Mazama raises $135M to drill superhot rock for AI data centers

Share
Mazama raises $135M to drill superhot rock for AI data centers

Two stories worth your time as the day turns over: the money chasing AI's power supply is now drilling into rock hot enough to melt lead, and Tesla's in-car assistant keeps deepening its ByteDance dependency.

Mazama Energy has raised $135 million to prove that the hottest rock on Earth can be a power plant for AI data centers. The Series B was led by Centaurus Capital — the firm founded by John Arnold — and Doerr Capital, with ConocoPhillips and Shell Ventures joining as new strategic investors, and repeat checks from Khosla Ventures and Gates Frontier. The company is drilling past 10,000 feet at its Newberry site in Oregon, on the way to roughly 15,000 feet, where it expects to hit 750°F (400°C) rock.

That temperature is the whole thesis. At those depths and pressures, water turns into a supercritical fluid that is neither fully liquid nor fully gas, absorbing far more energy than ordinary steam. Mazama says a single well can therefore produce up to 15 megawatts — up to ten times the output of conventional geothermal wells — and it is aiming for 200 megawatts from the first development by 2030, with the US Department of Energy backing Project Ceres as the near-term proof point. The longer-term ask is an independent certification of at least 10 gigawatts of resource potential at Newberry.

Why it matters for AI: data center builders have mostly bought their way out of the power crunch with natural gas, which is fast to permit and dirty but predictable. Enhanced geothermal has spent a decade as the promising-but-late option. What changed here is who is writing checks — Shell Ventures and ConocoPhillips bring drilling crews and subsurface know-how, not just capital, which is the part geothermal startups have historically lacked. If a single well really delivers 15 megawatts on a 24/7 basis, geothermal stops being a climate story and becomes a load-factor story, and load factor is what every AI campus actually needs.


Tesla pushed software version 2026.26.200.11 to its Chinese fleet on September 17, and the notable change is that the ByteDance Doubao large model now reaches more vehicle models as an in-car voice assistant. The assistant handles real-time information requests and open-ended conversation, and it ships with selectable voices — a livelier female voice and a steadier male one — plus role presets that range from trivia companion to music guide to storyteller. It sits behind Tesla's Premium Connectivity subscription.

The rest of the update is a normal Chinese-market OTA housekeeping pass: a pet mode with named pets, blind-spot warning ambient lighting, immersive audio processing across streaming sources, trip tracking moved into the Energy app, range estimates that now account for battery aging, a "familiar routes" navigation toggle, overnight automatic software installation, rear-screen navigation and lyrics, and camera previews while parked. Tesla also lists fixes for out-of-bounds access, an integer underflow, a CAN bus buffer overflow and a symlink race condition — the kind of security patch note the company rarely publishes in Western markets.

The strategic read is regional divergence. Tesla began putting Doubao inside its cars in August — Tesla puts ByteDance's Doubao AI inside its cars — while its overseas fleet builds around Grok, which the summer 2026.26 release turned into a vehicle-control hub. Two different assistant stacks, two different model suppliers, split by market and by regulator. Whoever wins the cabin will own the daily habit layer for both companies.

What to watch: whether Mazama's first commercial wells hold anywhere near 15 megawatts at sustained output, and whether the Doubao assistant moves down to the base fleet or stays a subscription feature.

If AI's appetite for power is what finally makes frontier geothermal investable, does that make the data center boom good news for the grid? Tell us in the comments.

Read more

Akhetonics says its all-optical CPU reaches a customer in 2026

Akhetonics says its all-optical CPU reaches a customer in 2026

Light-based computing keeps promising more than it delivers — but one Munich startup has just put a date on its bet, and the interview laying it out is doing the rounds on Hacker News this week. Akhetonics says it will deploy its first commercial machine with a major customer by the end of 2026, with several more planned for 2027. The company, founded by Michael Kissner and Leonardo Del Bino, is building a computer where data enters as light, is switched as light, and circulates through memory

The Week in AI — October 5–11, 2026

The Week in AI — October 5–11, 2026

Every big claim this week turned out to rest on fine print more interesting than the headline: revenue only the company reporting it can define, safety tests sandboxed while the product keeps the web, and a Pentagon phase-out nobody would confirm until reporters kept asking. The week's top 5 1. OpenAI's revenue was $20 billion below the numbers everyone quoted — and the gap was definitional. The Financial Times reported Thursday that OpenAI's annualized revenue runs roughly $20 billion unde

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Russia's largest tech company is learning what the AI era's infrastructure war looks like from the receiving end — three data centers in four days, and with them much of the cloud layer Russian businesses run on. A Ukrainian drone strike knocked out Yandex's data center in Vladimir early Sunday morning, the third of the company's facilities hit since October 8. The site — reported at roughly 50 MW and designed for about 2,880 server racks — stopped operating completely after the attack, Yandex

Agent teams cost up to 5x more, barely score higher

Agent teams cost up to 5x more, barely score higher

The multi-agent hype train hit a benchmark this weekend — and the grid and the trucking regulators had quiet weeks of their own. Vals AI put agent teams head-to-head with single agents on its Vibe Code Bench, and the teams cost between 1.8 and 5.1 times more for almost no extra quality. The evals company ran GPT-6 Sol and Claude Opus 5.5 solo and in teams across 50 apps at two reasoning efforts; out of four comparisons, only one was statistically significant — Sol at medium effort, where the t