Sean Kelly, co-founder and CEO of Amperon and a veteran power-markets trader, sits down with Rohit Yadav in this "AI Meets the Physical World" series episode to walk through where energy forecasting and grid intelligence actually stand today — across AI-driven load growth, weather volatility, the flexibility stack, hyperscaler and utility demand, and the broader grid transformation underway.

Whether you are a founder, operator, investor, or simply curious about where physical AI and the energy transition intersect, this episode will sharpen how you think about forecasting as the foundational infrastructure layer of a modern grid.

Chapters:00:00 — Introduction: The "AI Meets the Physical World" series00:44 — What Amperon is, and why AI was the thesis from day one01:32 — Scale and scope: 165 customers, 51 million meters, remote-first02:41 — AI as demand driver versus AI as the grid's stabilizer04:45 — A short history of forecasting: the 15-day model and the European model06:15 — Why weather got harder: one-in-100-year events every few months07:30 — Smart meters and electricity bills becoming mainstream conversation08:42 — Who Amperon serves: retailers, traders, IPPs, utilities, and data centers12:04 — What's next: becoming an end-to-end system of insight and action14:21 — Why the trading decision always stays with the customer16:39 — Grid reality: data centers, volatility, and demand response 2.019:17 — What it means for forecasting: EVs, rooftop solar, and home storage21:39 — The flexibility stack and the 50–100 hours that decide reliability23:26 — What's actually lagging: bureaucracy, interconnection queues, data, and talent25:59 — Why there will never be one global grid30:37 — Who captures the value: hyperscalers, utilities, incumbents, and startups33:01 — Amperon's superpowers: ensemble models, weather data, and AI in the workflow35:49 — Closing worldview: the problem and the solution are both AI

🔑 Key Insights You'll Walk Away With:➡️ The three structural forces reshaping how the grid operates compared to thirty years ago➡️ Why AI has to win as the grid's stabilizer over the next five years➡️ How energy forecasting evolved from a 15-day trading tool into grid-critical infrastructure➡️ Why Amperon runs six to eight models under every forecast, reweighted and retrained every hour➡️ Why flexibility is the only real alternative to blackouts — and the 50 to 100 hours that actually matter➡️ How data centers could become the new demand response, echoing the early-2000s playbook➡️ Why there will never be one global grid, and why local market expertise becomes a moat➡️ How hyperscalers and utilities operate on different clocks but represent two durable customer segments➡️ Why the same AI creating the demand problem is also the key to solving it

Links:Sean Kelly: https://www.linkedin.com/in/sean-kelly-0792626/Amperon: https://www.amperon.co/

Rohit Yadav: https://www.linkedin.com/in/rohityadav23/Newsletter: https://yadavrohit.substack.com/