Power demand from AI infrastructure does not arrive at the grid as one uniform category, even when different customers ultimately need the same commodity. A neocloud can approach a network operator with an immediate requirement for a defined block of capacity, potentially linked to its planned GPU deployment, customer commitments, and deployment schedule that can begin with a relatively modest electrical footprint. A hyperscaler can frame demand around a much larger portfolio strategy, where an individual site may represent one component of a wider infrastructure expansion plan and the eventual requirement may extend substantially beyond its initial energization. That distinction matters because network planning evaluates not only the magnitude of demand but also its certainty, timing, location, and engineering consequences. NESO has identified speculative and duplicated data center applications as a specific challenge when trying to establish a credible picture of future demand across Great Britain.

A 30MW request can therefore carry a different planning signal from a 1GW ambition, even when both customers expect substantial growth over time. The smaller request can define what the customer can energize, operate, and monetize within a known development phase, while preserving a credible pathway toward subsequent capacity additions. A gigawatt request requires network planners to consider a much larger combination of generation availability, transmission reinforcement, substation capacity, system stability, and competing connection requirements before the full requirement becomes actionable. Ofgem’s current proposals specifically target data center projects that occupy queue capacity without demonstrating sufficient readiness, using commitment fees and progress milestones to distinguish serious projects from speculative demand. The regulator has reported that demand connection applications increased from 41GW to 125GW in less than a year, with data centers accounting for at least 80GW of that demand.

The Phased Block Strategy That Fits Today’s Network

Phased capacity changes the engineering conversation because it separates the customer’s ultimate ambition from the amount of power required on day one. A neocloud could structure deployment around discrete blocks, such as an initial 20MW or 30MW requirement followed by additional capacity as computing demand, hardware procurement, and customer commitments mature. Each block can correspond to a defined electrical load, a defined equipment deployment, and a measurable operational milestone rather than leaving the network to accommodate an entire future estate immediately. This structure can reduce uncertainty around the timing of demand and give network companies a clearer basis for evaluating what infrastructure needs to exist at each stage. NESO’s data center work has already highlighted the difficulty created by repeated speculative applications and inconsistent descriptions of data center power requirements.

The commercial logic behind phased blocks matters as much as the electrical logic because AI compute infrastructure generates value when installed hardware becomes operational, not when an ultimate power allocation appears on a planning document. A customer that can commission one block, establish utilization, secure revenue, and then activate another block creates a sequence of operational milestones that can provide additional evidence of project progress. That structure does not eliminate transmission constraints, distribution constraints, or reinforcement requirements, but it can make the relationship between infrastructure investment and actual demand more explicit. NESO’s connections reform separates projects into different stages and categories, with transmission, large embedded, and distribution offers moving through defined phases rather than relying on an undifferentiated queue. The reform framework gives particular importance to projects that can demonstrate their readiness and connection requirements within the revised process.

When Speed to Connection Beats Scale of Connection

AI infrastructure economics place significant value on the date when computing capacity becomes usable, because an idle GPU allocation does not generate the same commercial value as an energized and commissioned installation. A neocloud may therefore have an incentive to optimize for the earliest practical block of power that can support production workloads rather than waiting for an entire long-term campus requirement to become available. The distinction becomes significant when network reinforcement cannot match the customer’s desired development schedule, since a smaller connection can potentially align with infrastructure that exists or can be reinforced within a defined project window. NESO’s own work recognizes that future data center requirements need better coordination because the scale and location of applications can create uncertainty for network planning.

Scale remains strategically important, but scale without a credible delivery path can lose practical value when the network cannot support the requested capacity within the required timeframe. Hyperscalers have the financial capacity and global demand base to justify very large facilities, yet their requirements can involve long development cycles because electrical infrastructure must match a broader portfolio strategy and substantial site requirements. A neocloud’s advantage comes from being able to monetize smaller increments of compute capacity across customers that need access to AI infrastructure without waiting for a complete large-scale buildout. Ofgem’s connection reforms reinforce the importance of projects demonstrating genuine readiness, which makes the relationship between committed demand and deployable capacity increasingly significant. The regulatory direction does not establish a preference for one business model over another, but it changes the value of demonstrating that requested capacity corresponds to a project that can progress through defined milestones.

How Two Different Customers Tell Two Different Growth Stories

Neocloud demand tends to be easier to express through a sequence of operational requirements because the business model centers on providing specialized computing capacity to customers that may need access without owning an entire hyperscale platform. A UK deployment can therefore be presented through committed workloads, GPU procurement, contracted customers, and staged electrical requirements that correspond to identifiable phases of capacity delivery. Hyperscalers generally have substantial financial capacity and global demand bases that can support very large facilities, yet their requirements can involve long development cycles because electrical infrastructure must match substantial site requirements and broader development plans. Neither model automatically produces a stronger connection case because network planning must evaluate physical requirements rather than commercial branding. The critical distinction lies in how clearly each proposal translates future business growth into a sequence of electrical requirements that network operators can test against network capability.

Distribution network operators and NESO encounter these narratives from different positions within the electricity system, so the same customer proposition can create different planning questions at different network levels. NESO must consider transmission-level system requirements and the wider sequencing of connection projects, while DNOs manage distribution infrastructure and the local engineering constraints that determine whether demand can connect at a particular point. The connections reform timetable reflects this distinction by separating transmission and large embedded offers from distribution offers and assigning different delivery windows to each category. A project with a clearly defined first phase can give both levels of the network a more precise starting point, while its later phases can remain subject to future engineering assessment and system conditions. This matters for AI developers because a commercial growth plan does not automatically translate into an immediate electrical entitlement at every stage.

The Grid Is Choosing a Different Kind of Customer

The emerging UK connection model is less about choosing between neoclouds and hyperscalers than about distinguishing different forms of demand certainty. A customer asking for 20MW today with a documented plan for additional capacity gives the network a different planning object from a customer reserving a much larger future requirement without equivalent evidence of near-term deployment. The distinction becomes more important as data center applications consume a growing share of available connection capacity and compete with other electricity demand and generation projects. NESO’s reformed pipeline similarly places greater emphasis on projects that are capable of moving through defined stages rather than simply holding a position in a large queue. The commercial consequence is straightforward: AI infrastructure developers must increasingly demonstrate not only how much electricity they want, but why they need it, when they can use it, and what they can prove at each stage.

Ultimately, the UK’s AI infrastructure market will be shaped by the interaction between computing economics and electrical system constraints rather than by the largest announced capacity figure. Neoclouds that can translate demand into phased, energizable blocks may be well positioned in a system that places stronger emphasis on project readiness, while hyperscalers can continue pursuing large-scale capacity where the network, site, and investment case support that trajectory. The more important variable is whether the customer’s deployment sequence provides network planners with a credible basis for allocating capacity, scheduling reinforcement, and reassessing subsequent demand. Britain’s policy direction already points toward tighter scrutiny of speculative data center applications as connection demand expands and the system faces increasing pressure to ensure scarce network capacity is allocated to viable projects.