Researchers at the University of Houston have demonstrated a carbon capture process operating at 90% efficiency, a figure they frame as a response to the surging electricity demand driven by AI data centers.

The number matters because capture efficiency is the gatekeeper metric for whether post-combustion retrofits make economic sense on power plants serving compute-heavy load. At 90%, the Houston team says its approach clears the threshold that has long kept capture from scaling beyond demonstration projects.

What did the University of Houston actually demonstrate?

The university reported a capture efficiency of 90% in its demonstration, positioning the result specifically against the backdrop of AI's energy appetite. Data center growth is pushing utilities toward extending the operating life of gas-fired generation, which in turn puts capture technology back on the table as a compliance pathway.

The claim comes from the university's own research team. That distinction matters for practitioners: this is a demonstrated result at research scale, not a permitted commercial facility with a nameplate tonnage. The gap between a 90% lab figure and a plant pulling millions of tonnes per year out of a flue stack remains the industry's core problem.

Why is AI demand the hook?

The framing is deliberate. AI workloads concentrate electricity consumption in specific grids, and the generation built or extended to serve them carries an emissions profile that corporate net-zero pledges have to account for. Capture technology that works at high efficiency gives operators of thermal plants a route to keep units running under tightening scrutiny — provided the cost per tonne holds up.

For circular-economy watchers, the material stream here is CO2 itself: captured carbon only closes the loop if it moves into utilization or storage at scale. A 90% capture rate upstream means little without offtake downstream.

What happens next?

The demonstration gives the University of Houston a data point to carry into scale-up discussions. The milestone that decides whether this becomes infrastructure rather than a headline is the first pilot on an operating power plant — with a published tonnage, a cost per tonne, and a regulatory docket behind it.