Latest / The Series A Show with Fexingo: Early-Stage Funding, Pitch Decks, and Startup Milestones / Why AI Startups Are Getting Asked About Their Energy Plan
Transcript
- Lucas: If you're raising a Series A for an AI startup in Q3 2026, there's a new question showing up in every due diligence memo. Luna: Let me guess — something about energy? Lucas: Exactly. Not just 'are your cloud costs under control' but 'what is your energy contingency plan if your data center loses power for 12 hours?' Founders who don't have an answer are seeing term sheets pulled. Luna: That's a real shift. And honestly, it ties into something I've been thinking about with this show. We've been doing over 130 episodes now — all ad-free, focused on the real mechanics of building a startup. A handful of listeners chip in monthly through buy me a coffee dot com slash fexingo, and that's literally what keeps us from having to run sponsor spots. So if this kind of specific, unsponsored analysis is useful for what you're building, that's where the support lives. Lucas: Yeah, it's a small group but it makes a big difference. And it keeps us focused on exactly this — the stuff that matters for founders right now. So, the energy question. It's not hypothetical. On July 15, a single fallen power line in Loudoun County, Virginia, knocked out two major AI data centers for nearly 18 hours. Luna: Loudoun County — that's basically the data center capital of the world. What were the downstream effects? Lucas: Some startups that relied on those facilities lost access to their models for almost a full day. For companies that had committed to uptime SLAs with enterprise customers, that's a breach. One founder I spoke to said their Series A term sheet was revised to include a clause requiring a backup power source — not just a UPS, but a dedicated microgrid or a contract with a second facility in a different grid region. Luna: So VCs are treating energy reliability as a risk factor on par with key-person risk or market size. Lucas: Right. And it's showing up in valuations too. I've seen two startups recently — both in the AI infrastructure space — that had similar revenue and team quality, but the one with a documented energy resilience plan got a 15 percent higher valuation. That's not anecdotal; that's the market pricing in operational stability. Luna: Let's tie this to some numbers we're seeing today. NVIDIA is trading around 207 dollars, Palantir around 123. Both are heavily tied to the AI compute story. If investors start worrying about energy constraints, does that cap the upside for these stocks? Lucas: It could, but not directly. For NVIDIA, the bigger risk is that if data center buildout slows because of grid bottlenecks, demand for their chips might plateau. Palantir is more about the software layer, but they rely on customers being able to run large-scale deployments. So energy reliability matters for the entire ecosystem. Luna: What about the startups themselves? How are founders responding? Lucas: The smart ones are adding a slide to their pitch deck. It's not just a risk mitigation slide — it's becoming a competitive differentiator. They're showing which regions they've picked based on grid stability, what redundant power agreements they have, even which renewables they're sourcing. Some are co-locating with green hydrogen projects. Luna: That's a big pivot from even two years ago, where the conversation was all about compute cost per token. Now it's about compute reliability. Lucas: Exactly. And here's the concrete takeaway: if you're a founder working on your Series A pitch this month, you need to be ready to answer three things. One, what is your current power source and what happens if it fails? Two, what is your projected power draw at 100x scale — not just now, but when you have 10 million users? Three, what is your unit economics per kilowatt-hour? Luna: Per kilowatt-hour — that's a metric I haven't seen in many pitch decks. But it makes sense. If your margin relies on cheap inference, a spike in energy costs can destroy your gross margin. Lucas: We're already seeing it. A computer vision startup I work with had to relocate their inference pipeline from Northern Virginia to the Pacific Northwest because the local utility couldn't guarantee enough capacity for their expansion. That move cost them three months and half a million dollars in migration costs. Luna: Right, and that delay can kill a Series A timeline. So VCs are getting ahead of that by asking the energy question early. Lucas: And there's a regulatory angle too. The Loudoun incident triggered a FERC inquiry into data center interconnection standards. So we might see new rules in the next 12 months that mandate backup power for facilities above a certain size. That would directly affect any startup that rents rack space. Luna: Interesting — so not just investor pressure, but potential compliance requirements. That's a lot for an early-stage company to handle. Lucas: It is, but it's also an opportunity. The founders who already have this figured out are standing out. I talked to a seed-stage startup building a software layer that optimizes inference scheduling to reduce peak power draw. They're getting meetings with top-tier firms specifically because their product solves this problem for other startups. Luna: So energy isn't just a cost — it's becoming a product category. Lucas: Exactly. I think we'll see a wave of startups emerge that focus on energy-efficient AI models, power-aware orchestration, and even hardware that can throttle based on grid signals. Some of those may become the next breakout companies. Luna: And for the founders listening, what's the one thing you'd say to prepare? Lucas: Get your energy numbers before you need them. Calculate your current power usage per training run and per inference. Document your providers' uptime history. And if you can, get a letter of intent from a second facility in a different region. That single piece of paper could be worth a lot in your next valuation discussion.