The distance between a customer name and a contract
Imperial's 330 MW is the requirement of Sebastian Rucci's proposed project, rather than the measured draw of an operating datacenter. Roughly 950,000 square feet and more than 130 gas generators cannot yet be counted as the inventory of a delivered computing service. References to training Google's Gemini models in Rucci's correspondence do not change that distinction. Google says it is not involved; Rucci acknowledges that no formal agreement exists. When a model's name stands in for a customer contract, the commercial foundation of physical capacity appears further advanced than it is.[1]
I do not take this as proof that the project can never find a customer. Discussions with prospective users can be ordinary business for an early-stage developer. But a discussion, an agreement and utilisation are different milestones. Google's direct response prevents counting the proposal as a confirmed Google investment. For the same reason, adding 330 MW to AI's already realised electricity consumption would be wrong. The missing demand-side milestone cannot be supplied by the developer's description of a named company's interest; it requires a binding commitment from the party that would actually use the capacity.[1]
A power plan does not supply water
Rucci says the project would use 750,000 gallons of water a day. That is the developer’s projected daily volume, not current consumption. While environmental review and permitting continue, describing the number as realised consumption changes the project’s stage. Power and water are also different quantities: 330 MW is a power requirement, while daily gallons describe a flow volume. Putting both in a capacity table does not explain how the cooling arrangement would operate. Effects on water and air quality are central to local objections. A generator count alone cannot answer those concerns.[1]
September's court decision requires a full environmental review before construction begins. That review covers the grid, water and emissions as well as alternative locations. Ordering more chips alone cannot remove this bottleneck: the environmental conditions of the project that would house them must first be resolved. Review is not the same as a ban; the project may adapt to changes. Nevertheless, assuming a favourable result and placing the proposed capacity on a delivery schedule does not remove the uncertainty carried by permitting. This legal and physical transition stands between acquiring land and operating a facility.[1]
The time scale of the local bargain
Rucci's expectations of more than 1,600 construction jobs and nearly 30 million dollars in annual tax revenue are the project's offer to local authorities. Both are the developer's estimates. A construction-period job count and revenue expected to recur every year do not share a time denominator. Water and air-quality concerns, meanwhile, concern daily life around the proposed facility. Evaluation therefore cannot rest on a total megawatt number alone. If the project proceeds, the period in which benefits arise and the operating conditions that produce environmental burdens need to be compared for the same site and stage.[1]
What Imperial shows today is how proposed AI capacity depends on several separate commitments. A customer agreement underpins commercial use, water supply supports cooling, and environmental review governs the move into construction. One milestone cannot serve as documentation for another. I would not count this project as 330 MW with a confirmed customer and commissioned operation. Progress is measured by resolving the separate dependencies, rather than by repeating the same number in more announcements. Treating local objections merely as delay also leaves part of the calculation out: the conditions under which a facility may operate belong to the physical definition of the computing service it would deliver.[1]