When I first read about Nvidia’s potential $3 billion investment in SB Energy, I couldn’t help but think back to the 2017 ICO craze. Back then, I was auditing utility tokens for a small fund, watching community Telegram groups light up over whitepapers that promised decentralized everything. But the real bottleneck wasn’t the code—it was trust. Today, the bottleneck isn’t trust either, at least not in the crypto sense. It’s something far more fundamental: electricity. And Nvidia, the king of AI chips, is quietly moving to own the grid.
Context: The Macro Energy Map
The news is deceptively simple: Nvidia is in talks to invest $3 billion in SB Energy, SoftBank’s renewable energy arm, to secure power for an OpenAI data center. The reported deal would bundle solar, storage, and GPU supply into one package. This is not a financial investment for immediate returns. Nvidia’s gross margin sits above 70%, and its cash pile is around $26 billion. A $3 billion outlay is a strategic hedge.
We’ve seen this before. During the 2020 DeFi Summer, I managed a fund allocating capital into Aave and Compound. The biggest friction wasn’t smart contract risk—it was user experience. People couldn’t figure out how to bridge assets. In the same way, the biggest friction for AI scaling today isn’t chip design—it’s power density. A single H100 GPU consumes about 700W. A cluster of 100,000 GPUs for GPT-5 training could draw 70 MW, or more. The International Energy Agency projects global data center electricity use could double to 1,000 TWh by 2026, roughly Japan’s entire consumption. The era of cheap, abundant compute is ending. The era of energy-aware compute is beginning.
Core: Nvidia’s Vertical Integration Play
Let’s break down the numbers. If SB Energy delivers 2 GW of solar-plus-storage capacity—a reasonable estimate for a $3 billion investment—that could power roughly 600,000 H100 GPUs annually. That’s far beyond what any single lab currently needs. This suggests Nvidia is thinking beyond training into inference, where power costs dominate over the lifetime of a GPU. In fact, over a 5-year lifespan, electricity can approach 50-100% of the hardware cost. Locking in cheap, clean power now is a way to lower total cost of ownership for OpenAI, making Nvidia’s GPUs more attractive than alternatives like AMD or custom ASICs.
But there’s a deeper layer. I’ve spent years watching how capital flows through infrastructure. In 2021, I invested $500,000 in Art Blocks generative art, focusing on community ownership. The lesson was: culture drives value, but culture needs a physical backbone. Nvidia’s investment is about building that backbone. The company is moving from selling chips to selling “AI factories”—integrated systems that turn energy into tokens. At GTC 2024, Jensen Huang repeatedly said, “AI factories are the new industrial revolution.” This deal is the first major step toward making that vision a reality.
History repeats, but liquidity decides the tempo. Right now, liquidity is flowing into energy assets because the market is waking up to the fact that compute is not free. The same pattern happened in Bitcoin mining: early miners who secured cheap hydro power (like in Sichuan or upstate New York) survived the 2018 bear market, while those who paid retail electricity rates got washed out. Nvidia is applying the same logic, but at a scale that dwarfs crypto mining.
Contrarian: The Decoupling Thesis
Here’s the counter-intuitive angle: I’m not convinced this investment will age well. The assumption that AI demand grows exponentially forever is baked into the deal. But what if the technology decouples? What if more efficient architectures (like sparse models or neuromorphic chips) reduce power requirements per unit of intelligence? Or what if OpenAI shifts to custom silicon, reducing its dependence on Nvidia? Then this $3 billion energy asset becomes a stranded cost.
Moreover, this deal centralizes power—both literally and figuratively. The crypto ethos is about decentralization, but Nvidia is building a gated infrastructure that only the largest players can access. It’s the opposite of permissionless. Culture is the code that compels human adoption. If the culture of AI becomes one of corporate monopoly over energy, we risk repeating the mistakes of the 20th-century utility oligopolies. The community will eventually rebel, just as they did against centralized exchanges after FTX.
Takeaway: Positioning for the Cycle
So what does this mean for us as crypto investors? The next cycle won’t be about which Layer 2 has the lowest fees. It will be about which projects understand the energy-compute nexus. Look for protocols that integrate renewable energy credits, or that offer decentralized compute markets that can source low-cost power dynamically. The real alpha is in the plumbing.
I’ll be watching Nvidia’s capital expenditure trajectory closely. If they start treating energy investments as a core part of their business, the entire AI stack gets revalued. The lesson from my 2022 bear market experience—where we held onto art blocks and focused on community trust—is that the fundamentals always win. Energy is the ultimate fundamental.
History repeats, but liquidity decides the tempo. Right now, liquidity is flowing into energy. Don’t fight the tempo.