How SMR Startups Are Using the Tesla Playbook to Disrupt Energy
Silicon Valley's favorite strategies — direct sales, vertical integration, iterative builds, and patient tech money — have arrived in nuclear, and the industry will never look the same.
The moment Sam Altman merged his SPAC with a nuclear startup named Oklo, it was clear something had shifted. Not just in who was funding nuclear energy, but in how these companies think about selling it. Nuclear used to be built by utilities for utilities, financed by regulated ratepayers, blessed by governments, and measured in decades. Then a group of engineers who grew up watching Tesla tear apart the car industry decided to apply the same logic to atoms.
The comparison is not just a slide deck flourish. The specific strategic moves that made Tesla successful — selling direct to customers, owning the supply chain end to end, building iteratively and learning in public, and securing early adopters wealthy enough to absorb first-generation premiums — are playing out in nuclear right now. Some of it is working. Some of it still needs to survive contact with a regulatory body that does not operate on Silicon Valley timelines. But the intent is unmistakable, and it is worth understanding precisely.
The direct model: skipping the utility altogether
Tesla did not sell cars through dealerships. It sold them directly to drivers, cutting out the middleman and controlling the full customer relationship. Oklo is attempting the same move in nuclear energy. Instead of building reactors and handing them to a utility, Oklo builds, owns, and operates its reactors, then sells power directly to customers under long-term contracts — what the company calls a “power-as-a-service” model. 💡
The implications are significant:
Utilities are not in the transaction at all
Oklo captures the margin from electricity sales rather than a one-time reactor sale
Customers like data centers get a predictable, locked-in power price for decades
Oklo can co-locate its Aurora Powerhouse directly on a customer’s site, eliminating grid interconnection costs
As of early 2026, Oklo has a customer pipeline of roughly 14 GW, including a supply agreement with data center giant Switch and a deal with Meta to supply 1.2 gigawatts to AI operations in Ohio. These are not letters of intent. They are structured agreements that include upfront funding from customers to accelerate construction.
This is, pretty explicitly, the Tesla pre-order model. Early adopters commit capital before the product ships, which funds the factory, which proves the concept for the next wave of buyers. It worked for cars. Whether it works for reactors, which have a dramatically longer build cycle and a much more complex regulatory environment, is the actual bet being placed here.
Think about what this means for the traditional nuclear industry: the assumption that utilities are the only viable customer for nuclear power has not just been questioned. It has been discarded. 🔬
Is this model genuinely viable, or is a 12-GW pipeline of signed contracts just optimism dressed in legal formatting?
Vertical integration: owning the supply chain like Apple owns its chips
Tesla builds its own battery cells, its own motors, its own software, and as of recently, its own chips. The point is not just cost control. It is speed. Every time you depend on an external supplier, you introduce a schedule risk you cannot manage. Kairos Power, the California nuclear startup that just became the only US company actively building an advanced SMR, understood this early. 🏭
According to E&E News reporting from July 2025, Kairos argues that vertically integrating its supply chain, doing everything from fuel fabrication to welding reactor vessels internally, gives it a better shot at staying on schedule. That is not a boast. It is a lesson learned from watching every other nuclear project fall apart when a single specialty subcontractor missed a deadline and the whole critical path unraveled.
Kairos’s vertical integration list is striking:
Fuel fabrication from its proprietary TRISO pebble fuel in-house
Reactor vessel welding performed internally rather than contracted out
Iterative reactor designs built at the same Oak Ridge, Tennessee, site, so infrastructure and personnel carry forward between projects
A direct revenue agreement with Google locked in before the first commercial reactor operates, providing financial certainty while construction advances
Oklo takes vertical integration a step further by closing the fuel cycle entirely. In March 2025, Oklo acquired Atomic Alchemy and subsequently announced plans for a $1.68 billion nuclear fuel recycling facility in Oak Ridge, Tennessee. The plan: convert the 80,000+ tons of spent nuclear fuel sitting at US reactor sites into HALEU to power its own Aurora reactors. That is not just vertical integration. That is turning a waste liability into a feedstock. Tesla wished it could recycle old cars into new battery cells this cleanly. ⚡
The strategic logic is identical in both industries: when you own the whole stack, you control your cost trajectory, and you build institutional knowledge that competitors cannot easily replicate.
Big Tech as the early adopter: who plays the role of the tech crowd that bought Teslas first
Tesla’s first customers were wealthy early adopters who could absorb a $100,000 price tag for a car that proved the concept for everyone who came after. SMR startups have found a customer base that is even more strategically useful: technology companies with virtually unlimited energy appetite, strong balance sheets, genuine carbon commitments, and the patience to fund first-of-a-kind builds. 🚀
The deal-making here has been extraordinary in scale:
Google and Kairos Power signed what Google described as the world’s first corporate agreement to purchase nuclear energy from multiple SMRs, targeting 500 MW by 2035, with Hermes 2 targeted for 2030 at the Tennessee Valley Authority grid
Meta announced agreements for 6.6 GW of nuclear energy projects in 2026, including the Oklo partnership
Oracle announced a gigawatt-scale data center powered by three SMRs, with CEO Larry Ellison stating construction permits were already secured
Microsoft signed a 20-year agreement to restart Three Mile Island at 835 MW, while also backing Oklo
These deals change the economics in a specific way. The tech companies are not just buying power. They are, as Google’s head of data center energy Amanda Peterson Corio said to CNBC, actively trying to help commercialize the technology: “We want to see how we can take this from something small to something bigger that we can deploy at scale.” That is a customer funding your learning curve. No nuclear company in history has had that.
It is also worth noting what kind of customer this is. A data center operator running on nuclear does not vary its load the way a grid operator does. It runs at near-constant draw, 24 hours a day, 365 days a year. That is the ideal customer for a baseload nuclear reactor, and it is exactly the profile that makes nuclear competitive against wind and solar in this specific application. The alignment between what SMRs produce and what AI infrastructure needs is probably the most consequential commercial development in nuclear’s history.
The iterative build strategy: failing fast, then failing less
Here is something Tesla did that is deeply unfamiliar in nuclear: it built cars that were not quite right, sold them to customers, used the feedback, and built better ones. The original Roadster was barely viable. The Model S was the first real product. The Model 3 was the one that changed the world. Each generation funded the next and taught lessons the previous one could not.
Kairos Power has adopted this strategy more explicitly than any other SMR developer. Its Hermes 1 reactor in Oak Ridge is a 35-megawatt thermal demonstration plant, deliberately non-power-producing, designed purely to prove the fluoride salt cooling technology and stress-test the supply chain. The Nuclear Regulatory Commission issued a construction permit for Hermes 1 in December 2023, the first non-water-cooled reactor approved for construction in over 50 years. 🔩 Then in November 2024, the NRC approved Hermes 2, a two-unit power-producing plant at the same site.
Kairos CEO Mike Laufer told CNBC that Hermes 2’s core purpose is to create a standardized reactor design that drives down the cost of future deployments. Each iteration is a learning vehicle, not a revenue vehicle. The Google deal is structured the same way — Google and Kairos bear the financial risk of the first-of-a-kind build, while TVA provides the revenue stream through a power purchase agreement. Future customers inherit a cheaper, better-proven design.
This is genuinely novel in nuclear. The traditional approach was to design everything perfectly on paper, then build it once and hope. Kairos builds small, learns expensively, and applies those lessons to the next build. The risk of the approach is that each build still takes years and costs hundreds of millions of dollars. This is not software iteration. You cannot ship a patch to a reactor. But the principle of structured learning across a product line, rather than treating every build as a unique custom project, is real progress.
Where the Tesla playbook could still break down
None of this means the outcome is guaranteed. And I think it is worth saying clearly where the analogy strains under pressure. 📈
Tesla’s most powerful advantage was speed. A car takes roughly 30 hours of manufacturing labor to build. A reactor takes years of construction and a regulatory process that the NRC measures not in sprint cycles but in 18-month review windows. The iterative model that works beautifully in software, and even in automotive, faces a different physics in nuclear:
Regulatory review of a new reactor design cannot be accelerated just because your customer has deep pockets
HALEU fuel, required by Kairos, Oklo, and TerraPower’s Natrium reactor, is still not available at commercial scale in the United States, and no tech company checkbook can manufacture that supply chain overnight
First-of-a-kind construction costs are genuinely high; Kairos’s first commercial units will not benefit from learning-curve cost reductions until multiple reactors are built
The NRC’s licensing process, even on an accelerated timeline, still requires years that Tesla’s product cycles simply do not
The Bloomberg feature on Oklo from October 2025 noted that critics worry Oklo’s political connections and fast-moving ethos could pressure the NRC to approve designs before they are fully proven. That is a real concern. The nuclear safety record is excellent precisely because the regulatory process is slow and rigorous. The goal is not to remove oversight. It is to rationalize it for new reactor types that have fundamentally different risk profiles than the 1970s designs the current framework was built around.
The SMR startups that will succeed are probably the ones that internalize both sides of this. Move fast where you can: supply chain, factory setup, customer agreements, software. Move carefully where you must: safety case, regulatory submission quality, fuel cycle planning. Kairos, which got its Hermes 1 permit in 18 months after submitting what the NRC described as a thorough application, seems to understand this distinction. Others that assume the Tesla playbook transfers wholesale to a nuclear regulator may find the analogy has limits.
The question worth sitting with is this: when the first American SMR actually produces commercial power, which company will own that moment, and will their approach look more like Silicon Valley’s or more like the traditional nuclear industry’s?