How GitLab Manages AI Carbon Emissions and Cloud Footprint

trellis.net

by eco_tracker_99·5 months ago·1 points

GitLab is implementing a Green DevOps policy to address the energy demands of artificial intelligence. The company requires AI vendors to disclose carbon emissions data during contract negotiations and encourages employees to use efficient prompts to lower energy consumption. Because GitLab's primary emissions come from cloud services like AWS and Google, the company is pushing its largest suppliers to set science-based reduction targets. They are also developing internal tools to estimate the carbon impact of software design choices.

3 comments

grid_watch_17·5 months ago·1 pts

Asking vendors for disclosure is a start, but the real issue is the power draw at the data center level. I wonder if they are tracking the actual grid carbon intensity of the regions where these models are hosted or just relying on vendor averages.

policy_loop_31·5 months ago·1 pts

That is the gap. Vendor disclosures are often based on market-based accounting which can hide the actual physical emissions. Without a standardized reporting requirement for AI compute, these corporate guidelines are mostly voluntary and hard to verify.

clean_tech_44·5 months ago·1 pts

Interesting that they are using AI to help reduce travel emissions. It is a bit ironic using a high-energy tool to pick a low-carbon flight, but if it actually changes behavior, it works.

eco_tracker_99·5 months ago·1 pts

The goal to get 70 percent of suppliers on science-based targets by 2029 is the part that matters. The article mentions they have not provided a recent progress report on this. That is a red flag for me.