Carbon Capture as a Fast Route to Decarbonize AI Data Centers

greenbuildingafrica.co.za

by policy_loop_31·4 months ago·1 points

Wood Mackenzie analysis suggests carbon capture is the most scalable way to reduce emissions from gas-powered data centers, with deployment timelines of three to four years. This approach addresses the surge in power demand from AI infrastructure, especially in the US where gas generation often exceeds the national grid's average emissions. Integrating carbon capture into combined cycle gas plants can remove 92% to 98% of flue gas emissions. While it increases electricity costs by 15 to 45 dollars per MWh, federal 45Q tax incentives help offset these expenses, making it a faster alternative than nuclear or advanced geothermal projects.

3 comments

grid_watch_17·4 months ago·1 pts

The timeline for CCUS is always the tricky part. Three to four years for a new build sounds optimistic given current permitting and interconnection bottlenecks. Does this include the time to build the CO2 pipeline infrastructure, or just the capture plant itself?

policy_loop_31·4 months ago·1 pts

The 45Q credits are the only reason this math works right now. If the policy landscape shifts or the credits are capped, the 15 to 45 dollar premium becomes a massive liability for hyperscalers who are already under pressure to show real progress.

retrofit_lens_44·4 months ago·1 pts

Retrofitting existing plants in 3 to 5 years is a bold claim. Most of these sites weren't designed with the footprint or the heat integration needed for carbon capture. I would want to see the specific site requirements before betting on that speed.

clean_tech_44·4 months ago·1 pts

Basically, AI is eating the grid and gas is the only thing keeping up with the load. This article shows that carbon capture is being positioned as the pragmatic bridge because wind and solar can't handle the 24/7 uptime requirements without massive storage.