The Path to Scalable, Affordable Carbon Removal

Scaling Direct Air Capture to gigaton scale won’t happen by accident. It requires deliberate engineering, rapid iteration, and a team built for speed.

That’s what we’ve been focused on from day one.

First, we are building a team that is mission-driven. We want people who are here to fight for the mission. When that’s true, it removes ego from the equation and enables better decisions, always guided by what’s best for the mission.

Second, we are building a lean team of exceptional engineers. You can do far more, far faster, with a small team of outstanding people. That’s critical, given the urgency baked into climate change. We think of the team like a professional sports team, where everyone pushes themselves and each other to perform at the highest level. Engineers think from first principles, challenge each other, and take extreme ownership. We trust their judgment over rigid processes that slow them down.

The first decision we had to make was choosing the right technological path.

The key to unlocking Direct Air Capture is cost. Most systems today cost between $500 and $1,000 per ton. To scale globally, we need to bring that down to $100–200. So how do we choose the right technology to get there?

First, we look at the theoretical cost floor: what does physics and chemistry tell us is possible? This helps rule out paths that will always be too expensive.

Second, we ask which technologies get cheaper the fastest. According to Wright’s Law, for every cumulative doubling of units produced, costs fall by a predictable percentage, known as the learning rate. Some technologies have high learning rates because we figure out how to improve them quickly. Others improve more slowly, even if they scale. Picking the right path means betting on the technology that will move fastest down the cost curve.

Ramez Naam highlights two key factors behind fast learning rates: how many moving parts the technology has, and how much of it can be built in a factory rather than on-site. Simpler systems and controlled environments allow for faster iteration, and speed of iteration is paramount.

In fact, the learning rate can matter more than the starting cost. A technology that begins more expensive, or even has a slightly higher cost floor, can still win if it improves faster. We’ve seen this before: crystalline silicon solar panels aren’t the most efficient or theoretically best photovoltaic technology, but they dominated the market because they improved quickly and scaled early.

That’s why we chose to focus on solid sorbents with a temperature–vacuum swing adsorption process. We already know that solid sorbents work, which means we can skip the lab-scale uncertainty and get straight to building, testing, and improving. It may start higher on the cost curve than other approaches, but it offers the steepest learning curve.

The process handles only gases and liquids, which are far easier to move and control than solids. This allows us to design systems that are simple and low-complexity, with very few moving parts.

Solid sorbents also lend themselves well to modular systems, which we can make plug-and-play. That means they can be built on a manufacturing line and deployed with minimal work on site. More on this later.

So that’s the path we picked. But “solid sorbent” is a broad category. So which one should we use?

‍

At the heart of our machines is the solid sorbent, a chemical filter that captures CO₂ while the rest of the air passes through. There are many types, including amine-functionalized materials, zeolites, and metal-organic frameworks. Each comes with tradeoffs in capture capacity, kinetics, stability, reusability, and cost. These properties have a major impact on the overall economics of DAC.

The field is still in its early days, and it’s too soon to pick a single winner. Materials discovery is accelerating, driven by growing financial incentives, rising interest from major OEMs, and AI-driven materials modeling that accelerates discovery and screening. As a result, more teams than ever are racing to develop better sorbents.

That’s why we designed our machines to be sorbent-agnostic. They can run with a wide range of materials, which means we can always use the best sorbent available without redesigning our system. This flexibility lets us ride the wave of material innovation instead of betting on a single chemistry.

And that approach is already paying off. New players are bringing forward sorbents that outperform early standards across the board: higher capacity, better stability, faster kinetics, lower thermal mass, lower cost. We’re now seeing serious entrants from adjacent industries. Major automotive suppliers and some of the largest chemical companies in the world are stepping into the space. These breakthroughs are just getting started, and we’re here for it.

The sorbent gives us the chemistry. Now we need the machines to bring it to life. And we need to think hard about what will let us deliver with speed and scale.

The key to unlocking Direct Air Capture is cost. Most systems today cost between $500 and $1,000 per ton. To scale globally, we need to bring that down to $100–200. So how do we choose the right technology to get there?

First, we look at the theoretical cost floor: what does physics and chemistry tell us is possible? This helps rule out paths that will always be too expensive.

Second, we ask which technologies get cheaper the fastest. According to Wright’s Law, for every cumulative doubling of units produced, costs fall by a predictable percentage, known as the learning rate. Some technologies have high learning rates because we figure out how to improve them quickly. Others improve more slowly, even if they scale. Picking the right path means betting on the technology that will move fastest down the cost curve.

Ramez Naam highlights two key factors behind fast learning rates: how many moving parts the technology has, and how much of it can be built in a factory rather than on-site. Simpler systems and controlled environments allow for faster iteration, and speed of iteration is paramount.

In fact, the learning rate can matter more than the starting cost. A technology that begins more expensive, or even has a slightly higher cost floor, can still win if it improves faster. We’ve seen this before: crystalline silicon solar panels aren’t the most efficient or theoretically best photovoltaic technology, but they dominated the market because they improved quickly and scaled early.

That’s why we chose to focus on solid sorbents with a temperature–vacuum swing adsorption process. We already know that solid sorbents work, which means we can skip the lab-scale uncertainty and get straight to building, testing, and improving. It may start higher on the cost curve than other approaches, but it offers the steepest learning curve.

The process handles only gases and liquids, which are far easier to move and control than solids. This allows us to design systems that are simple and low-complexity, with very few moving parts.

Solid sorbents also lend themselves well to modular systems, which we can make plug-and-play. That means they can be built on a manufacturing line and deployed with minimal work on site. More on this later.

So that’s the path we picked. But “solid sorbent” is a broad category. So which one should we use?

One of the unique advantages of Direct Air Capture is that it can be deployed almost anywhere. CO₂ is evenly distributed in the atmosphere, so there’s no need to “go where the emissions are.” That opens up the possibility to site DAC systems where the economics work best.

We look for sites with abundant, cheap clean energy. Powering DAC only makes sense if the energy is both clean and cheap, that’s what will make the unit economics attractive and ensures we remove far more CO₂ than we emit. That energy must either already be in surplus, or we must be helping to build more.

We also need land. Removing billions of tons of CO₂ will require deploying a lot of machines, which means a lot of space. We focus on areas where we can scale without displacing people or ecosystems.

Finally, we look for places where we can store the CO₂ directly on-site. Transporting CO₂ is expensive and adds complexity. By colocating with our storage partners, we eliminate that cost entirely and simplify operations.

The good news is that there are a lot of places that meet these criteria. Kenya is one of the best examples. The grid is already over 90 percent renewable, powered by geothermal, hydro, and wind. Its geothermal plants regularly curtail excess heat and electricity, which we can put to work. There’s ample non-arable land in the Rift Valley, and the Rift itself is rich in basalt formations, ideal for permanent CO₂ storage through mineralization. Most importantly, local partners are eager to collaborate, and the energy infrastructure is ready to grow.

Alright, let’s bring it all together. We said that cost was the biggest bottleneck. So where does our approach actually land?

We already see a clear path to getting below $200 per ton by 2030. And the reason we’re confident is because the cost-down has already started, and it’s happening fast.

Our modular design accelerates everything. Because each container is an independent unit, we can iterate fast, improving energy efficiency, improving our designs, and driving down costs with every build. In just the last 12 months, we’ve reduced the capex per container by 4x.

Then there’s scale. Several of our suppliers have already given us pricing tiers that drop significantly as volumes grow. The price of key components drop more than 50% as volumes increase, just from known cost curves. And as clean energy gets cheaper globally, the cost of running our machines keeps falling too.

Finally, because our system is sorbent-agnostic, we’re ready to absorb improvements as they come. Over the past year, our sorbent cost has dropped by 3x thanks to new suppliers and better materials entering the market, and we already see their next-generation products bringing another 50% cost reduction. Because we can integrate those materials without redesigning the system, we capture those benefits immediately. And beyond that, there may be breakthroughs we haven’t even accounted for. We’ll be uniquely positioned to take advantage of them when they arrive.

We’re confident in the tech and the cost curve, but some will argue that we should wait until the cost is lower before we scale. But that’s not how progress works. The fastest way to reduce cost is to deploy, learn, and iterate. And in the meantime, early deployments generate revenue, build credibility, and prove the model. Did Tesla wait until he had the cheapest EV before shipping something? Of course not. They started, improved, and scaled, and that’s exactly what we’re doing.

Let’s wrap up.

In 1968, Stewart Brand wrote to the hippies who wanted to build a new world: “We are as gods, and might as well get good at it.” By 2009, he was reflecting on a new reality: nature is no longer in balance, and ecosystems won’t simply repair themselves if we leave them alone. Instead, we need to take responsibility. “We are as gods and have to get good at it.”

That’s what we are about, and this is the master plan to get there:

1. Build our first modular DAC units in the form of a shipping container — Done.
2. Deploy those containers to build capture facilities — first site coming Q4 2025.
3. Iterate on the technology as fast as possible to drive down cost.
4. Scale our capture facilities in places with the best economics.