where permit due diligence and grid diligence converge. A connection may be contractually allocated yet exposed to a route, easement or permitting delay.

5. Network and customer works under construction

Physical progress changes the quality of evidence. Procurement, factory slots, enabling works, cable routes, substations and protection systems can be checked against an integrated schedule.

The owner should maintain one critical path that joins network milestones to building completion, commissioning and customer readiness. Our commissioning-readiness framework is relevant here: energisation is not a standalone date but an input to testing, load-bank strategy and handover.

6. Energised capacity with an evidenced ramp

The strongest evidence is delivered power at the required voltage and quality, together with a credible ramp to the target IT load. Even then, diligence should distinguish initial energisation from full contracted capacity.

For AI facilities, the ramp matters because capital deployment may run ahead of usable power. That is the central risk described in capital is not capacity.

Four dates belong in every investment model

A single “power date” is not enough. At minimum, the model should track:

1. connection-offer or agreement date; 2. utility works completion; 3. initial energisation; 4. full-capacity availability.

Between those dates, the sponsor should state the capacity available in each tranche and the probability assigned to it. The base case must not silently assume that a 90 MW end-state is available when only an initial block is scheduled.

This is especially important when the commercial plan contains a pre-lease, cloud commitment or accelerator procurement. The obligations in a data centre pre-lease should be tested against the same grid milestones used in the technical schedule.

What to do when the grid date moves

A delayed grid connection does not automatically destroy a project. It changes the sequencing problem.

The first response is to protect the facts: update the evidence register, identify the specific dependency and separate confirmed utility information from sponsor inference. The second is to re-run the phasing model. A smaller first phase may have a different delivery path from the ultimate campus load.

Alternative supply can then be assessed, but without treating it as a slogan. Behind-the-meter generation, batteries or private-wire structures have their own fuel, emissions, permitting, resilience and bankability constraints. Our review of behind-the-meter power for data centres sets out the main questions.

At policy level, Europe recognises that electricity networks have become a central infrastructure constraint. The European Commission’s European grids overview describes the investment and planning challenge, while the Commission’s 2026 proposal on data centres and grid participation points toward more explicit treatment of flexibility and system impact. These initiatives may improve the framework, but they do not replace project-level evidence.

A compact control pack for owners and lenders

For each site, the decision pack should contain:

• a one-page connection summary with MW, voltage, tranche and date;
• source documents and their version history;
• a utility-contact log with open decisions;
• a map of connection assets, routes and land rights;
• dependencies between utility works and customer works;
• deposits, securities and lapse conditions;
• permitting status for every off-site component;
• a dated risk register with named owners;
• downside cases for delayed and staged energisation.

This pack should be short enough to use and detailed enough to audit. It also creates a common interface between development, investment, procurement and commercial teams.

For portfolio investors, consistency matters more than false precision. A comparable grid-evidence score across markets can reveal which sites are genuinely ready and which remain option value. That comparison is one of the reasons we built the European data centre development pipeline study.

My view: the grid claim should age faster than the land claim

Land ownership can remain stable for years. Grid evidence can deteriorate in weeks if milestones slip, queue policy changes or network works are reprioritised.

For that reason, I would not treat a connection letter as a permanent attribute of a site. It is a time-stamped claim that should be reverified at every material development, financing or leasing decision. The stronger the commercial commitment, the fresher the grid evidence should be.

PowerlandMap tracks projects, power milestones and source dates so teams can compare markets without collapsing every claim into a single MW number. If you are screening a portfolio or testing a site’s power case, request access and we can walk through the evidence model with you.

*Matthieu Gallego*