Doug Bernauer

The a16z Podcast | Participants: Doug Bernauer (Radiant), Drew Baglino (Heron), Erik Torenberg, Erin Price-Wright

1. Key Themes

The Grid's Real Problem Is Delivery, Not Generation

The conventional wisdom is that we need more power plants. The actual bottleneck is getting power where it needs to go. New generation capacity is being added at record rates, but the transmission and distribution infrastructure — built on century-old mechanical principles — cannot handle the demands of data centers, electrification, and reindustrialization.

"New power is not the problem. Delivery is the problem... This year is by far going to be the highest capacity addition to the grid ever. So new power is not the problem. Delivery is the problem." — Drew Baglino [00:47:21]

"The grid is breaking. We're so bottlenecked today on the lines that run crisscross across the country. I mean, it's this very complicated, giant organic machine." — Erik Torenberg [00:00:00]

The Grid Must Become Software-Defined, Bidirectional, and Decentralized

For decades, the grid has operated as a one-way, top-down, centrally planned system built on mechanical infrastructure with no real-time software control. The coming wave of generation from edges (solar, micro-nuclear, batteries) and loads that spike unpredictably (data centers, EVs) requires a fundamentally new paradigm: a software-defined, bidirectional grid that can respond dynamically.

"Part of that is because the underpinning of the grid is these mechanical systems that are not fast responding, that don't have a lot of telemetry... if there is software, it's very slow to respond. And so you don't, it is harder in a world like that to imagine a bidirectional grid." — Drew Baglino [00:24:02]

"The idea that the grid can grow and move from the edge is just not something that we've really been able to process for the last 50 years in the U.S. Like the grid has been a very top-down project." — Erik Torenberg [00:23:36]

Nuclear Is Pre-Industry — The Kitty Hawk Moment Is Now

There is no functioning nuclear industry in the modern sense. Multiple new reactor designs are being built from scratch and will go critical for the first time by July 4th. This is the equivalent of pre-Kitty Hawk aviation excitement. The category is real but still unproven at scale. The key unlock is treating nuclear reactors as manufactured products — built in factories, shipped on trailers, deployed in 48 hours — rather than mega civil engineering projects.

"There is no nuclear industry. That's really true. It's almost like we're getting excited about flight before Kitty Hawk, right?" — Doug Bernauer [00:14:50]

"We're doing nuclear reactors as products for the first time ever. And it's so that you can just say, yeah, I want nuclear power and we deliver it, it operates. And then when it's done operating, we take it away and there's no waste or other tricky consideration you have to make." — Doug Bernauer [00:29:13]

2. Contrarian Perspectives

Data Centers Will Actually Lower Electricity Rates for Everyone

The popular narrative is that data centers are driving up electricity costs for consumers. The opposite is likely true. Electricity rates = total grid costs ÷ total kilowatt-hours delivered. Data centers are near-constant, high-utilization loads — the ideal customer — dramatically increasing the denominator and spreading fixed infrastructure costs across more kWh.

"The data center customer is like the ideal customer. They're consuming like near their maximum power almost all the time compared to like your house where you're like maybe at 10% of the maximum power of your house, like an hour a day. So the more data center load, like more loads we have like data centers, like factories, that are steady constant loads, the cost of serving electricity to everybody will go down." — Drew Baglino [00:46:34]

Uranium Is a Free Resource Being Wasted — Leaving It in the Ground Is the Dangerous Choice

The cultural framing around nuclear waste treats uranium as inherently dangerous to use. Bernauer inverts this: uranium undergoes spontaneous fission regardless of whether humans use it. When it decays unused in the ground, it produces radon — which is actually harmful to human health. Using it in a reactor is the more responsible choice. We've already lost 99.2% of our original uranium-235 endowment.

"When the earth formed four billion years ago, we had like 128 times as much uranium-235. So we better get it before it's gone. Take it out of the ground, use it. Don't let it turn into radon and other stuff that is actually harmful to someone's health. It's like it's more dangerous from a health perspective to leave it in the ground." — Doug Bernauer [00:30:09]

EV Supply Chain Slowdown Is an Opportunity, Not a Problem

EV growth is decelerating, which the market reads negatively. But for power electronics manufacturers, this frees up a massive, already-scaled supply chain — semiconductors, capacitors, ferrites, manufacturing capacity — that can be redirected to an even larger opportunity: grid infrastructure, data centers, and industrialization.

"Electric vehicle growth is slowing down, which means we can take that momentum and bring it into a new problem statement, which is power for data centers, power for industrialization, power for economic growth and prosperity, and for sustainable energy." — Drew Baglino [00:43:44]

Data Centers Are Currently Destabilizing the Grid — But Can Easily Fix It

The received wisdom treats data center grid disconnects as a feature (uptime protection). At small scale this was fine. At gigawatt scale, instantaneous two-gigawatt disconnections are a grid stability crisis. The fix is not to restrict data centers — it's to add software-defined controls and modest storage so data centers actively stabilize the grid instead.

"They are designed to date to do that, right? They want to keep their compute up... But when you're building gigawatt data centers, it starts to really matter. And the grid stability is at risk. And so that is very solvable with software, modern power electronics, you know, dynamic grid forming controls." — Drew Baglino [00:44:43]

3. Companies Identified

Radiant Nuclear

Description: Startup founded in 2019 by Doug Bernauer (ex-SpaceX 12 years). Building portable, factory-manufactured micro nuclear reactors — one megawatt, trailer-sized, meltdown-proof fuel, deployable in 48 hours, operating for five years (equivalent to 2 million gallons of diesel). Building a production facility on an 80-acre site in Tennessee targeting one reactor per week.

Why Mentioned: Only company currently permitted to bring a new-design reactor to full power (not just criticality). On-schedule for 2026 full-scale reactor test. Positioned to serve off-grid markets (military, islands, disaster relief, industrial edge) where diesel costs exceed ~$6.50/gallon — a massive global addressable market.

"We are now the only reactor permitted, of these new reactors, to go to full power... And we're on schedule to do that, which is kind of wild." — Doug Bernauer [00:16:09]

"One per week coming off of a production line from our Tennessee facility, which is an 80-acre site we just signed for in October." — Doug Bernauer [00:17:05]

Heron (formerly referenced as Drew Baglino's company)

Description: Founded by Drew Baglino (ex-Tesla VP of R&D, nearly two decades). Building solid-state transformers — replacing century-old oil-filled magnetic transformers with high-frequency power semiconductors and software. First product: Heron Link, a five-megawatt bidirectional solid-state transformer (800–1500V DC to 34kV AC). First factory targeting 40 gigawatts per year of capacity.

Why Mentioned: Addresses the delivery bottleneck directly. Modular, fault-tolerant (30 modules per unit, fail-operational), software-defined. Targets the 500+ gigawatt and growing DC market (data centers, solar, batteries) first, then expands to AC-to-AC for full utility grid applications.

"We're building a 40 gigawatt factory and people are like, wow, that's a lot. It's about 10 to 15% of the ex-China market for our product category. And it's equivalent to half the state of Texas in peak power." — Drew Baglino [00:27:32]

"It's a five megawatt bidirectional solid state transformer that goes from DC anywhere from 800 to 1500 volts DC to 34,000 volts AC, which is effectively the sub-transmission voltage of data centers, of large battery power plants, of solar facilities." — Drew Baglino [00:18:44]

4. People Identified

Description: Founder and CEO of Radiant Nuclear. Spent 12 years at SpaceX (joined 2007, worked on first successful Falcon 9s, Grasshopper reusable rocket, Hyperloop, Boring Company, Mars colony design). No prior company founding experience — slow-rolled into Radiant after Elon Musk pointed him toward nuclear while he was stuck on Mars power systems using solar.

Why Mentioned: Combines rare deep regulatory experience (launched rockets from military bases, now navigating NRC), hardcore manufacturing thinking (one reactor/week production target), and long-term commitment (stated 2026 timeline in 2020 and is on schedule). Currently the only new-design reactor permitted to full power in the U.S.

"In 2020, I said in 2026, I will put a full scale reactor and get it critical and get it up to full power. And we're on schedule to do that, which is kind of wild." — Doug Bernauer [00:16:09]

Drew Baglino

Description: Founder and CEO of Heron. Spent nearly two decades at Tesla as VP of R&D, responsible for Megapack product development, manufacturing, and business. Co-authored Tesla's Master Plan Part 3 — the global sustainable energy feasibility study. Background in whole-system energy modeling (undergraduate thesis on New Zealand achieving Kyoto commitments as a microgrid study).

Why Mentioned: Brings rare combination of power electronics depth, large-scale manufacturing execution (Megapack), and whole-grid systems thinking. Actively working to reshore ferrite and thin-film capacitor supply chains to the U.S. — strategically important and underappreciated.

"I had the opportunity to work on this project called the Master Plan Part 3 project where we were effectively studying, okay, now let's do it for the whole globe. Can we have a sustainable all-electric or largely all-electric sustainable energy future? Is it feasible? Are the resources there? Is the investment reasonable? And the answer was like yes, resounding yes in many ways." — Drew Baglino [00:07:29]

5. Operating Insights

Automate Last — Earn the Right to Automate Through Manual Reps

Both Bernauer and Baglino explicitly follow the same manufacturing philosophy: build by hand first, learn what the process actually is, delete unnecessary steps, then and only then automate. This mirrors lessons from Tesla's Gigafactory evolution. Premature automation locks in bad processes at scale.

"You know the process, right? Automate is last... if you're trying to delete the steps and done all the other smart stuff." — Drew Baglino [00:34:15]

"Parts of the production line that can be automated and should be will then go in those newer buildings." — Doug Bernauer [00:34:35]

Minimize In-Field Work Content as a Core Product Design Principle

Reducing site work isn't just a convenience — it's a competitive moat. Less site work means faster deployment, lower total cost, simpler permitting, less community resistance, and faster revenue realization. Baglino credits this explicitly from Megapack learnings: eliminating concrete pads where seismically permissible, using soil nails instead, removing architectural scope. Bernauer's reactor deploys in 48 hours from wheels-stop.

"The less involved the onsite project is, just the faster everything about it will go... we even worked past the pad, like we got rid of the concrete, wherever the seismic would allow. It's awesome. It just did soil nails. Because again, you're disturbing less dirt." — Drew Baglino [00:31:15]

Find the One Miracle That Leads to Revenue — Then Earn the Next One

Both founders explicitly constrain their companies to solving one hard problem at a time. Bernauer declined to also solve permanent in-ground nuclear installations. Baglino targets the DC market first before AC-to-AC. The discipline of identifying which single miracle unlocks product-market fit and revenue — rather than stacking miracles — is a key operating principle.

"One miracle that leads to then a product and revenue, right? That's the way. And then you have time to think about another miracle." — Doug Bernauer [00:37:30]

"Series miracles. You don't want to have too many in a startup." — Drew Baglino [00:37:33]

6. Overlooked Insights

The DC Microgrid Convergence: Every New Technology Is Natively DC

This point was made almost in passing, but it is architecturally profound. Solar is DC. Batteries are DC. Compute (data centers) is DC. Micro-nuclear (Radiant) produces DC as its native output. Heron's first product converts DC to AC — but the deeper implication is that a fully DC microgrid combining all four is not only technically cleaner, it eliminates multiple conversion steps, reduces losses, and may represent the default architecture for edge deployments, space infrastructure, and eventually new grid segments. Nobody has named this convergence explicitly as a category.

"Compute also is natively DC... compute, batteries, solar — all the new technologies — micronuclear, yeah, actually all DC." — Drew Baglino and Doug Bernauer [00:25:32]

"Our reactor fundamentally makes DC power because we actually run a really compact power generator... we do it at such a high speed that we then use active rectifiers to convert to DC as the first step." — Doug Bernauer [00:26:23]

The U.S. Has Lost Its Power Electronics Knowledge Base — and Nobody Is Talking About It

Baglino briefly mentioned that talent from GE migrated to Korea, Japan, and China as those became the growth markets for power infrastructure. This represents a multi-decade hollowing out of institutional knowledge in exactly the domain now most needed — grid power electronics. This isn't just a supply chain risk; it's a human capital crisis that will constrain how fast companies like Heron can actually scale, and represents a structural moat for anyone who has retained or rebuilt that expertise.

"Decades of not a lot of change means like the brightest and most talented people are like, I got to go do something else more impactful with my life. And a lot of those folks actually, for example, a lot of the talent in GE went to Korea, right, or Japan, because those were the growth markets. And so you got this almost like hollowing out, or China, right, hollowing out of the knowledge base." — Drew Baglino [00:11:37]