Mazama Energy’s $135M Superhot Geothermal Bet on AI Power
AI does not have a software problem. It has a power problem. Mazama Energy just raised $135M to drill deeper into superhot rock beneath an Oregon volcano.
AI does not have a software problem. It has a power problem.
Mazama Energy just raised $135 million to drill deeper into superhot rock beneath an Oregon volcano because a chatbot is useless when the grid cannot feed the data centre running it.
While half of Silicon Valley is busy funding another software layer for AI, Mazama is spending serious money on the physical constraint that could matter more than the model: electricity.
The $135 million bet is not really about geothermal
Mazama Energy announced $135 million in new capital, including an oversubscribed Series B, to push its superhot-rock geothermal project at Newberry Volcano in central Oregon toward commercial power. Centaurus Capital and Doerr Capital led the financing. ConocoPhillips, Shell Ventures, Khosla Ventures and Gates Frontier were among the backers.
That investor list matters.
You have oil majors, climate capital, John Doerr’s money and Vinod Khosla’s money all sitting at the same table. These people do not agree on much besides one thing: dependable power is about to become an absurdly valuable asset.
Mazama’s idea is straightforward enough to explain over a beer, though brutally hard to execute. Drill deep into very hot rock, engineer a reservoir underground, circulate fluid through it, bring the heat back to the surface, and turn it into electricity. Unlike solar, it does not clock off at sunset. Unlike wind, it does not care whether the breeze has had a sick day. Unlike gas, it is not dependent on a fuel pipeline or a politician pretending energy security is sorted.
The company says it reached 629°F, or 331°C, in an engineered geothermal system at Newberry in 2025 — which it describes as the hottest such system achieved to date. Its next target is more confronting: push beyond 750°F and develop horizontal wells capable of generating 15 megawatts per well.
Fifteen megawatts is not a rounding error. It is the difference between a science experiment and a useful building block for a power plant.
Mazama says its Ceres project, supported by the US Department of Energy, is intended to demonstrate generation in 2027. Its longer-term Newberry plan targets an initial 200-megawatt commercial project, with first power targeted for 2029 and full output in 2030. The company also says Newberry may hold more than 10 gigawatts of resource potential. That last number is an ambition requiring independent certification, not a contract-ready power supply. Sensible people should keep the distinction clear.
Why this round matters more than another AI funding headline
The AI boom has created a funny form of collective denial. Investors talk about models, chips, agents and data. Then, somewhere near the bottom of the pitch deck, they remember that all of it needs power.
Not theoretical power. Actual megawatts, at an actual site, on a timetable that does not stretch to 2040.
Data-centre developers have increasingly looked to natural gas because it is dispatchable and familiar. That is understandable. A data centre cannot tell its customers it is offline because the weather was moody. But gas projects still need pipelines, turbines, permits, fuel supply and often a grid connection that is already oversubscribed.
Superhot geothermal is trying to offer a different deal: firm, carbon-free power with a relatively compact land footprint. The appeal is not that it sounds green in a company annual report. The appeal is that it could run around the clock where power-hungry industrial loads need it.
The catch, naturally, is that the Earth does not hand out easy wins.
At these temperatures, drilling gear gets punished. Reservoirs need to be engineered, not merely found. Materials, well design, fluids and diagnostics all have to work in conditions that would happily destroy ordinary equipment. The Department of Energy has backed work at Newberry specifically because superhot enhanced geothermal systems still need to prove they can create commercially viable reservoirs under these extreme conditions.
So do not read $135 million and assume Mazama has solved geothermal. It has not. Read it as sophisticated investors paying for the next, far more meaningful test.
The overlooked point: this is an oil-and-gas capability transfer
The lazy take is that geothermal is the wholesome alternative to fossil fuels.
The smarter take is that the same industries that became very good at drilling, subsurface engineering, logging wells and managing ugly operational complexity may be central to making geothermal work at scale.
ConocoPhillips and Shell Ventures joining Mazama’s round is not merely a nice logo slide. It is recognition that subsurface capability has value beyond extracting hydrocarbons. If superhot geothermal succeeds, the winners will not necessarily be the companies with the prettiest climate branding. They will be the teams that can drill faster, survive harsher conditions and make a reservoir perform repeatedly.
Mazama has called its integrated system MUSE: drilling, reservoir development, diagnostics and heat harvesting assembled as one platform. That is the right instinct. In hard-tech businesses, the product is rarely the shiny component. The product is the whole operating system: hardware, field crew, supply chain, permitting, data, maintenance and commercial contracts all working without drama.
I have seen this mistake plenty of times in business. Founders think they have invented a product when they have really invented one impressive piece of a painful process. The market does not pay you for being impressive. It pays you for removing the entire headache.
Mazama’s real product cannot be “a hotter well.” It has to become dependable, financeable electricity.
This is where the hype needs a hard slap
There are plenty of reasons to be excited about geothermal. There are also reasons not to get carried away.
First, engineering success is not commercial success. A well reaching extraordinary temperatures proves capability; it does not prove that electricity can be sold cheaply enough, reliably enough and repeatedly enough to justify billions of dollars in future build-out.
Second, the numbers get bigger much faster than the evidence. Mazama’s 15-megawatt-per-well ambition is a target. Its 10-gigawatt Newberry resource estimate is a target awaiting independent certification. Its 200-megawatt commercial project is a target. Anyone treating those as operating assets is either confused or trying to sell you something.
Third, capital intensity changes the game. Software investors can back ten companies from a laptop. Geothermal needs land, rigs, specialists, equipment, permits and years of work before a meaningful revenue stream arrives. That means the best founders in this category need to be part engineer, part operator and part project financier. Being great at raising a seed round will not cut it.
But here is the contrarian bit: those difficulties may be precisely why this opportunity is interesting.
Easy markets attract a swarm of competitors. Hard markets create a moat out of competence. If Mazama can turn extreme drilling and reservoir creation into a repeatable module, it will not be competing with another AI wrapper launched from a co-working space. It will be competing in one of the few markets where execution remains deeply, stubbornly scarce.
What Mazama is really selling: certainty
Mazama is not selling heat from rocks. It is selling a possible answer to a much bigger corporate problem: how do I secure power I can count on?
That question is becoming central for hyperscalers, manufacturers, utilities, miners, hydrogen projects and anyone trying to build large, energy-intensive infrastructure. The value of reliable supply rises when the grid is congested and demand is surging.
This is why the location of the power source matters. A company that can deliver dependable generation near an industrial load may be more valuable than one with an impressive national resource map. The winning energy businesses of this decade will not merely generate electrons; they will solve timing, location, permitting and reliability.
Mazama has a long way to go before it earns that status. But the $135 million round says the market sees a credible path worth funding.
And that is a far more serious signal than another founder saying their AI product will change everything.
What this means for you
If you are a founder: stop treating infrastructure as somebody else’s problem. Whether you run a software company, factory or marketplace, ask what scarce input could throttle your growth in three years. It might be power, data rights, distribution, skilled labour or regulatory approval. Build around the constraint early, before it becomes expensive.
If you are an operator: separate impressive technology from an operating system. Make a list of every dependency between your product and customer value: suppliers, implementation, training, compliance, uptime, support and cash collection. Your business is only as strong as the weakest boring link.
If you are an investor: do not confuse a massive market with a solved business model. In hard tech, demand can be obvious and returns can still be terrible if the company cannot execute projects, raise follow-on capital or turn technical milestones into contracted revenue. Ask what must be true at each stage — and what proof would kill the thesis.
If you are building anything exposed to AI demand: price power and infrastructure risk into your plans now. Not once your landlord, cloud provider or utility sends you the bill. The cheap electricity era is not something you should assume will last just because your spreadsheet says it should.
Mazama Energy has not proven that volcano-adjacent rock will power the AI economy. But it has put $135 million behind a much sharper idea: the next great tech bottleneck may not be intelligence at all.
It may be the ability to keep the lights on.