Google's AI Power Fix: Drill Into a Volcano, Melt Rock
Google's agreement with Quaise is, per Bernstein, 'a bet on a first-of-a-kind technology' - and it shows how far hyperscalers will go for electricity.

Google has signed an agreement with Quaise, a company using rock-melting drilling technology to reach superhot, volcanic geothermal energy, a deal a Bernstein analyst called 'a bet on a first-of-a-kind technology.' For decision-makers, it signals that AI's electricity bottleneck is now being attacked with unproven science, not just cheap solar and wind.
Google has struck an agreement with Quaise, a company chasing one of the strangest ideas in energy: drilling into the planet with rock-melting technology to reach the superhot energy trapped deep underground, including in volcanic formations. The deal is Google's latest attempt to lock down electricity for the data centers that power its AI models. And the reaction from Wall Street was not a shrug. A Bernstein analyst called the agreement "a bet on a first-of-a-kind technology." That phrase carries real weight. First-of-a-kind means no one has proven this works at commercial scale, and Google is effectively buying an option on a physics experiment.
Why would the search giant take that bet? Because AI has turned electricity from a boring line item into a strategic bottleneck. Training and running large models consumes enormous amounts of power around the clock, and the grid in most developed markets is not adding generation fast enough to keep up. Historically, hyperscalers solved this by signing long-term contracts for solar and wind. Those remain cheap, but they are intermittent, and a data center cannot pause mid-training because the sun went down. That mismatch, between always-on demand and weather-dependent supply, is the problem every big tech power strategy is now trying to solve.
Quaise's answer is to go where the energy is hotter and more constant. For context, conventional geothermal today typically means drilling a few kilometers into the crust to reach hot rock and using that heat to spin turbines. Go deeper and temperatures climb sharply. The catch is that steel drill bits historically fail under those conditions, wearing out or melting before they get far. Quaise's approach, as the company has described it publicly, replaces the mechanical bit with high-power millimeter waves, essentially industrial microwaves, that vaporize rock instead of grinding it. The prize is superhot rock, often located in volcanic regions, where the energy density is far higher than in conventional geothermal wells. The volcano in the headline is not a metaphor.
Google has historically ranked among the most aggressive corporate buyers of clean power, and its recent dealmaking has stretched across solar, geothermal, and advanced nuclear. That portfolio approach is deliberate. No single technology can be scaled fast enough to meet AI demand alone, so the strategy is to place many small bets and let the winners emerge. A first-of-a-kind drilling agreement fits that logic: the downside is contained, and the upside is a source of power that does not depend on weather, pipelines, or foreign fuel.
The risk side of the ledger is just as real. First-of-a-kind technologies historically take years, often a decade or more, to move from pilot projects to commercial scale, and the path is littered with cost overruns and engineering surprises. Bernstein's word choice, "a bet," is a polite way of saying the outcome is uncertain. Google is not buying guaranteed megawatts here. It is buying early access, technical learning, and a seat at the table if the technology works.
That has second-order consequences well beyond one company. Utilities and grid operators are already fielding record volumes of interconnection requests from data center developers, and regulators in several markets are debating who pays for the transmission upgrades that AI demand requires. If deep geothermal becomes viable, it changes the map of where power can be generated, opening up regions with hot rock but no gas, coal, or strong sun. If it does not, hyperscalers will lean even harder on nuclear and natural gas, and the emissions math for AI gets messier.
There is also a competitive dynamic at work. Every major cloud provider is racing to secure generation, and the constraint is no longer capital. It is time. A solar farm or battery array can be built in a couple of years; a novel drilling technology cannot be rushed on the same schedule. That is why the smartest operators are stacking options across technologies rather than betting on one winner, and why an agreement like this one is best read as a hedge rather than a headline-grabbing commitment.
For executives and boards, the lesson is that electricity has moved from procurement to strategy. Companies that treat power as a commodity to be bought at the lowest price are now competing against rivals signing decade-long agreements for technologies that do not fully exist yet. The question is no longer whether AI is worth the energy. It is who will control the energy, and on what terms, when the demand arrives.
This story's Key Insights and Take-aways are locked.
Create a free account to unlock Executive Actions for one credit.
Register to UnlockAlways free for Executives Club members. Join the Club
More in Technology
Anthropic mathematician claims Claude helped crack the Riemann hypothesis
If the proof holds, AI has toppled math's most famous open problem; if it does not, it is a case study in announcing breakthroughs before anyone verifies them.
AI's $4 Trillion Wipeout Erases Nasdaq's Gains for the Year
A brutal week took roughly $4 trillion off the AI sector and left the Nasdaq flat for the year, and the damage ran the full length of the supply chain.
Anthropic mathematician says AI proved the Riemann hypothesis. Verification will take time.
A single unverified claim from inside Anthropic now sits against more than 160 years of open mathematics, and the gap between asserting a proof and surviving peer review is the entire story.


