Big Tech is going nuclear in its pursuit of artificial intelligence. Power-hungry datacenters are just one part of a broader freak-out over how the US grid will handle even bigger loads including electric vehicles and re-shored factories, plus withstand extreme weather, all while decarbonizing at a reasonable cost.
The obvious answer is that we must build more supply. But we can also be much smarter on the other side: demand.
The grid is struggling to balance three imperatives: affordability, reliability and sustainability. In the five years through June, average household bills outpaced inflation in 14 states accounting for half of US residential electricity demand.
The big driver is utilities’ spending on transmission and, especially, local distribution networks, which increased at about 9%-10% per year between 2017 and 2022, according to Sector and Sovereign Research LLC. Yet performance has deteriorated: The average customer experienced 14% more outages in 2023 than a decade before, with each one lasting more than two hours longer. Utilities are about as popular as lawyers these days.
The go-to solution is to build more capacity (if NIMBYism allows). This comes with an inherent inefficiency since the grid must be overbuilt to handle peak demand, with that extra capacity sitting idle most of the time. Complicating this further, the timing and nature of those peaks is changing as renewable energy introduces intermittency and distributed resources like rooftop solar and new loads such as EVs shift demand patterns.
This is where smarter demand management comes in. Rather than just building over the peaks, we should flatten those peaks out. We do this already in limited ways, ranging from (often clunky) time-of-use pricing to programs that reward customers (typically commercial or industrial) for curbing demand on peaky days. But there is much more potential. (Disclosure: My wife founded a company developing aggregated, actively managed portfolios of distributed energy resources, known as virtual power plants.)
Consider Octopus Energy Ltd. Less than a decade old, it is now the UK’s largest electricity supplier (it entered Texas via an acquisition in 2020). Its in-house software, Kraken, is a customer relationship, enterprise management and grid optimizing platform all in one. Processing and forecasting a multitude of parameters including weather, detailed demand and grid congestion, Kraken enables Octopus to manage or nudge customers to be flexible in order to take advantage of cheaper prices when renewable energy is most productive and shift away from peak hours.
One product, Saving Sessions, will typically alert customers via an app to curb demand for a short period and thereby earn rewards; something Chief Executive Greg Jackson likens to daily specials in the supermarket, when you might spot a discount and alter your dinner plans. At the more sophisticated end, Octopus can link Kraken to your EV, learn your driving habits and manage charging on your behalf in order to top up your battery as cheaply as possible (it can also lease you the EV and help install a home charger). Octopus thereby effectively controls 1.4 gigawatts of aggregate demand, equivalent to about 3% to 5% of UK peak demand, that it can shift around to some degree to help the grid operator. And that’s just today, with EVs comprising less than 5% of the UK vehicle fleet, and only a subset of those signed up.
Similarly, Base Power Co., a start-up co-founded by Zach Dell — son of a certain well-known tech pioneer — installs heavily discounted, and oversized, batteries in Texan homes. Base is thereby building a grid-scale battery that just happens to be spread throughout homes and linked with software, minus the NIMBYism and bottlenecks. Base gets a resource to soak up excess solar and wind power cheaply and sell it when grid prices are high, which also helps to foster more renewable generation. In return, customers get cheaper bills and, with scars from the 2021 winter blackouts still fresh, a welcome backup power source. “Most people don’t care about the ESG stuff,” Dell tells me, referring to environmental, social and governance considerations (this is Texas, after all).
Other markets do value the environmental benefits, though. For example, Holy Cross Energy, a co-op serving about 42,000 residential customers in a patch of Colorado that includes Aspen, targets 100% clean energy by 2030. It offers no-money-down deals on home batteries and EV chargers, paid for over time with adders to monthly bills, in return for being allowed to manage them collectively to balance increasing renewable energy generation.
Possibly the biggest pool of value here involves building less grid infrastructure due to smoothing out the peaks in demand. These avoided costs, being a counterfactual, are also the hardest to quantify, though modeling suggests they could be substantial.
One much-discussed example from a few years ago estimated California could save $120 billion over 30 years. An analysis published last year by the Brattle Group, a consultancy, for Google concluded that a virtual power plant could meet the grid’s need for a given amount of resources at a 60% discount to the cost of a natural gas peaker and a 40% discount to a grid-scale battery — before accounting for climate benefits. Meanwhile, the International Energy Agency’s latest long-term projections anticipate demand-side measures reducing peak electricity demand in advanced economies in 2035 by 15%. As an aside, the fact that big technology firms like Amazon.com want to co-locate datacenters with power plants, and thereby skip grid fees, points to the value of avoided costs.
All these numbers can be debated. What cannot be debated is that power bills are a problem already but more electrification is imperative to meet climate goals. In addition, clean technology prices point down even as grid-power prices rise, strengthening the economic case for distributed resources in a self-reinforcing process. “It’ll just be cheaper to install solar and storage,” says Kevin Kang, an analyst at energy-sector analytics firm Enverus and co-author of a recent study projecting almost half of US households will adopt rooftop solar by 2050.
That fundamentally changes the fabric and nature of the grid, from the top-down approach built to handle ever-higher peaks to one awash with distributed resources and smarter loads that both draw upon and support the grid. New business models emerging from the likes of Octopus show what is possible, although quantifying and harnessing the full spectrum of value is as complex an undertaking as managing the grid itself. Maybe AI could help with that.
