AI is changing electricity demand faster than the grid was built to accommodate it. But the more interesting question may not be how much power data centers need.
It may be what that electricity is worth to them.
If AI workloads can economically justify power prices far above those of traditional industrial customers, that creates enormous economic headroom. Could that help pay for grid modernization, new capacity, flexibility and cleaner generation? And what happens if utilities, developers and investors build for demand that ultimately arrives more slowly, or not at all?
That is the question at the center of this episode.
Hans Royal began thinking about it while on sabbatical from Schneider Electric. He came across the traditional gas heat-rate concept and started asking whether there was a similar way to measure the conversion of electricity into AI intelligence. That led him to create Compute Heat Rate, a metric for estimating how much different AI workloads can afford to pay for power.
David Siap, Senior Manager for Energy at CREO, brings the other side of the equation. His work asks what it actually costs to add reliable capacity, how investors should evaluate the AI-driven demand boom, and which investments still make sense if the forecasts prove too optimistic.
Peter Kelly-Detwiler, co-founder of NorthBridge Energy Partners and a longtime demand-response operator, tests both ideas against the realities of power markets, interconnection, flexibility and ratepayer risk. Episode 976: Thursday, October …
Expect to learn:
🔹 Why AI may behave very differently from traditional industrial electricity customers
🔹 How Compute Heat Rate helps explain the economics of AI power demand
🔹 Why flexibility may depend more on grid access than electricity price alone
🔹 Which investments still look sensible if data center demand falls short of today's forecasts
🔹 How AI could become a source of capital for a stronger grid, and what has to go right for that to happen
If you're trying to understand where AI, power markets and clean energy investment are actually headed, this conversation gives you a much better map.
Are there other technologies you’ve scouted on the frontlines of the Clean Energy Revolution that you think we should be covering here on SunCast?
Hit us up - [email protected] with your feedback & recommendations.
If you want to connect with today's guest, you’ll find links to their contact info in the show notes on the blog at https://suncast.media/episodes/.
Our Platinum Presenting Sponsor for SunCast is CPS America!
You can learn more about all the sponsors who help make this show free for you at www.suncast.media/sponsors.
Remember, you can always find resources, learn more about today’s guest and explore recommendations, book links, and more than 950 other founder stories and startup advice at www.suncast.media.
We'd love if you'd leave us a 5 ⭐ rating & review and it's never been easier: https://www.ratethispodcast.com/suncast
Subscribe to Valence, our weekly LinkedIn Newsletter, and learn the elements of compelling storytelling: https://www.linkedin.com/newsletters/valence-content-that-connects-7145928995363049472/
You can connect with me, Nico Johnson, on:
Twitter - https://www.twitter.com/nicomeo
LinkedIn - https://www.linkedin.com/in/nickalus
(00:00) The Explosive Surge in Data Center Load Growth
(02:33) What Is a Megawatt Hour Worth to AI?
(05:53) Defining Compute Heat Rate: Converting Electrons to Intelligence
(09:14) Quantifying AI’s $5,630/MWh Willingness to Pay
(10:41) Frontier Models vs. Commodity Inference Workloads
(14:41) The Economic Paradox of Data Center Flexibility
(18:09) Speed to Power and Interconnection Delay Economics
(22:35) Beyond LCOE: The Marginal Cost of Firm Capacity
(27:41) How AI Capital Distorts Clean Energy Investments
(35:21) Grid Interconnection Rules and Jurassic Park Analogies
(38:19) How Training vs. Inference Loads Impact the Grid
(45:07) Proving Empirical Demand Response for Large Data Centers
(48:53) Phantom Loads and Capacity Price Spikes in PJM
(55:09) Social License to Operate and AI’s Power Future
