AI Power Demand Just Turned Geothermal Into a Wall Street Story
Reuters reported Wednesday that Fervo Energy secured a $10.21 billion valuation after its shares jumped 33% in their Nasdaq debut, a strong signal that public markets are now pricing power supply as part of the AI trade. The Houston-based geothermal developer raised $1.89 billion in an upsized IPO, selling 70 million shares at $27 each, above its marketed range, before opening at $36. Reuters tied the listing directly to surging electricity demand from artificial intelligence, rapid data-center expansion, electric vehicles, and domestic manufacturing. Fervo develops advanced geothermal systems intended to provide round-the-clock carbon-free electricity, using horizontal drilling and multi-stage hydraulic fracturing at its pilot project. Its first commercial facility is expected to begin delivering power by the end of 2026, and the company says it has roughly $7.2 billion in potential contracted revenue backlog through power-purchase agreements. Alphabet, Southern California Edison, Shell, Breakthrough Energy Ventures, and Devon Energy are all connected to the company’s financing or agreements.
The interesting part of Fervo’s debut is not that investors like geothermal. The interesting part is why they suddenly care so much. For years, the technology story was told as if software lived outside the physical world. Cloud computing was weightless. AI was magic. Platforms scaled because code copied itself at almost no marginal cost. That was always an incomplete story, but the market could ignore the missing pieces while power demand stayed manageable.
AI is forcing the physical layer back onto the front page. Models need data centers. Data centers need electricity. Electricity needs land, transmission, permits, cooling, fuel, turbines, wells, transformers, and long-term contracts. The platform economy is discovering that the bottleneck is not just chips or talent. It is reliable power delivered at industrial scale.
That is why a geothermal IPO can become an AI story. Fervo is not being valued only as a green-energy company. It is being valued as a possible answer to a constraint that now sits underneath the entire AI buildout. If hyperscalers cannot secure steady power, the promise of endless compute becomes a spreadsheet fantasy. If they can secure it, the winners in the AI race may be decided as much by energy procurement as by model architecture.
This is where the public-policy angle matters. The AI boom is already being sold as national competitiveness, productivity, and technological destiny. But ordinary citizens experience the buildout through utility bills, local permitting fights, grid congestion, water fights, and the quiet transfer of infrastructure priority toward the largest customers. When a data center wants the power footprint of a small city, someone has to decide who pays for the wires, reserves, and backup capacity.
Geothermal may be a cleaner and more reliable piece of that puzzle than many alternatives. It can run around the clock, uses a smaller land footprint than sprawling renewables, and fits the political mood better than some older green-energy narratives. But none of that removes the governance question. A technology that helps solve a bottleneck can still become part of a bigger allocation fight.
The market is doing what markets do: finding the scarce input and bidding up exposure to it. The political system now has to do what it usually avoids: decide whether the costs of the AI power race are being assigned honestly. If utilities socialize the grid upgrades while private platforms capture the upside, the public will eventually notice. If states bend permitting and rate structures for data-center demand while households absorb rising bills, the AI revolution will feel less like progress and more like extraction.
Fervo’s IPO is a reminder that AI is not floating above the economy. It is drilling into it. The next phase of the platform story will be written in megawatts, interconnection queues, power-purchase agreements, and local rate cases. That is less glamorous than a chatbot demo, but it is where the real incentives live.