For most of the last century we built three great infrastructures independently.
Communications networks moved information.
Electricity networks moved energy.
Compute transformed information into decisions.
That separation is now ending.
Artificial intelligence, distributed energy resources, electric vehicles, batteries, smart appliances and ubiquitous communications mean these systems are no longer independent industries. They are becoming one integrated cyber-physical system.
Many people are beginning to recognise this from different directions.
Jensen Huang recently argued that electricity networks possess enormous amounts of unused capacity most of the time because they are designed for rare worst-case events. Rather than building ever more infrastructure simply to satisfy rigid reliability guarantees, he proposes that data centres should become flexible loads, accepting differentiated levels of service and reducing demand when required. In return they gain access to cheaper power and faster connections.
Meanwhile Bloomberg recently highlighted another fascinating development: the emergence of markets for compute itself. Rather than negotiating bespoke contracts, compute may increasingly be allocated through transparent exchanges using sophisticated combinatorial auctions capable of matching thousands of buyers and sellers simultaneously.
These are not isolated trends.
They are manifestations of the same underlying idea.
Every scarce resource eventually becomes programmable
Communications evolved from fixed telephone circuits to packet-switched Internet routing.
Cloud computing evolved from dedicated servers to virtualised, dynamically allocated compute.
Spectrum evolved from administrative allocation to sophisticated auctions.
Electricity is now following exactly the same trajectory.
Yet much of today's electricity market still behaves as though the Internet was never invented.
We continue to centrally optimise an entire country every half hour before broadcasting a handful of prices and expecting millions of independent devices to coordinate themselves around them.
It is remarkable that we would never attempt to operate the Internet this way, yet we still attempt to operate electricity like this.
The real product is not electricity
We often describe electricity as a commodity.
It isn't.
People do not buy electrons.
They buy heating, transport, refrigeration, lighting, manufacturing, computing and increasingly AI inference.
Likewise, organisations do not simply buy compute.
They buy the completion of workloads.
Both electricity and compute are service delivery systems whose value depends upon:
how much is available,
where it is available,
when it is available,
and how reliably it is delivered.
Those four dimensions cannot be captured by a single energy price.
Dynamic pricing is only one piece of the puzzle
Many discussions around electricity market reform begin and end with dynamic pricing.
Dynamic pricing is necessary.
It is not sufficient.
If prices alone solved coordination problems, communications engineers would never have invented routing protocols.
Modern electricity systems require at least three interacting mechanisms.
First, dynamic prices should continuously communicate the real-time value of energy throughout the network.
Second, dynamic capacity allocation should determine who has the right to consume or inject power when networks become constrained.
Third, fairness mechanisms should ensure that over time no participant is systematically disadvantaged simply because of geography, income or historical circumstances.
These three mechanisms work together.
Prices provide incentives.
Capacity provides physical guarantees.
Fairness maintains long-term legitimacy.
Remove any one of them and the system begins to fail.
AI makes this possible
Twenty years ago such a market would have been unimaginable.
Today every household possesses multiple Internet-connected devices.
Millions of assets can communicate continuously.
AI agents are beginning to negotiate on behalf of users.
Bloomberg's article makes exactly this point for compute markets: AI can transform high-level user objectives into expressive bids that sophisticated allocation mechanisms can optimise automatically.
Exactly the same concept applies to electricity.
Consumers should not manually chase prices.
Software should simply optimise against user preferences while respecting network constraints and contractual service levels.
Capacity should become a product
One of Jensen Huang's observations particularly resonated with me.
He argues utilities should offer different reliability contracts rather than assuming everyone requires identical service. Some users would happily accept lower guaranteed availability in exchange for earlier access or lower cost.
This idea extends far beyond data centres.
Electricity markets should explicitly sell:
Energy
Capacity (availability)
Reliability
These are distinct products.
Someone operating an aluminium smelter has different requirements from someone charging an electric vehicle overnight.
A hospital has different requirements from cryptocurrency mining.
Today's market largely pretends these differences do not exist.
Tomorrow's market should embrace them.
Fairness matters
Perhaps the biggest omission from many discussions is fairness.
Markets do not exist solely to maximise efficiency.
They exist to allocate scarce resources in a manner society accepts as legitimate.
If flexibility payments systematically reward wealthy households while poorer households bear higher risks...
If network constraints permanently disadvantage particular regions...
If participants repeatedly lose access despite behaving responsibly...
...then the market will ultimately lose public support regardless of its theoretical efficiency.
Fairness is therefore not an optional social add-on.
It is part of the engineering specification.
It is every bit as important as stability, reliability and efficiency.
The convergence has already begun
Communications networks coordinate information.
Compute coordinates decisions.
Electricity networks coordinate energy.
Increasingly these are becoming one programmable system.
The future electricity market will look less like today's wholesale market and more like the Internet:
continuously operating rather than periodically clearing;
distributed rather than centralised;
service-aware rather than commodity-focused;
software-defined rather than infrastructure-defined;
dynamically allocating energy, capacity and reliability simultaneously.
The technologies already exist.
The communications infrastructure already exists.
The compute exists.
AI is arriving rapidly.
The only component still largely designed for the twentieth century is the market itself.
That, more than any shortage of renewable generation, batteries or transmission lines, is why I believe electricity market reform has become one of the defining engineering challenges of our time.