Powering growth without overburdening ratepayers 

A single connection request can now arrive measured in hundreds of megawatts. The hard part isn’t the engineering. It’s who pays, whether the load even materialises, and whether the public accepts the answer.

A single grid-connection request can now arrive measured in hundreds of megawatts. Data centres, low-carbon industrial projects, hydrogen, green metals: the new sources of demand are large, fast-moving and far less predictable than gradual residential load growth. Networks built their planning processes around that steady growth. These connections are firm requirements in the near term but progressively less certain in the out-years. That is exactly the kind of demand that breaks a planning process tuned for steady, forecastable growth.

Connecting them is a real but solvable engineering problem; networks know how to build capacity. The harder problem never appears on a one-line diagram: who pays, and whether the public will accept it.

Two customers, one bill, a lot of tension

The tension is simple to state. Serving a wave of energy-intensive projects requires billions in investment, and households are acutely sensitive to seeing it surface on their rates. Cost recovery for large reinforcements is never automatic: the network still has to demonstrate the spend is prudent and proportionate to the benefit before recovering any of it. In regulatory terms, the investment has to survive a prudence review (and, in many jurisdictions, a used-and-useful test) before a dollar of it reaches the rate base.

The mechanism deserves plain statement. If a 300 MW load cancels after the network builds the new bulk-supply point, the capacity doesn’t vanish from the books. Its depreciation flows into the rate base carried by the customers who remain. Stranded-asset risk is the precise channel through which households pay for growth that never arrived. That’s why load uncertainty is a financial question before it is a forecasting one.

Get the tariff design wrong and you create a political problem that no technical excellence will solve. A region can welcome industrial growth in principle and reject it in practice. The shift happens the moment residential customers believe they’re subsidising it. The perception alone, that ordinary bill-payers are funding a multinational’s data centre, can stall a project regardless of its underlying economics. Public consent, once lost, is slow and expensive to rebuild.

So executives aren’t really wrestling with “can we serve this load.” The real question is “how do we manage rapid, unpredictable, large-scale demand while protecting households from significant rate increases.” Those goals pull in opposite directions. The work is finding the path that honours both rather than sacrificing one to the other.

Where the load lands changes everything

Not all connection requests cost the same to serve, and the spread is enormous. A project sited beside existing capacity might need only modest reinforcement. The same project a few kilometres away might trigger a new bulk-supply point, fresh transmission capacity and a long, contested consenting process. When the delta runs to hundreds of millions, where a hyperscaler or industrial plant connects becomes one of the highest-leverage variables in the entire plan.

And the network can influence that variable rather than merely receive it. Guiding, sequencing and pricing connections can steer a prospective customer toward a node the network can serve cheaply. Every other customer benefits, because they would otherwise help fund the larger build. A connection offer isn’t a binary yes/no; it’s a price signal that shapes where load lands.

Phantom load: the queue is a set of probabilities

The queue itself creates a second uncertainty. Large customers routinely file connection requests with several utilities or jurisdictions at once, then commit to one. The queue therefore overstates real demand, sometimes dramatically. Planners call the difference phantom load. The discipline that follows is simple: a connection queue is a set of probabilities, not a sum. Probability-weighting the queue, rather than adding it up, is the difference between building for demand and building for paperwork.

That means the analysis can’t stop at “can we connect this.” It has to model marginal cost-to-serve node by node: the incremental reinforcement each candidate connection point triggers, traced through the network. It also has to test rival growth scenarios with proper sensitivity and uncertainty analysis. A 300 MW request that may or may not materialise in year four can’t sit in the plan as firm demand. Then the analysis has to trace how each scenario propagates through to both industrial and residential bills. The connection point easiest to engineer rarely protects the public’s willingness to accept the project. The modelling exists to find the lowest-marginal-cost viable connection. The gap between that and a merely feasible one is exactly the capital that ends up on someone’s bill.

Pricing the risk before pouring the concrete

Modelling tells you where the risk sits; contracts decide who carries it. The instruments are well established. Financial commitment milestones and capacity deposits make a speculative request expensive to park in a queue. Contribution in aid of construction (CIAC) shifts a defined share of the reinforcement cost onto the customer who triggers it. Take-or-pay or minimum billing demand provisions keep revenue flowing even if consumption ramps more slowly than promised. And flexible or curtailable connection agreements let a customer connect early on a non-firm basis while the network firms up capacity at its own pace.

None of these is new. What has changed is the scale of the bet each one protects, and the fact that the right mix is node-specific. A deposit calibrated for a 20 MW industrial connection is noise against a 300 MW campus that may trigger a new bulk-supply point. The instrument mix should follow the marginal cost-to-serve, which is exactly what node-by-node modelling reveals.

Social acceptability is a planning input, not a PR afterthought

The networks navigating this well treat rate impact and fairness as variables inside the planning model, not as something to manage with messaging after the decisions. By the time a decision reaches the public, the modelling behind it has usually already spent, or preserved, the room to protect households.

A fair tariff mechanism lets industrialisation proceed without quietly loading the cost onto households. It is an analytical design problem before it is a communications one. It turns on cost causation: how to allocate the reinforcement a new load triggers between that load and the wider customer base. Shallow connection charging socialises most of the network cost; deeper, cost-reflective charging assigns it to the customer who caused it. The choice between them determines whether a large connection proceeds cleanly or provokes a backlash. And it has to satisfy the regulator and the public at the same time.rcement triggered by a new load is allocated between that load and the wider customer base. Shallow connection charging socialises most of the network cost; deeper, cost-reflective charging assigns it to the customer who caused it. The choice between them determines whether a large connection proceeds cleanly or provokes a backlash, and it has to be defensible to a regulator and acceptable to the public at the same time. 

So how do you protect the household and the load?

Increasingly, that design question decides whether a growth strategy is viable at all. Bring social acceptability into the optimisation alongside cost and risk, as a constraint the plan must satisfy rather than a reaction to manage afterward. Then “can we afford to connect this load” and “will the public accept it” stop being two conversations in two different rooms. They become one analytical question, answered before the commitment rather than defended after.

How Direxyon does this. Direxyon’s Asset Investment Planning (AIP) platform models marginal cost-to-serve node by node. It runs rival growth scenarios through Monte Carlo simulation, so it treats a 300 MW request that may or may not materialise in year four as the uncertain thing it is: probability-weighted, not firm. Rate impact and social acceptability enter the optimisation as explicit constraints, alongside cost and risk. A value-based multicriteria framework puts fairness on the same analytical footing as engineering prudence. The output is concrete: a ranked view of candidate connection points by marginal cost-to-serve, a probability-weighted rate impact for each growth scenario, and a cost-allocation rationale that stands up in a prudence review. The platform produces it before the commitment, not afterward to defend it. 

Frequently Asked Questions