Google bets on Voltus to power data centers with a virtual grid

Google just became the first named customer in Voltus’s new 'Bring your own capacity' program, and the setup is simple in concept but tricky in practice: Voltus will build a virtual power plant in PJM, the grid that covers much of the US East Coast, and Google will foot the bill for setting it up. In return, the extra capacity created by aggregating distributed energy resources like electric vehicles and smart thermostats will help run Google’s data centers in the region. Voltus says the plant should be operational in 2027 and could aggregate up to 100 megawatts of distributed energy resources each year. This is one of the clearest real-world examples yet of a tech giant trying to use flexible demand as part of its power strategy, instead of relying only on new generation and transmission projects.

That matters because the power problem for data centers is not just how much electricity they use, but when they use it. The grid has to be built for peak demand, not average demand. MIT Technology Review points back to a Duke University study from last year that found if data centers agreed to decrease energy demand for roughly 40 hours per year, about 100 gigawatts of new load could come online without requiring new power plants or transmission equipment. That is the lure here: if a facility can dial back during the handful of brutal hours when everyone else is maxing out the system, it may free up room for more computing the rest of the year. In plain English, the grid can handle a lot more if the biggest users agree not to slam it all at once.

But this is where the elegant theory runs into the messy reality of incentives. Data centers are not like a thermostat or an EV charger that can always be shifted around without consequences. Google itself has said there are limits on how flexible a data center can be, and not every facility will be able to ramp down power demand. The company already has agreements with utilities across the US to limit or shift its own energy demand, but it has also acknowledged that there is no single solution for expanding grid capacity. Michael Terrell, Google’s global head of advanced energy, put it this way in an emailed statement: “There is no one solution for expanding grid capacity and we’re continuing to explore all options, including the many avenues for load flexibility.”

Voltus’s program is an attempt to solve the incentive problem from a different angle. Instead of asking data centers to do all the flexibility themselves, the company is offering a way for them to finance flexibility on their local grid. Voltus announced the program in September and says it allows data centers to pay for other people to be flexible. Under the deal, Voltus will pay customers who agree to participate in the virtual power plant, and those customers will help provide the storage or demand reduction that can be tapped when the grid gets stressed. The idea is attractive because it turns a hard infrastructure bottleneck into something more like a market transaction. If you cannot quickly add wires, transformers, or a new power plant, maybe you can pay people with batteries, EVs, and smart devices to help you get the same result.

Still, the source points to a big unresolved question: will people actually sign up? That question is not abstract. A recent study in California looked at managed electric-vehicle charging, a model where people get paid to give up control over when they charge their EVs so demand can be smoothed out. With no economic incentive, only 1% of EV owners enrolled. Even at $40 per month, which the study says was about 15% of their power bill, only 4.6% signed up. The region is different, the program is different, and the Google-Voltus arrangement is not disclosing participant pay, but the pattern is useful: money helps, but not always enough. And in this case, the exact price is not public, which means the real participation test is still ahead of the business model.

The regulatory backdrop makes the stakes even bigger. One proposal in the US would let new data centers come online years sooner if they agree to lower demand when the grid is nearing its max. Texas has also moved, with a new law requiring large users to switch to backup power or curtail demand in emergency situations. That tells you the policy world is converging on the same idea from two directions: either data centers become flexible voluntarily, or they get pushed to be flexible by rule. Google’s deal with Voltus sits in the voluntary camp, but it is clearly designed to operate inside a broader world where grid access is becoming conditional, negotiated, and increasingly strategic.

There is also a social angle executives cannot ignore. A recent Gallup poll found that about 70% of Americans oppose AI data centers in their area. That does not automatically block projects, but it does explain why “we will figure out the power later” is not a very comforting pitch anymore. For operators, the message is that energy strategy has become a reputation strategy. For boards, the important question is not just whether a data center can be powered, but whether the company can secure local buy-in, regulatory tolerance, and enough participating households and businesses to make flexibility real. Google’s deal with Voltus is concrete, early, and unusually explicit about how a tech giant might try to buy its way into more grid capacity. The next test is whether the model works outside the slide deck, because if it does, it could become a template for data-center growth. If it does not, the sector is back to the same old fight over permits, power plants, and patience.