Google Wants a Virtual Power Plant

Who Gets to Control the Switch?

Artificial intelligence needs electricity on an industrial scale, and the technology companies are beginning to look at our homes, businesses and everyday energy consumption as part of the solution. It may be ingenious, but it raises some rather uncomfortable questions about who benefits from the digital revolution.

There is something faintly amusing about the way we have been sold the digital future. We were promised a world in which information would flow effortlessly, machines would take care of the tedious business of living, and technology would make almost everything more efficient. The future, apparently, would be cleaner, faster and considerably more convenient than the slightly inconvenient present in which most of us still have to remember where we put the car keys.

What we were not told quite so enthusiastically was how much electricity this wonderful new world would consume.

Now Google is proposing something that sounds rather like science fiction, although the engineering behind it is considerably more practical. The company has announced an expanded agreement with energy technology firm Voltus to develop a virtual power plant capable of providing approximately 1.07 gigawatts of flexible electricity capacity across the PJM grid region in the United States. That is a tenfold expansion of an initial 100-megawatt arrangement announced in June 2026, with the expanded contract expected to run through 2040.

The plan is to aggregate energy resources from homes, businesses and industrial facilities, using equipment such as batteries, smart thermostats and systems that allow electricity consumption to be reduced or shifted when the grid comes under pressure. Google funds the arrangement, Voltus coordinates the resources, and participating customers receive financial benefits for making their energy consumption more flexible.

The companies describe this as a way to strengthen the electricity grid while helping to accommodate the growing power demands of data centres. It is an interesting development, and there is a genuine engineering argument in its favour. Yet I find myself looking beyond the technology to a rather more important question. When the electricity demands of the world’s largest technology companies begin to influence how ordinary people use energy in their own homes, what exactly are we building here?

The cloud has acquired an enormous appetite

For years, the internet has been presented as something almost weightless. We talk about storing information in the cloud, streaming entertainment, asking artificial intelligence to write our letters and summarise our documents, and allowing increasingly sophisticated software to perform tasks that once required human attention. The language makes it all sound wonderfully ethereal, as though our digital lives exist somewhere above the mundane concerns of the physical world.

They do not, of course. The cloud is a collection of buildings filled with computers, surrounded by cooling systems, electrical infrastructure and a considerable amount of hardware. Those computers need power every second they operate, and the facilities housing them must be designed to keep working reliably even when the weather, the electricity grid or the local infrastructure would prefer to have a quiet afternoon.

Artificial intelligence has intensified the problem. The expansion of computing capacity requires substantial investment in data centres, and those facilities need dependable electricity supplies. In regions where demand is already pressing against available capacity, securing sufficient power has become a practical obstacle to expansion.

The PJM electricity market, which serves approximately 67 million people across thirteen states and the District of Columbia, is one example of the problem. Demand from data centres, alongside other changes in electricity consumption, has put pressure on the system. New power stations and transmission infrastructure take time to develop, obtain approvals and construct. Technology companies, meanwhile, are operating in an intensely competitive environment in which waiting several years for a grid connection is not necessarily an attractive business strategy.

Google’s arrangement with Voltus is an attempt to address that bottleneck by making more effective use of resources that already exist. Rather than relying entirely on the construction of new power stations, the company can help finance a coordinated network of smaller energy resources that collectively provide capacity when it is needed.

It is a sensible idea in principle. If thousands of businesses can temporarily reduce consumption, batteries can discharge stored electricity, and participating households can adjust certain appliances, the combined effect can help relieve pressure during periods of peak demand.

But notice what has happened here. The demands of the digital economy are no longer simply a matter for the companies building the computers. They are becoming a consideration for the electricity system serving everyone else.

That is where the conversation becomes considerably more interesting.

A power station that does not look like a power station

The term virtual power plant can sound rather grand, conjuring images of a vast computer system directing electricity around the country. The reality is both more mundane and more sophisticated.

A traditional power station produces electricity from a central location. A virtual power plant brings together a collection of distributed resources and coordinates them so that they can provide a useful service to the grid. Those resources might include batteries in commercial premises, domestic energy storage systems, controllable heating and cooling equipment, or industrial processes that can temporarily alter their electricity consumption.

Imagine a hot afternoon when electricity demand is approaching the available supply. A conventional response might involve bringing additional generation online, importing electricity, or investing in infrastructure capable of meeting the peak. All of these options can be necessary, but they can also be expensive, particularly when the additional capacity is required for only a relatively small number of hours.

A virtual power plant offers another possibility. Some participating buildings reduce their consumption, batteries supply stored energy, and industrial users temporarily adjust operations where it is practical to do so. Software coordinates these individual contributions, allowing them to be offered to the electricity market as a combined resource.

No single household has to behave like a miniature power station. The collective effect is what matters.

There is considerable merit in making better use of existing infrastructure. Electricity networks have to be designed around periods of high demand, even though those conditions may occur for only a fraction of the year. Reducing the size or frequency of those peaks can help avoid some expensive investments and improve the utilisation of the system.

Google is therefore not necessarily buying electricity from every participating household in the conventional sense. It is funding an arrangement that makes electricity capacity available when the grid needs it, with Voltus coordinating the participating resources and their contribution.

That distinction matters. A gigawatt of flexible capacity is not the same as a gigawatt of continuous electricity generation. The resource may be available for particular periods, and its actual contribution depends on how much capacity is enrolled, how reliably it performs, and how long individual resources can sustain their response.

Nevertheless, the scale of the proposed arrangement is significant. Utility Dive reported on 22 October that Voltus expects to build towards the full capacity over approximately three to four years, with the agreement extending to 2040. The commercial terms have not been fully disclosed, so it would be premature to assume that every projected benefit will materialise.

The concept, however, is no longer merely an interesting experiment. It is becoming part of the commercial infrastructure being developed to accommodate the expansion of large data centres.

And that leads us to the part I find most intriguing.

When your home becomes part of somebody else’s power station

For anyone interested in off-grid living, energy independence or simply reducing household bills, the underlying technology deserves attention.

A household with solar panels and a battery can generate electricity, store some of it for later use, and reduce its dependence on the grid. With the appropriate equipment and contractual arrangements, it may also be possible to participate in programmes that reward customers for allowing their stored energy or flexible consumption to support the wider electricity system.

The household becomes more than a passive customer. It becomes a potential contributor to the management of the network.

That is a significant change in the traditional relationship between consumers and electricity suppliers. For generations, the arrangement has been fairly straightforward. A power station generates electricity, the network delivers it, and the customer pays the bill. There have always been exceptions, including private generation and industrial demand management, but the broad model has been one of centralised supply and largely passive consumption.

Distributed energy resources change that relationship. A battery in a home can provide a service to the grid. A business can earn money by agreeing to reduce consumption at certain times. A collection of small resources can be coordinated to perform a task that would otherwise require a much larger piece of infrastructure.

In principle, this creates opportunities for households and businesses to earn additional income, reduce bills and make better use of equipment they have already purchased. It could also help defer some of the costs associated with expanding the electricity network.

But there is a difference between participating in an energy market and surrendering control over the way one uses energy. The distinction depends on the contracts, the technology, the incentives and the degree of control retained by the customer.

A smart thermostat may make small adjustments to heating or cooling. A battery may discharge when the system requests it. An industrial business might temporarily alter a process that can safely be moved to another time. These arrangements can be entirely reasonable, provided the customer understands what has been agreed and receives fair compensation.

Yet I would want to know precisely what that agreement permits.

Can the customer override a request when circumstances change? How much notice is provided? Are there limits on how frequently equipment can be controlled? What happens when a battery is needed for the household’s own purposes? Are the payments predictable, or do they depend on market conditions that the average participant has little ability to understand?

These are not arguments against virtual power plants. They are the questions that should accompany any system that turns private equipment and household behaviour into commercial energy resources.

Technology companies are very good at making complicated systems appear effortless. That is often part of their appeal. The difficulty comes when the effortless experience conceals a contractual arrangement that deserves considerably more attention than the small print normally receives.

I would rather see a customer understand exactly what they are selling before their thermostat becomes an employee of Google.

The question of who pays

The strongest argument in favour of Google’s approach is that it may help reduce the cost of meeting electricity demand.

Building new power stations, expanding transmission networks and upgrading local distribution infrastructure can be expensive. If some of that investment can be avoided or deferred by coordinating existing resources, the savings could be substantial. The companies argue that participating customers can receive direct financial benefits and that a more flexible grid can help improve affordability for electricity users more broadly.

Those benefits are possible. They are not automatic.

The crucial question is how the costs and savings are distributed. A company may fund a virtual power plant, but that does not mean every cost associated with accommodating its electricity demand disappears. The grid still needs maintenance, transmission capacity, dependable generation and sufficient reserves to deal with periods when demand is high and flexible resources cannot provide enough support.

Nor does a private agreement, by itself, guarantee that electricity bills will fall. The final outcome depends on the rules governing the electricity market, the cost of the resources recruited, their actual performance, and the way savings are passed through to customers.

There is a broader issue here, too. The growth of data centres creates demand for electricity, and the infrastructure required to meet that demand has consequences for the surrounding community. New generating capacity may be required. Transmission lines may need to be upgraded. Local planning decisions may become more complicated. If the system is under pressure, the costs of maintaining reliability have to be met somehow.

The question is whether those costs are borne primarily by the companies creating the additional demand, or whether ordinary electricity customers end up sharing a disproportionate part of the burden.

Google’s agreement is an attempt to provide capacity through a funded arrangement rather than simply expecting the grid to accommodate new demand without additional measures. That is a constructive step. It would be unfair to dismiss it merely because the company is also pursuing its own commercial interests. Businesses are entitled to seek reliable electricity for their operations, and private investment can help solve genuine infrastructure problems.

But the public interest requires more than a reassuring announcement about affordability. It requires evidence that the arrangements actually improve the system, that the financial benefits are distributed fairly, and that the costs are not quietly transferred to people who have no meaningful opportunity to participate.

A household without solar panels, a battery or a smart heating system may have little ability to earn money through these schemes. A tenant may not be able to install the equipment even if they want to. A small business operating on tight margins may be unable to adjust its consumption without disrupting its work.

If the benefits flow mainly to those who already possess the necessary equipment, there is a risk that the emerging energy economy will reward those who can afford to participate while leaving others to pay their bills as before.

That is not an inevitable outcome. It is simply something that should be measured rather than assumed away.

A clever solution can create room for a bigger problem

There is another question that deserves attention, and it concerns the relationship between efficiency and growth.

Suppose Google’s virtual power plant works exactly as intended. It recruits sufficient resources, provides reliable capacity during periods of high demand, rewards participants and helps avoid some expensive infrastructure investment. The grid becomes more flexible, and additional data centres can be connected more quickly.

That would represent a genuine achievement.

But what happens next?

If the additional capacity makes it easier to build more data centres, and those data centres create still greater demand for electricity, the system may find itself repeating the same exercise on a larger scale. The immediate bottleneck is relieved, investment accelerates, and demand continues to grow. Eventually, further capacity is required.

This is not an argument that improving efficiency is pointless. Quite the opposite. A more efficient system is preferable to one that wastes resources unnecessarily. Better coordination, energy storage and flexible demand can all reduce costs and improve reliability.

The difficulty is that efficiency gains do not automatically reduce total consumption. If making a service cheaper or easier to provide encourages much greater use of it, the overall demand for resources can continue to rise.

We see this in other areas of everyday life. More efficient vehicles can reduce fuel consumption per mile, but if people drive considerably further, the total saving may be smaller than expected. Better insulation reduces the energy required to heat a home, but the result depends partly on how the household uses the improvement. Faster computers make new tasks possible, which in turn creates reasons to buy even more computing capacity.

Artificial intelligence could follow a similar pattern. If computing becomes more efficient, its applications may expand into activities that were previously too expensive or impractical. The total demand for computing power may therefore grow even while the energy required for an individual task falls.

Google’s virtual power plant should be judged partly against this wider context. Does it help create a more efficient electricity system that benefits everyone, or does it primarily remove an obstacle to continued expansion?

The honest answer may be that it does both. Commercial incentives and public benefits are not mutually exclusive. The challenge is to ensure that the public benefits remain substantial as the industry grows.

There is little point celebrating a more efficient way of accommodating demand if the resulting growth continually outruns the savings. We would simply be getting better at running faster towards the next infrastructure problem.

The return of energy independence

Living in rural Spain encourages a rather practical view of electricity. Away from the reassuring assumptions of urban infrastructure, energy becomes something you have to think about. Generation, storage, consumption and reliability stop being abstract policy questions and become part of the ordinary business of living.

An off-grid household has to make decisions that are largely invisible to someone who simply expects electricity to emerge from a socket whenever required. How much energy is available? What can be stored? Which appliances are essential? When should consumption be reduced? What happens if the weather refuses to cooperate?

These are not new questions. People who lived without reliable electricity learned to manage resources according to availability long before anyone thought of putting the word smart in front of a thermostat.

The difference now is that the same principles are being applied on a much larger scale, using digital systems to coordinate thousands of individual resources. A home battery that once existed solely to improve the household’s own resilience can potentially become part of a network providing a commercial service. A business can treat its ability to reduce consumption as an asset with financial value.

That creates an interesting tension between energy independence and participation in a centralised market.

A household may become more self-sufficient by generating and storing its own electricity, yet it may also find that the equipment it owns becomes more valuable when connected to a wider system. Participating in that system could provide additional income, but it may also introduce contractual obligations and new dependencies.

There is nothing inherently wrong with that arrangement. Independence does not have to mean isolation, and a well-designed energy market can reward people for making useful resources available to others.

The important point is that participation should remain a choice, and that the value of the service provided by the household should be recognised fairly.

For those who can afford the initial investment, distributed energy systems may offer a combination of resilience, lower bills and potential income. For those who cannot, the benefits will depend on whether the wider system becomes genuinely more affordable or whether the opportunities remain concentrated among equipment owners and commercial operators.

Energy independence is not simply a question of owning a battery. It is also a question of retaining meaningful control over the resources upon which daily life depends.

What happens when the grid is under real pressure?

The virtual power plant model has another limitation that is worth remembering. Flexibility is useful, but it is not unlimited.

A battery can discharge only the energy it has stored. A household cannot endlessly reduce electricity consumption without affecting comfort or essential activities. An industrial facility may be able to reschedule some operations, but not every process can be interrupted safely or economically. A network of distributed resources can provide substantial support, but its actual contribution depends on the availability and reliability of those resources at the moment they are needed.

This is why a virtual power plant should be viewed as one component of a wider electricity system rather than a replacement for everything else.

New generation will still be required as demand grows. Transmission and distribution networks will still need investment. Storage will be important, but storage does not create energy from nothing. It moves energy through time, and the system must first obtain that energy from somewhere.

The same applies to gas-fired generation, nuclear power, wind, solar and other sources. Each has different characteristics, costs, environmental consequences and reliability considerations. A resilient electricity system requires careful planning across all of them rather than a belief that one clever technology will make the difficult decisions disappear.

Google’s proposal is significant because it may demonstrate that distributed resources can be contracted at a scale relevant to major industrial consumers. If Voltus can deliver the promised capacity reliably, other companies may follow. Long-term contracts can provide the certainty needed to finance batteries and other equipment, allowing a market to develop around resources that were previously too fragmented to compete effectively.

That is potentially a meaningful change in the economics of energy infrastructure.

However, the proposed 1.07-gigawatt capacity is expected to be built up over several years. The full arrangement will have to demonstrate that it can perform at the promised scale, particularly during periods when the electricity system is under strain. Until that happens, the announcement should be treated as a substantial commitment rather than proof that the problem has been solved.

The distinction matters because energy infrastructure is not judged by how impressive it looks in a press release. It is judged by whether the lights stay on when everyone needs them.

The uncomfortable question of control

There is one final consideration that I think deserves more attention than it is likely to receive.

As electricity systems become more digital, the ability to coordinate energy consumption becomes increasingly valuable. A company capable of aggregating thousands of batteries, thermostats and industrial resources can influence when electricity is used, how much capacity is available to the market, and which resources are called upon during periods of high demand.

That is not necessarily sinister. Coordination is precisely what makes the system useful. But the more important a service becomes to the functioning of the electricity network, the more important its governance becomes as well.

Who determines the rules? Who has access to the data? What safeguards protect participating customers? What happens if the company operating the system changes its commercial priorities, suffers a technical failure or withdraws from the market? Can customers leave without unreasonable penalties? Are the arrangements transparent enough for regulators and the public to understand their consequences?

These questions apply to energy markets generally, not just to Google. The difference is that large technology companies have the financial resources, computing infrastructure and commercial incentives to become influential participants in systems that were once dominated by utilities and traditional energy suppliers.

We should not assume that private ownership makes an arrangement bad, any more than we should assume that a technologically sophisticated solution must be good. The relevant questions concern accountability, competition, resilience and the distribution of benefits.

Electricity is not an optional luxury in modern life. It supports heating, refrigeration, communications, healthcare, transport and much of the infrastructure upon which communities depend. Any system that influences its availability deserves appropriate oversight.

A virtual power plant may be an excellent tool. That does not mean the public should have to accept every commercial arrangement built around it without scrutiny.

So where might this all lead?

I suspect we are seeing the beginning of a broader change in the relationship between energy producers, consumers and the technology industry.

For decades, electricity demand was largely treated as something that the supply system had to accommodate. As computing, transport and heating become increasingly electrified, that approach is becoming more expensive and, in some places, increasingly difficult to sustain. The next phase will involve greater coordination between generation, storage, consumption and the infrastructure connecting them.

Virtual power plants are likely to become part of that development. They offer a way to make existing resources more useful, provide new revenue opportunities for participating customers, and potentially reduce some of the costs associated with meeting peak demand. Google’s agreement with Voltus could help establish a commercial model that other large electricity consumers will want to follow.

If it works well, the benefits could extend beyond the data centres that prompted the investment. A more flexible grid could be more resilient, and the ability to use existing infrastructure more efficiently could help avoid unnecessary expenditure.

But there is a condition attached to that optimistic outcome. The benefits must be demonstrated, the costs must be distributed fairly, and the growth in electricity demand must be managed rather than simply accommodated at any price.

It is entirely possible for a company to develop a genuinely useful technology while pursuing a commercial objective that does not automatically serve every public interest. That is not a contradiction. It is the reason regulation, transparency and independent scrutiny exist.

Google is entitled to find a better way of securing the electricity its data centres require. The public is equally entitled to ask whether the resulting arrangement makes the wider electricity system better for everyone else.

For those of us interested in energy independence, the development is worth watching closely. The prospect of households and businesses earning money by providing flexibility to the grid is attractive, particularly if it helps make distributed energy systems more affordable. Yet the long-term value will depend on the terms of participation and whether customers retain meaningful control over their own resources.

Perhaps the most revealing aspect of this story is that the technology industry’s future is becoming less abstract by the day. The growth of artificial intelligence is now influencing decisions about power stations, transmission networks, household batteries and the way people use electricity in their homes.

The cloud has arrived at the electricity meter, and it would appear that it has brought a rather substantial appetite with it.

The question is not whether we should use clever technology to manage our resources more effectively. Of course we should. The question is whether the technology serves a wider public purpose, or whether the public is gradually reorganised around the needs of the technology.

That distinction may become one of the defining questions of the coming decade. We are building a world in which our homes, our appliances and even our decisions about when to consume electricity may become part of an enormous interconnected system. There is real potential in that arrangement, but also a responsibility to ensure that convenience does not quietly become dependence, and that efficiency does not become an excuse for endless expansion.

I would like to think that we can have the benefits without surrendering the choices that make them worthwhile. Whether the people designing these systems share that priority will be rather more important than whatever they decide to call the next generation of smart infrastructure.

After all, a virtual power plant may be able to balance electricity supply and demand, but it cannot decide what a fair deal looks like. That remains a human responsibility… at least for now.