Why C&I storage is finally taking off

Kann sets the episode up with a cautionary history. In the mid-2000s the US solar market was mostly commercial and industrial, and the first big acquisition in the space was SunPower buying PowerLight, a C&I developer, for $330 million in 2007. Then residential solar scaled, utility-scale solar scaled, and C&I simply did not. Storage has so far repeated the pattern exactly: under 2 gigawatts on the US grid going into 2020, about 40 gigawatts by 2025, of which roughly 90% is utility scale and 10% residential, which as Hade points out leaves basically nothing for C&I. His diagnosis is that the residential share is not purely an economic decision, since much of it is bought for resilience, whereas C&I customers buy on economics alone and the economics were not good enough. The pioneers of the 2015-2020 era, Stem, Green Charge, AMS, went after demand charge management and sometimes came close to penciling, but the value proposition was an edge case that mostly needed state subsidy to work. His phrase for the average customer is that the juice was not worth the squeeze: you could build an NPV-positive project, and the bill savings still were not large enough to justify the time a business would have to spend on something that was never going to save the business.

On the revenue side, Hade counts four streams, and Kann inserts an important qualifier before he starts: rising retail rates are inherently good for behind-the-meter generation but not inherently good for storage, because what matters is how the rates are structured. Time-of-use arbitrage now exists in many utility territories where it did not a decade ago. Demand charges, the part of a commercial bill set by the single highest power draw rather than by total energy, have been rising faster than the energy component, which makes shaving them worth more. Ancillary services, once the whole market, are now a minor share. The fourth, capacity, is the one Hade says has changed most in 18 months: PJM capacity prices are up elevenfold in three years, and the auction has hit its price cap without clearing for several rounds running, which he describes flatly as a market not functioning properly. That dysfunction is precisely what creates the opening, because hyperscalers can contract bilaterally for capacity outside the capped auction and pay whatever they think an incremental megawatt is worth. Kann then presses the counterargument, that storage erodes its own revenue: in ERCOT the arbitrage spread has been flat to declining because so many batteries were built, and on a recent record net-load day 10 gigawatts of batteries dispatched and prices never cleared $250 a megawatt-hour. Hade concedes the point and generalizes it into the episode’s sharpest unresolved problem. Storage has demonstrably helped every grid where it has been deployed at scale, and in the markets where it helped most it has not necessarily been a good investment. He says market design has to fix that over the next five to ten years and that we are not there yet.

The cost side is where his case is strongest and also where measured results and forecasts have to be separated carefully. A typical C&I battery today runs about $800 per kilowatt-hour after the tax credit: $300 to $400 of hardware, about $100 of software, about $200 of installation and about $200 of transaction costs, meaning interconnection, permitting, financing, customer acquisition and the people who do all of it. The tax credit itself is the first piece of good news, 30% at base, plus 10% for domestic content and another 10% in an energy community, so up to 50%, and it came through the 2025 budget law unchanged when the solar and wind credits did not. Hardware is the second, down about 40% at the pack level and 30% at the system level in 18 months. But soft costs are the historical killer, because a C&I project carries roughly utility-scale complexity amortized over a far smaller asset, and this is where Hade’s specific bet lives: transaction costs are repetitive workflows, financial modeling, contracts, interconnection filings, and agents can be trained to do them. Projects his team runs today cost $25 to $50 per kilowatt-hour in that bucket against $200 historically, and he adds that they are not even good at this yet. He forecasts about $20 within three years. Installation he is visibly less confident about, calling it harder to predict and offering a learning-curve argument: today you pay an electrician both to learn how to install a C&I battery and to install it, and as volume arrives that halves, to maybe $100 over five years. Put together, $800 becomes something like $510, which he thinks pushes many more customers past the threshold where this gets interesting. His heuristic there is worth carrying: 5% off a customer’s net electricity spend makes the conversation worth having, above 10% is a really good project, and rising rates have moved the pitch from vitamin to painkiller.

The claim that C&I grows fastest does not rest on C&I being good. It rests on utility-scale being stuck. A new utility-scale battery joins the same interconnection queue as large loads, and Hade puts the average wait at about six years in PJM, maybe nine in CAISO and maybe four in ERCOT, so demand for batteries over the next three to five years has to be met behind the meter, where interconnection is fast. Kann’s objection is the obvious one and Hade does not answer it: the queue is long precisely because it is full of storage projects, and those will eventually come online too. The remaining problem is scale. Hyperscalers think in gigawatts, and a single one-megawatt commercial site is irrelevant to them, while a thousand of them is a competitive advantage. Bridging that gap is aggregation into a virtual power plant, which is why Hade’s company ended up inside Voltus, and it is also the part he explicitly says he does not understand, suggesting his boss come on the show to explain how bring-your-own capacity actually works. Underneath all of it sits an unresolved hardware gap in exactly the size range most C&I facilities need, which he can only say he hopes more manufacturers fill.