What Demand Response Actually Looks Like at Scale

Most demand response programs are small and slow. A new generation of commercial platforms is changing what large energy users can actually do.

Ask most energy professionals to describe demand response and you will get a version of the same answer: utilities call large industrial customers on hot summer days and ask them to reduce consumption in exchange for bill credits. Aluminum smelters, paper mills, large refrigeration warehouses — facilities with interruptible loads that can go offline for a few hours without operational damage. The utility avoids firing up peaking capacity. The customer gets a check. Everyone wins.

That model is real and has been running for thirty years. It is also marginal in scale, slow in response time, and fundamentally limited to a small subset of industrial facilities. The question that PeakDemand AI — our Series A portfolio company — is trying to answer is different: what does demand response look like when you have automated, real-time load control across thousands of commercial and industrial facilities simultaneously?

The Commercial Building Gap

Commercial buildings — offices, retail, hotels, warehouses, data centers, healthcare facilities — represent roughly 18 percent of total US electricity consumption. Their loads are largely controllable: HVAC systems, lighting, refrigeration, and EV charging can all be adjusted without meaningfully disrupting building occupants if done with sufficient intelligence about timing and duration. The problem is that most commercial buildings have never been enrolled in any demand response program, and most facility managers do not have the tools to participate even if they wanted to.

The participation barrier is partly technology — building automation systems from different vendors often do not communicate with utility demand response signals without significant custom integration work. It is partly operational — a facility manager at a hotel chain does not have the bandwidth to manage manual participation across 200 properties. And it is partly financial — the historical payment structures for commercial demand response have not been compelling enough to justify the setup cost for smaller facilities.

The addressable demand response resource in commercial buildings is roughly 80 GW nationally. Current enrollment is under 10 GW. That gap is not primarily a technology problem — it is a product and market development problem.

What Automated Load Control Actually Does

The core innovation in modern commercial demand response platforms is automated, sub-minute response. Traditional programs require a utility to call a customer, the customer to manually implement load reductions, and a verification process afterward. Total response time: 30 to 60 minutes. Effective for dispatch of large industrial loads. Useless for the frequency regulation and fast-ramping services that a high-renewables grid increasingly needs.

Automated platforms connect directly to building management systems via standard protocols — BACnet, Modbus, OpenADR — and can implement pre-programmed load adjustments within seconds of receiving a signal. A commercial HVAC system with a standard temperature setpoint of 72 degrees can pre-cool a building to 70 degrees in the two hours before a demand event, then let the temperature drift to 74 degrees during the event without running compressors. Building occupants notice nothing. The grid gets 15 to 30 percent load reduction for two to three hours at a critical moment.

The Market Structure Is Improving

Several developments have made the commercial demand response market meaningfully more attractive over the past two years. FERC Order 2222, despite slow implementation, has created clearer pathways for aggregators to stack multiple small commercial loads and participate in wholesale capacity and energy markets. Some independent system operators — PJM, CAISO, ERCOT — have improved their dispatch interfaces and settlement processes for demand response resources, reducing the administrative burden for aggregators.

More importantly, the economic value of demand response has increased in high-renewable grids. As wind and solar generation ramp up, the frequency of grid stress events — periods where supply and demand imbalance requires fast response — has increased in most markets. Capacity market prices, which compensate demand response providers for being available to respond, have risen in markets like PJM. The revenue per enrolled megawatt is higher than it was five years ago, which improves the business case for investing in customer enrollment infrastructure.

The Enrollment Challenge Is the Real Work

Building a demand response platform that technically works is the easier part. The harder part is building an enrollment and customer success operation that can acquire, integrate, and retain thousands of commercial facilities. The economics require scale — a platform with 500 MW of enrolled load has a fundamentally different unit economics profile than one with 50 MW, because the fixed costs of market participation, settlement, and operations do not scale linearly with load.

PeakDemand AI has been building enrollment infrastructure with a focus on portfolio accounts — property management companies, REIT operators, retail chains, and healthcare systems that control dozens or hundreds of facilities. Landing a single corporate account can add 10 to 50 MW of enrolled capacity with a single contract. The customer success model then focuses on performance and reliability: making sure the automated dispatch events work seamlessly, monitoring for false positives, and ensuring the facility staff understand what is happening and why.

Churn risk in demand response is real. A facility manager who experiences one unexpected dispatch event that disrupts operations will un-enroll and not come back. Building reliable, transparent automation that respects operational constraints is the product requirement that everything else depends on.

What This Means for the Grid

A commercial demand response market with even 50 GW of fast-automated load control capability would change the grid's ability to absorb variable renewable generation materially. It would reduce the frequency of curtailment events, reduce peaker dispatch, and lower the cost of maintaining grid reliability. That benefit flows to all electricity consumers, not just the facilities enrolled in programs.

We think about this as the demand-side complement to storage. Storage shifts energy across time. Demand response shifts consumption across time. Together, they address the intermittency challenge without requiring more generation capacity. The investment thesis is straightforward — and the market size, if you can actually unlock it, is large.