Direct Air Capture sounds like a straightforward engineering problem: build a machine, pull CO₂ out of air. In practice, it is a fight against the second law of thermodynamics — and the second law does not negotiate.
Why DAC is hard by definition
Atmospheric CO₂ sits at roughly 420 ppm today. That means roughly 1 out of every 2,400 molecules in the air is a CO₂ molecule. To capture it, you have to separate a highly dilute component from a vast mixture — and thermodynamics tells you exactly how much that must cost at minimum.
The minimum work required to separate CO₂ from a mixture comes from the Gibbs free energy of mixing. For an ideal gas mixture, separating one mole of component from a mixture where its mole fraction is requires:
At 400 ppm () and 298 K:
Converting to practical units: 19,500 J/mol ÷ 44 g/mol × 1000 kg/tonne = approximately 440 MJ/tonne, or ~120 kWh per tonne of CO₂.
This is not a technology specification — it is a physical limit. No matter how clever the chemistry, no matter how much money is invested, no DAC process can ever fall below 120 kWh/tonne. The second law is the floor.
What "only 120 kWh/tonne" actually means at scale
The climate community talks about needing to capture billions of tonnes of CO₂ per year by mid-century. Let's be specific about what that requires.
At 1 billion tonnes per year — a modest target in the context of 37 billion tonnes emitted annually — the minimum energy requirement is:
That is roughly 3% of total US electricity generation in 2022. At the theoretical minimum.
For the more ambitious scenarios — cumulative removal of 100–200 billion tonnes between now and 2050 needed to stay near 1.5°C — the cumulative energy requirement runs to several petawatt-hours. Under optimistic emissions trajectories, this approaches 5% of total projected global annual energy consumption, sustained for decades.
And this assumes 100% thermodynamic efficiency, which is impossible.
The efficiency gap
The minimum work assumes a perfectly reversible process. Real processes are irreversible: there is heat loss, friction in compressors and fans, incomplete reactions, regeneration losses in sorbents. Current leading DAC systems (liquid solvent and solid sorbent approaches) achieve roughly 7–8% of the theoretical minimum efficiency.
That means actual energy consumption is approximately:
Real-world deployments confirm this range. Climeworks' Orca plant in Iceland reported energy consumption around 2,000 kWh per tonne of CO₂. Carbon Engineering has published figures in the 1,400–2,000 kWh/tonne range depending on heat integration.
At 2,000 kWh/tonne, that 1 billion tonne/year scenario requires not 120 TWh/year but 2,000 TWh/year — about 50% of current US electricity generation, dedicated entirely to one billion tonnes of carbon removal.
The asymmetry argument
This is where the thermodynamics becomes a policy argument.
Emitting a tonne of CO₂ releases the energy content of the fuel and costs nothing additional. Capturing a tonne back from the atmosphere at 400 ppm requires 1,500–2,000 kWh of new energy input. The ratio is brutally asymmetric.
Every tonne emitted today that must be removed later requires that 1,500–2,000 kWh burden. Every tonne not emitted avoids that burden entirely. The physics strongly favors emission prevention over atmospheric removal — not as a moral argument, but as a resource accounting statement.
Stopping emissions is not easier than DAC because of political will or economic incentives. It is easier because the thermodynamics is fundamentally different. You are working with the second law rather than against it.
What this doesn't mean
None of this says DAC is useless. There are emissions pathways — aviation, cement, agriculture — where the residual carbon cannot easily be avoided and where removal will genuinely be necessary to achieve net-zero. DAC is likely an essential component of that picture.
But the framing matters. DAC is expensive by the laws of physics, not just by current technology immaturity. Efficiency improvements will help — getting from 7% to 30% efficiency would cut energy demand by 4× — but the floor remains 120 kWh/tonne, and the scale of deployment required for meaningful climate impact is enormous regardless.
The honest summary: DAC is a last resort that will probably be necessary, and the less we need of it, the better — not just economically, but thermodynamically.