Summary
- Microsoft’s Essex decision and a Dutch manufacturer’s lost customer illustrate the commercial costs of delayed grid connections.
- Grid queue reforms focus on readiness and delivery, but may miss businesses whose investment deadlines expire before a useful answer arrives.
- Amir Naser Hojati proposes an 'answer-in-time rate' and follow-up on abandoned applications to measure whether connection information arrives early enough.
Microsoft withdrew in October from an agreement to anchor Nscale’s planned 50-megawatt AI data centre in Essex. The site was due to open by 2027, but sufficient grid power was not expected until the 2030s. Microsoft says the decision concerns this site alone and continues to work with Nscale elsewhere.
Across the North Sea, Dutch mustard maker Marne Mosterd spent four years seeking another 250 kilowatts of electricity. In March, network operator Enexis identified additional capacity, clearing the way for a fourth production line in Groningen. But a prospective customer had already chosen a competitor. Director Paul de Vries said the customer could not wait five years.
The projects are vastly different, yet both reveal how commercial plans can change before electricity needs are resolved. Nscale still intends to develop its site without its intended anchor customer. Marne can expand without the order it lost. Sometimes the decisive deadline belongs to a customer the network operator never meets.
Grid operators work towards a connection date. Businesses face an earlier one, when financing must close, equipment be ordered or a contract signed. An estimate arriving afterwards can be accurate and still commercially useless.
Amazon Web Services told Reuters in February that grid timing had become one of the biggest factors in its European investment decisions. A data centre can take about two years to develop, while a transmission connection can take up to seven. AWS said grid congestion or missing connections had made planned projects unfeasible in many countries.
Regulators are tackling a real but different problem: speculative applications. In Britain, contracted demand-connection offers jumped from 41 gigawatts in November 2024 to 125GW by June 2025, with around 73GW attributed to data centres in Ofgem’s consultation. That dwarfs Britain’s February peak demand of roughly 45GW, although connection offers are not actual consumption.
Ofgem has proposed commitment fees and progress checks for large data-centre projects. The EU’s Grids Package aims to accelerate network planning and permitting, while Commission guidance favours assessing project readiness.
These are deadlines for delivering infrastructure. Customers have deadlines too. Clearing a queue cannot show whether investors received an answer while they still had a choice.
In the first half of 2026, Dutch regional operators granted capacity to 1,699 requests; another 613 left waiting lists for other reasons. Not every departure represents lost investment. The published totals cannot distinguish routine withdrawals from postponed expansions or investment moved elsewhere. Nor do they count firms deterred before applying.
Britain’s ADE: Demand describes a manufacturer that abandoned electrification plans after an indicative eight-year connection wait exceeded the remaining life of its furnace. A business facing that choice might replace ageing equipment with another gas-fired furnace. Its abandoned electricity demand would never enter the queue.
Money compounds the problem. Dutch state-backed investor Invest-NL reported in October that companies may have to pay 10 to 20 per cent of connection costs upfront, sometimes three to seven years before knowing whether expansion can proceed. A manufacturer may need a credible electricity timetable to raise finance, while readiness checks demand evidence of financing first.
Such tests may favour applicants able to absorb uncertainty without revealing whose commercial deadline comes first.
The European Commission estimates that electricity grids will need roughly €1.2tn of investment by 2040. Planners already look beyond connection applications. But where those applications carry substantial weight, businesses discouraged from applying risk disappearing from forecasts of future demand. That can obscure investment that better infrastructure might have attracted.
A practical reform begins with one question: by what date must you receive credible connection information for it to influence your investment decision?
Operators could record that date confidentially, alongside realistic connection windows, indicative costs and flexible-supply options. Regulators could then test an answer-in-time rate: the share of participating applicants receiving a useful assessment before their stated deadline.
Companies might exaggerate their urgency, and genuine deadlines change. Record the original date and subsequent revisions, grant no queue priority for claimed urgency, and check whether information proved useful rather than merely prompt.
This still misses firms that never applied. Regulators should follow a sample of enquiries, applications and withdrawals, asking what businesses intended to build and what happened instead.
Berkeley Lab offers a precedent in tracking American generation-connection requests. Active queued capacity fell about 10 per cent in 2025, amid withdrawals exceeding 750GW. Generators are not consumers, but the principle holds: a shorter queue tells little without knowing why it shrank.
Ofgem’s proposed demand-capacity register, open for consultation until November 6, offers an immediate place to test the idea.
Europe needs more cables and substations. It also needs to give answers while businesses can still use them.
In Groningen, Marne can move ahead with its fourth production line. The extra electricity can support future orders. It cannot bring back the customer that left.
About the author
Amir Naser Hojati is a futures trader and fintech entrepreneur with more than 15 years of experience in financial markets. He writes about European economic policy, capital allocation and investment risk. He is the author of a SUERF policy brief on systemic risks in AI-driven financial markets. His work has appeared in Central Banking, Funds Europe, The TRADE and Brussels Signal.
