1. The Prompt for My Position: The AI Energy Tsunami
My decision to invest in Oklo wasn’t just about nuclear. It was about anticipating the second-order effects of the AI revolution.
In late 2024, I took a small but deliberate position in Oklo—not because they had revenue (they don’t), and not because they were already producing energy (they aren’t)—but because they sit at the intersection of two undeniable trends:
- The exponential growth in AI model inference and training demand, and
- The desperate global need for clean, compact, reliable, and scalable energy.
Let’s be clear: Large Language Models (LLMs) are not free to run. Every ChatGPT query I make today consumes 10x the energy of a traditional Google search.
Now extrapolate that:
- Millions of AI agents working 24/7
- Real-time inference across billions of endpoints
- Autonomous robotics, smart infrastructure, neural rendering, simulation, and more
I’ve seen it firsthand. My own Google searches dropped by over 90%, replaced entirely by LLM workflows. I’m just one person. Imagine what happens when 100 million people change their behaviour.
2. The Coming Energy Bottleneck
Between 2025 and 2030, we are going to see a megawatt-hour explosion driven not just by more data centres, but by:
- Edge AI compute and 5G/6G real-time inference
- Increased power density per rack (accelerated by GPUs, TPUs, and custom silicon)
- AI model retraining and fine-tuning in enterprise contexts
- Autonomous systems operating continuously in physical space
This won’t just strain grids. It will transform them.
The new constraint won’t be how much compute you can deploy—it’ll be how much power you can provision, how fast, and how sustainably.
That’s why I’ve been scanning the energy sector for over two years looking for a solution.
3. Why I Chose Oklo
There are many ways to play the AI–Energy theme:
- Invest in utilities (like NextEra Energy or Brookfield)
- Back gas explorers and pipelines
- Look at battery storage, grid software, or HVDC transmission
But none of those models fundamentally solve the problem of:
“How do we create dense, clean, always-on power at the data centre edge, without waiting 10 years for approvals or grid upgrades?”
This is where Oklo shines.
✅ Small Modular Reactors (SMRs)
Oklo is pioneering compact, fission-based nuclear reactors that can be deployed like modular units. Their Aurora powerhouse is designed to deliver clean, safe, stable power for 10+ years with no refuelling, minimal waste, and a dramatically smaller footprint than traditional nuclear.
✅ Deep government roots
Their board and leadership include former U.S. Secretaries of Energy and top nuclear scientists. Their regulatory knowledge and access to Department of Energy (DoE) infrastructure are unparalleled.
✅ AI-native positioning
Most importantly, they are not trying to power towns—they’re trying to power data centres. Their early partnerships, including with OpenAI and private compute infrastructure investors, position them as a nuclear-native hyperscaler supplier.
✅ Networked deployment potential
Oklo’s modularity allows for clustering of reactors, adaptive to demand. You could build a nuclear-powered data centre campus with redundancy, load balancing, and edge replication.
4. The Case for Asymmetry
Let me be clear: Oklo is pre-revenue. This is not a G1 Compounder or even a G2 Future Leader. It is a G3 Moonshot.
But the asymmetry is what makes it powerful:
- Their market timing is impeccable
- Their TAM is growing faster than anyone expected
- Their product-market fit is becoming clearer by the quarter
And most importantly, their leadership and board connections suggest a real ability to win DoE grants, secure long-term Power Purchase Agreements (PPAs), and establish first-mover contracts with the Fortune 500.
Oklo passes what I call the “Speculative Snap Test”:
If Oklo succeeds, the world looks meaningfully different—faster AI, more local resilience, decarbonisation at scale.
And if it fails?
I’ve allocated with eyes wide open, within my high-conviction G3 exposure.
5. The Bigger Picture
This investment is also about how I think emerging infrastructure gets funded.
When the internet was born, no one knew which companies would win. But the early believers in internet infrastructure (Cisco, Equinix, Level 3) saw decades of asymmetric returns.
I believe energy is the internet infrastructure of the AI era. The companies who win here won’t just sell power—they’ll define where and how intelligence lives in the physical world.
Final Thought
Oklo is not just a bet on nuclear. It’s a bet on:
- The explosion of AI power demand
- The need for location-flexible, sovereign, and clean energy
- The underappreciated speed at which AI infrastructure is decoupling from legacy utilities
I don’t expect Oklo to be a straight line. But I do expect that over the next 5–15 years, some company will redefine nuclear for the AI age.
And I’d rather be a year too early than five years too
