A demand-side framework for an increasingly intelligent electricity system
Before we build against the next increment of forecast electricity demand, how well do we understand the demand we are planning to serve?
The electricity industry has become extraordinarily sophisticated at forecasting demand, procuring supply, building generation, expanding networks and maintaining reliability.
Demand-side thinking is not new. Energy efficiency, demand response and the economics of “negawatts” have demonstrated for decades that the lowest-cost solution is not always another unit of supply.
What is changing is our ability to execute against that insight.
Smart meters, connected devices, electric vehicles, distributed generation, batteries, advanced controls and artificial intelligence increasingly allow us to understand and influence the characteristics of demand—not simply treat demand as an aggregate quantity that must always be supplied.
The demand-side opportunity is not new.
Our ability to automate and orchestrate it increasingly is.
A load forecast tells us how much. Strategy needs to ask what kind.
A forecast may tell us that the electricity system needs another 500 megawatts, one gigawatt or five gigawatts.
But those numbers can obscure strategically important differences.
Some demand is continuous. Some is intermittent. Some is highly time-sensitive. Some can shift by minutes or hours. Some can be stored. Some can be avoided through efficiency. Some requires exceptionally high reliability. Some has significant economic consequences if interrupted.
A megawatt of demand is therefore not necessarily economically or operationally equivalent to every other megawatt of demand.
That distinction matters because infrastructure decisions can last for decades.
Before asking how much new supply we need, we should increasingly ask:
What are the characteristics of the demand we are planning to serve?
EVs illustrate the opportunity
Electric vehicles provide an accessible example.
If millions of vehicles begin charging at roughly the same time in the early evening, they create one type of electricity-system requirement.
If those same vehicles can charge when renewable generation is abundant, when network capacity is available or when electricity is less expensive, they create a very different system requirement.
The underlying customer requirement—mobility—has not necessarily changed.
The characteristics of the electricity demand have.
Smart charging, automation and potentially vehicle-to-grid technologies begin moving us from simply informing customers about energy toward orchestrating energy around the outcome the customer actually requires.
This raises an interesting question for personalization:
What if the best personalized energy experience is one the customer barely needs to interact with?
Personalization then stops being principally about better communications and begins contributing to better system outcomes.
Data centers: characterize before you capitalize
The rapid growth of AI and data-center electricity demand makes the distinction even more consequential.
The obvious question facing utilities and planners is:
Where will all the additional electricity come from?
That is an important question. But it may not always be the first question.
Before building against a forecast, we should understand the characteristics of the demand we are building for.
I have worked with RE24Energy as the company explored access to the U.S. data-center market. Its work has included matching data-center electricity demand with renewable resources at a more granular, hourly level rather than treating energy only as an annual quantity.
That experience reinforces an important point.
Once we stop thinking about electricity solely as an annual volume, the timing and characteristics of both demand and supply become much more important.
Data centers combine computational workloads, cooling systems, redundancy, storage, onsite resources and different requirements for timing and reliability.
None of this means large data centers will not require substantial new electricity supply.
It means that a forecast for 500 megawatts of new data-center demand should not automatically be interpreted as 500 megawatts of identical, immutable demand.
Characterize before you capitalize.
Understanding the load may change the scale, timing or nature of the supply-side investment ultimately required.
AI changes the orchestration capability
Artificial intelligence becomes particularly interesting when the number of variables being coordinated becomes too complex for conventional rules.
An increasingly connected electricity system may need to coordinate distributed generation, storage, EV charging, building controls, weather, market prices, network constraints, data-center workloads and reliability requirements.
AI therefore has the potential to become more than an information or forecasting technology.
It can increasingly become an orchestration capability for complex and flexible energy systems.
The strategically interesting question is not simply:
Can AI predict demand more accurately?
It is:
Can increasingly intelligent systems help us satisfy more demand with the infrastructure and resources we already have?
AI becomes strategically important because it increases our ability to coordinate complexity—not because it eliminates the need to understand the underlying system problem.
Move upstream of the capital decision
The traditional electricity-system planning model can broadly be characterized as:
CONVENTIONAL PLANNING
FORECAST → BUILD → SERVE
As demand-side capabilities become more sophisticated, we have an opportunity to insert additional strategic steps before committing to incremental supply.
EBS VIEW
CHARACTERIZE → ORCHESTRATE → SATISFY → BUILD THE RESIDUAL
This does not eliminate the need for generation, transmission or distribution investment.
It improves the question being asked before those investments are made.
- What portion of the requirement can be addressed through efficiency?
- What portion is flexible?
- What can move in time?
- What can be stored?
- What can be coordinated?
- What can existing infrastructure accommodate?
Only after asking those questions do we determine the residual requirement that genuinely requires additional supply-side investment.
Utilities do not have a shortage of technology.
They may have a shortage of questions that come before the technology or capital decision.
The strategic opportunity is to move one step upstream and ask whether the problem itself has been framed correctly.
Toward Satiated Demand
I describe the emerging objective as Satiated Demand.
Satiated Demand does not mean suppressing consumption or denying customers energy.
It asks whether the underlying requirement creating the demand has been successfully satisfied before assuming that every forecast increment of electricity consumption requires an equivalent increment of new supply.
A household needs comfort, mobility, hot water and functioning appliances.
A commercial building needs productive space and reliable operations.
An industrial facility needs production.
A data center needs computational outcomes within defined performance and reliability requirements.
Electricity is an input used to satisfy those requirements.
The strategic opportunity is therefore to understand how intelligently those requirements can be satisfied before determining how much additional infrastructure must be built.
For decades, the electricity industry has become extraordinarily sophisticated at making supply respond to demand.
The emerging opportunity is to become equally sophisticated about the demand side.
The larger strategic question is no longer simply:
How do we produce enough electricity to meet forecast demand?
It is increasingly:
How intelligently can we satisfy demand before determining what additional supply we actually need?