AI Helps Scale Direct Air Capture With Geothermal Power

A direct air capture (DAC) machine that removes 60 tons of CO₂ per year doesn’t sound like much. But that number is the point: it’s small enough to be real, testable, and brutally honest about what still needs to be proven.

In Kenya’s Great Rift Valley—where geothermal steam already generates about a quarter of Kenya’s electricity—startups are trying to turn excess heat and uniquely reactive basalt rock into permanent carbon storage. The bet is bold: build a “Great Carbon Valley” that makes DAC cheaper, cleaner, and easier to scale than the versions struggling elsewhere.

For readers following our AI in Energy & Utilities series, this story lands close to home. The biggest bottlenecks in geothermal and carbon removal aren’t just mechanical—they’re operational: predicting output, scheduling maintenance, verifying CO₂ storage, managing power demand, and proving results to skeptical buyers. That’s where AI-driven analytics and modern digital services matter, whether you’re in Kenya, California, or the Gulf Coast.

Why geothermal + direct air capture is a serious bet

Geothermal-backed DAC is compelling for one reason: energy quality. DAC needs lots of energy, and not all energy is equally useful.

Kenya’s Rift Valley offers two advantages that many regions can’t pair in one place:

1. Abundant renewable energy from geothermal resources—plus “waste” thermal energy that isn’t fully converted into electricity.
2. Basalt-rich geology that can mineralize CO₂ into stable solids (think calcite and magnesite), which is the kind of durable storage carbon credit buyers increasingly demand.

In practice, that means you can power capture with low-carbon heat and then store the CO₂ in a way that’s easier to argue is permanent.

The hard truth: DAC is still expensive and politically exposed

Even supporters admit DAC is not a near-term substitute for emissions cuts. Costs remain high, and the market is fragile.

• Typical compliance carbon prices have been far below DAC credit prices. One benchmark cited widely in the market: DAC credits averaging around $450/ton while major carbon markets trade far lower.
• Total global DAC removals remain tiny—on the order of tens of thousands of tons per year, not billions.
• In 2025, climate policy whiplash—especially in the United States—has pushed uncertainty into the entire carbon removal purchase pipeline.

That’s why any “DAC hub” story has to be evaluated like an energy project and like a digital trust project. If the measurement, reporting, and verification (MRV) isn’t airtight, buyers disappear.

What’s actually happening in Kenya’s “Great Carbon Valley”

The Great Carbon Valley vision is part brand, part real industrial strategy: attract multiple carbon removal and clean industrial players to a geothermal region so shared infrastructure (power, permitting, transport, storage) lowers costs over time.

Several companies are lining up pilot efforts. One high-profile thread is Octavia Carbon, a Nairobi-rooted startup testing modular DAC units near Lake Naivasha.

Here’s the technical core, simplified:

• Fans move air across a chemical filter (amine-based) that binds CO₂.
• The system heats the filter (roughly 80–100°C) under vacuum to release CO₂ as a gas.
• CO₂ is then pressurized and injected underground.
• Basalt chemistry helps mineralize CO₂ into rock-like solids.

A notable detail: Octavia emphasizes that its approach uses mostly thermal energy (over 80% thermal, per the reporting), aligning well with geothermal heat that might otherwise be vented.

This is the kind of “boring operational fit” that makes or breaks climate tech: matching the energy form (heat vs. electricity), climate conditions (humidity and temperature), and logistics (modular containers) to the place.

Where AI fits: turning pilots into reliable infrastructure

AI won’t make DAC cheap by magic. What it can do is reduce the waste, downtime, and uncertainty that kill early projects. In energy and utilities terms, AI helps move a system from “prototype” to “plant.”

AI for geothermal optimization and heat utilization

The first scaling problem is energy management: geothermal production varies, equipment degrades, and grid demand shifts.

AI helps operators:

• Forecast steam and brine conditions using time-series models tied to sensor data (pressure, temperature, flow).
• Optimize dispatch between electricity generation and process heat users (like DAC) when heat is abundant but grid demand is low.
• Detect anomalies early—valve drift, scaling, pump inefficiency—before they become shutdowns.

If you run a utility, this is familiar: it’s the same playbook as AI for wind forecasting or transformer health, applied to geothermal wells and heat exchangers.

AI for DAC process control (the hidden cost center)

DAC performance lives and dies on tight operational control: fan speed, sorbent loading, temperature ramp rates, vacuum cycles, and humidity management.

AI-driven control systems can:

• Tune cycles to local weather conditions (especially humidity), maximizing capture per kilowatt-hour.
• Predict sorbent degradation and schedule replacement based on performance curves.
• Balance throughput vs. wear so units don’t “optimize themselves to death.”

A practical stance: the winners in DAC won’t be the teams with the fanciest chemistry—they’ll be the teams who run plants like high-uptime industrial systems. AI is how you do that at scale with lean teams.

Digital MRV: the difference between a climate project and a financeable asset

Carbon markets are under heavy scrutiny. Nature-based credits have faced repeated allegations of over-crediting, and buyers now demand durability and proof.

For mineralized storage, digital MRV can combine:

• Continuous monitoring of injection volumes and pressures
• Geologic models updated with real field data
• Audit-ready data pipelines that prove chain of custody from capture to storage

In U.S. terms, this resembles what industrial clients ask for in regulated environments: immutable logs, clear data lineage, and third-party verification readiness.

If you’re selling carbon removal, MRV isn’t paperwork—it’s your product.

The real scaling constraint: demand, not prototypes

A lot of climate tech commentary over-focuses on engineering milestones and under-focuses on buyers.

Kenya’s strategy tries to solve a core market problem: build a hub where renewable energy is abundant and “extra,” so DAC doesn’t look like it’s stealing clean power from homes or factories.

But even if the energy case works, DAC still needs durable demand. Right now, demand is fragmented:

• Voluntary buyers are inconsistent and prone to “net-zero backtracking.”
• Governments are uneven—some jurisdictions are building compliance markets, others are retreating.
• High-quality removal purchases often come through structured programs (like advance market commitments), not routine procurement.

What would make demand more stable (including in the U.S.)

From an energy & utilities lens, these are the market signals that turn pilots into buildouts:

1. Standardized MRV requirements that reduce buyer uncertainty.
2. Long-term offtake contracts (5–15 years), similar to power purchase agreements.
3. Policy-backed procurement where governments require removals for specific sectors (hard-to-abate industries).
4. A credible pathway to “no double counting” so credits aren’t a reputational risk.

U.S. tech and digital services firms can play a direct role here—not as donors, but as customers who demand strong verification and stable delivery.

The community factor: infrastructure can’t be “extractive” and still succeed

Most companies get this wrong: they treat community engagement like a checkbox, and then act surprised when projects stall.

In the Rift Valley, distrust isn’t abstract. Indigenous Maasai communities have faced displacement tied to energy development, and many residents still lack access to the electricity generated nearby.

Here’s the stance I think serious builders should take:

If a project sells “climate benefit” globally while locals can’t get reliable power, it’s not a climate solution—it’s a credibility problem waiting to explode.

AI and digital tools can help here too, but only if they’re used for accountability:

• Public-facing dashboards showing local electrification progress and service reliability
• Transparent grievance tracking and response SLAs (yes, like customer support)
• Workforce development tracking that shows job quality, not just job counts

Trust is a project dependency, just like permitting or financing.

Practical lessons for U.S. energy and digital leaders

Kenya’s Great Carbon Valley effort is a mirror for U.S. decarbonization projects: ambitious tech meets messy reality. If you’re building AI in energy & utilities—or buying carbon removal—these are the practical moves that matter.

1. Treat MRV like a first-class product requirement. If you can’t audit it, you can’t sell it.
2. Use AI to raise uptime, not just optimize a single metric. Reliability beats lab efficiency.
3. Design for local energy constraints. Thermal integration, water use, and grid impacts decide viability.
4. Buy in ways that create predictable revenue. Long-term contracts beat one-off pilot purchases.
5. Measure local benefit as seriously as CO₂ removed. Power access, wages, safety, and land rights aren’t “nice to have.”

What to watch in 2026: the signal that DAC hubs are real

The next year will separate “press release hubs” from durable infrastructure. The clearest signals won’t be slogans—they’ll be operational proof:

• Verified capture performance over seasons (wet vs. dry periods)
• Documented mineralization results and monitoring integrity
• Falling operating costs through learning curves and automation
• Evidence that new geothermal buildout improves grid access locally

If that happens, Kenya’s geothermal DAC story won’t just be about carbon removal. It’ll be a case study in how AI-driven operations, industrial digital platforms, and trustworthy data turn climate ambition into something financeable and repeatable.

And for U.S. companies building AI-powered technology and digital services, the question isn’t whether these projects matter. It’s whether you’re ready to support (and demand) the level of operational rigor that makes them work.