Washington wants AI and firm power. The DOE closed a $26.5B loan package for Southern Company’s utility. Last week it committed up to $4.4B more to restart 3 shuttered nuclear plants. FERC rejected PJM’s “bring your own capacity” interconnection rules for large loads and sent it back to refile. This week it went even further: show-cause orders hit PJM, MISO, SPP, CAISO, ISO-NE, and NYISO simultaneously, demanding every major grid operator justify how large load interconnection costs get allocated. This may sound a little academic and political. The real world implication is the Michigan regulators approved weeks ago to let DTE serve 2.4GW combined to Google and Oracle data centers is now facing an open ratepayer revolt over a potential $6.4B bill. Whatever it decides in the next 2 months may reset the economics of data center deals.
The corporates energy teams are hedging. Google just bought a 1GW VPP from Voltus in PJM, running through 2040, 10x the size of a 100 MW deal it signed in June. Imagine a mid size warehouse or office building. The owner has never once thought about “load flexibility.” The HVAC runs on a timer, lights come on at 7am, and nobody’s tracking what happens on hot afternoons when the whole grid is straining. Then a Voltus rep calls. They’ll install a couple of connected thermostat and a small control box, point out you’ve got maybe 50-100 kW of load that could be nudged, pre-cooling 1 hour before 3pm instead of exactly at 3pm, delaying a chiller cycle 20 minutes during the 40 hours a year PJM actually gets tight. You almost change nothing 325 days a year. On the handful of days it matters, Voltus sends a signal, your systems respond automatically, and you get paid, a standing payment just for being enrolled and available, plus a bonus per event. You were never going to go find this deal yourself. Someone had to show up, install the hardware, handle the utility paperwork, and aggregate your kilowatts with thousands of other buildings’ until it added up to something PJM and Google actually wanted to buy. That aggregation is the entire business.
After enough years in this industry, it is easy to dismiss an idea fast “that is hard to scale”. Kudos to the Voltus team for proving they can aggregate this kind of distributed “micro flexibility” up to GW scale and sell to Google as a single contracted resource running through 2040. Small assets, stitched together, became infrastructure. That is good entrepreneurship and good hustle. Afterall, Costco started as a warehouse in an airplane hangar too. Maybe time to take another fresh look at financing industrial energy efficiency and load (process) flexibility in this firm power constrained world.
I spent this week in NY catching up with friends and mentors. The opinions seem to converge 1) we all want to invest in the physical layer, but everyone is a little afraid that the AI funded music will stop at some point and much of today’s equipment scarcity is not durable. 2) fish where the fish are. Capital at any moment is finite, so if it piles into the hottest trades, good businesses elsewhere with differentiated products and strong teams get ignored and undervalued. We’ve seen some of listed renewables traded at 20% discount to NAV. The private market has even more pockets like that, for anyone willing to look. 3) compounding rests on one fundamental assumption: don’t lose what you already have. : )