Earlier this month, a friend working in distributed energy operations at an investor-owned utility posted on social media to express gratitude. Half a million customers had shown up to voluntarily reduce or shift energy use during the regional heatwave that broke records and tested grid limits.
A few hundred commercial and industrial customers and a few thousand small businesses provided half of the total load reduction. The rest came from homes with grid-edge technologies, like home battery systems, smart thermostats, and electric vehicles. This all-of-the-above deployment in the face of emergency embodied the state of grid flexibility: both how it has evolved, and how far it has yet to go.
Making the most of this flexibility, though, will require addressing two pieces of unfinished business. The first requires us to unlock more of the underutilized distribution and transmission grids in their built form today — to use the system we already have for more hours of the year, and to avoid burdening it during the clogged, congested hours that create system delivery risk. The second calls on us to activate everything that is signal-accepting and signal-responsive: anything that can curtail, idle, pre-consume, or discharge, either for predictable and everyday demand, or in the precious hours that could avert an emergency.
The heart of the challenge statement has little to do with grid technology itself; we have the tools we need to use both distributed energy resources and large sources of load as grid resources. But their use is not yet widespread. The challenge is instead about securing the capital to bake technologies into grid solutions — and about who our champions will be. Who will guarantee that a broad, distributed choreography of flexible infrastructure assets can move in syncopated rhythm with the grid’s living system, and not just coexist beside it?
The framing in the early years of grid flexibility was binary: the status quo versus the insurgent technologies, like behind-the-meter storage and building controls. Those innovations mostly lived outside the system, in startup offices and pilot apps, aggregation launches, and informal market demos led by empowered consumers. They weren’t, for many years, employed by grid owners and operators.
And the result was cycles of problem-solving decadence: siloed action that cannibalized both nearly viable DER business models and the possibility of shared momentum between grid operators and grid innovators.
Today, though, grid owners and operators are getting involved, which means these technologies actually have a shot. But a crucial hurdle remains: Utility cost-plus accounting practices tell them nothing about the value to be had to society if they get this right. Cost-of-service ratemaking has little guidance forthcoming on how to best take advantage of grid flexibility, but it’s becoming clear that spending a little money now will save millions of dollars in five years.
And the same can be said for our regional grid operations. Forthcoming PJM capacity auction results are piling on price signals that we need to do more faster on the demand side.
The rise of grid-edge technology in grid operations
Slowly, in the rooms where grid operators are responsible for keeping things from falling apart in real time, DERs have started to earn a place by showing up reliably during actual system stress. Reliability operators are beginning to trust the technology, because its potential has now been demonstrated across multiple utility grids.
In California’s 2022 record heatwave, a mass text alert prompted rapid voluntary reductions in power use across millions of households, dropping peak demand by more than 2,000 megawatts in about 30 minutes. Rolling blackouts were narrowly avoided across the state — despite record demand of roughly 52 gigawatts — thanks in part to these demand reductions and battery support. And far fewer localized utility outages occurred.
The same year, in Texas, the legacy of failure was still raw. 2021’s Winter Storm Uri had left more than 4 million customers in the dark, many for over 72 hours, while industrial and data center loads rode out the disaster with backup power or contractual exemptions. Nearly 1,400 MW of large customer load was protected or voluntarily curtailed, which certainly helped the grid in the aggregate. But that privilege of choice did not extend to the everyday power user.
In the wake of that crisis, many households took control of their own energy. Thousands of generators, home batteries and rooftop solar systems were installed in 2022 — and my own work to enable market support to compensate people for their investment in grid reliability followed as well.
The rise of the data center power user
But since 2022, the evolution of the grid has been driven not by households considering DER purchases, but rather by large grid users like the data center development teams seeking power for artificial intelligence infrastructure.
These teams are moving with capital intensity and strategic clarity. They know what they need. They know that power is the constraint. And, crucially, they are a different group of power users than the millions of residential and small commercial customers who have been courted by DER technologists and program innovators of the past.
Data centers are a major source of load, and potentially a major source of flexibility as well. Think, for example, about the role of data centers in helping the grid ride out the heatwave earlier this month. If several sub-gigawatt commercial loads could provide half of what a stressed grid needed when demand spiked in the heat, that implies that curbing a single data center’s load for a few hours by 25% could prevent entire neighborhoods from having to minimize air conditioning use in that time period.
There is commonality between these sectors. When the grid cannot promise capacity in the near term, the instinct — for both individual households and for massive data centers — is to self-provide. Both groups are purchasing energy-efficiency software, backup or behind-the-fence generators, and solar and batteries; data centers are also exploring temporary and self-built substations.
But beyond self-provision, what are our options for bettering the power system? They include better grid access and compensation programs, social license-to-operate thinking inside tech companies, and better grid design models to integrate both grid-edge and large power user support, instead of creating thousands of grid-defection islands of various shapes and sizes.
The human context
We are at risk of witnessing the quiet birth of two energy systems: one that responds with infinite urgency to the demands of big data and another that fails the basic needs of people. And to prevent it, we must be intentional about the human experience of the grid.
The same level of coordination, orchestration, and responsiveness now being built into grid-edge DER architectures to improve peoples’ experience during grid stress, must extend to large sources of load as well. Namely, AI and conventional compute data center power systems — site controls, storage, grid-responsive intelligent software stack components, and even gas turbines — should also be flexible assets that support sites and be invited through innovative grid policies to freely share electrons or curtailment support to the grid.
Every major heatwave should trigger a system-wide response from all of these assets. Every event in which these assets show up can help the local grid operator manage real-time problems, improve short term forecasts, and transform outage management and prevention planning.
The good news is that the software exists for the grid to make use of these assets and the talent exists to integrate responsiveness into the standard operating plan for loads on grids. Those of us who have lived this world in its past and present are ready to shape its future.
In my own work, I advise teams building both distributed energy and AI campus tools. Some are proving that computation can flex without compromising AI workload performance. Others are designing portfolio solutions that will unlock thousands of grid-edge assets to respond reliably to system stress. Still others know that grids will fail people for many more years than they will fail well-capitalized corporations — remember how this coop survived Hurricane Sandy — and so they are building solutions for multi-day resilience at critical infrastructure and neighborhood levels.
At scale, the work is to motivate teams of talented communities, grid operators and technologists to shape how neighborhoods back themselves up, how emergency responders stay well through extreme grid restorations, and how we keep AI systems online when the grid is strained and alleviate that strain. If we are going to further legitimize AI as a people-enabling solution, we need to legitimize the participation of this technology with grid policy that also enables human welfare.
The need for proactivity
This moment is exciting for both the energy and compute communities I work in, these strange bedfellows now charting shared territory.
But it also tests the darker side of my thinking: that we may not act boldly until another round of grid failures forces us to. Rolling outages. Economic losses. Recovery efforts that stretch for days. With early DER programs, our stakeholders were deeply reactive, and my own work often came in the aftermath of crisis, like in the case of Texas. We cannot afford to be so with data center load integration.
When commentators point to the success of a DER program, they often talk about the numbers. Markets moved. Megawatts delivered. But I see what moved us: what had to fail and burn to push the work into being. I am animated by the probing questions I hear from my students in STEM programs: young people thinking about the energy sector not just as a career, but as a place to make meaning. They ask how to hold on to their values when institutional change is slow. They worry about wasting years working inside systems that lag behind what is possible. They ask me as their teacher, can we really do this better?
Of course they can. Every one of them will be a piece of the machine we are now designing as grid policy technologists to fix what is wrong. They will run power companies, serve on institutional investor boards and reshape power utility governance, lead regulatory agencies, cast deciding votes, and arrest decadence in human ingenuity. And by the time they have children and elders to care for in a storm, I hope that they will have figured out how to keep breastmilk cold, life support machines on, and families safe.
The question has never really been about what is possible. It is about choice. Who decides what gets built, who gets backed up, who gets left behind. The systems we design now — how they allocate resilience, how they value flexibility, how they respond under pressure — will become the lived reality of millions.
Arushi Sharma Frank is just a human being and a mom: one person who decided that her work in the world of energy must solve for both the AI boom and grid-edge resilience, so our kids can have the best AI-enabled cancer treatments and keep their lights on. Her bio is here, if you’d like to know more about her work. The opinions represented in this contributed article are solely those of the author, and do not reflect the views of Latitude Media or any of its staff.
