Engineers at the University of Houston have recorded approximately 90% carbon dioxide capture efficiency in laboratory tests of a system that fits inside standard 40-foot shipping containers. The team, led by Assistant Professor Mim Rahimi with Harvard postdoctoral associate Ahmad Hassan, published the results in Environmental Science & Technology.

The technology, termed Small Modular Electrochemical Containers (SMECs), targets a specific and fast-growing emissions source: natural gas power plants feeding AI data center buildouts. Modeling in the paper indicates that 20 modular SMEC units could capture up to 98% of CO2 emissions from a 360-megawatt natural gas facility.

What does the system replace?

Conventional point-source capture demands capital-intensive infrastructure. Plants install absorption towers exceeding 100 feet in height, supply heavy thermal energy to regenerate the solvent, and maintain specialized membranes that foul and add cost. SMECs strip all three requirements out of the process.

The core component is a membraneless electrochemically mediated amine regeneration (EMAR) cell. The design combines CO2 capture and release in a single electrochemical unit, removing the mechanical and structural bottlenecks that define tower-based amine systems. The separation runs on electricity rather than heat.

Containerization changes the deployment math as well. Because operators can activate or deactivate individual containers, capacity scales in discrete increments. For power plants balancing natural gas generation against intermittent renewables, that modularity matches capture throughput to fluctuating load.

How was it validated?

The researchers tested the architecture in the laboratory under simulated natural gas and coal exhaust conditions. Two results anchor the case for scale-up:

• Approximately 90% CO2 capture efficiency across the test campaign.
• Stable performance maintained over 300 operational cycles.

The commercial deployment model is grounded in a named project. The team modeled the system against a 360-megawatt natural gas facility linked to OpenAI's planned Stargate AI data center project in Abilene, Texas. The analysis found that an array of 20 shipping-container units could capture 98% of that plant's carbon dioxide while occupying a minimal footprint on the site.

What does this change for capture economics?

The honest answer at this stage: not much yet, and potentially a great deal later. The figures come from laboratory testing and modeling, not from a commissioned unit on an operating plant. The 300-cycle stability result and the 90% capture efficiency establish technical feasibility; the 98% figure for Abilene is a projection, not measured performance.

What the architecture does offer is a credible path around the cost structure that has stalled tower-based capture. A 40-foot container is a manufactured product, not a construction project. If EMAR cells hold up beyond the laboratory cycle count, capture capacity could be procured, sited, and expanded in increments that match data center power contracts.

The team is not stopping at flue gas. According to the publication, the UH researchers are adapting the electrochemical cell design for direct air capture (DAC), which would extend the platform to decentralized industrial sites without a concentrated point source.

What happens next?

The milestone to watch is the transition from simulated exhaust to a field demonstration on an operating gas plant. Sustained cycle life beyond 300 operations, verified capture rates on real flue gas, and a published cost per tonne would move SMECs from an academic result to a procurement option for the developers now contracting gigawatts of gas-fired capacity behind AI campuses.