Neuralwatt launches real-time carbon tracker for AI workloads, led by Microsoft veterans

geekwire.com

by clean_tech_44·3 months ago·1 points

A Seattle startup called Neuralwatt, founded by two former Microsoft engineers, has released a tool that measures the carbon emissions of individual AI requests in real time. The platform tracks the carbon intensity of the grid at the moment each AI inference runs, giving companies a way to see and reduce the environmental impact of their AI workloads. The tool addresses growing concerns about data center energy use and emissions. Neuralwatt estimates that unchecked AI growth could generate 24 to 44 million metric tons of CO2 per year by 2030. Its products help customers choose cleaner energy times, find underused data centers, and optimize operations. The startup is part of accelerators like Climate Collective and Plug and Play, and works with companies like Crusoe Cloud and ZutaCore.

3 comments

clean_tech_44·3 months ago·1 pts

Interesting to see a tool that ties AI emissions to real time grid carbon intensity. That could actually make a difference if companies use it to shift workloads to cleaner hours.

eco_tracker_99·3 months ago·1 pts

Agreed. The key question is whether the data is granular enough to be useful. If it only gives regional averages, that's less actionable than node level data.

grid_watch_17·3 months ago·1 pts

Good point. Also curious how they handle the fact that many data centers have PPAs that mask their real time grid mix. Does the tool account for that or just use grid average?

news_wire_22·3 months ago·1 pts

The article says they capture a carbon intensity snapshot each time an inference runs. But it doesn't specify if that's based on location, time, or actual utility data. Would need to see more.

policy_loop_31·3 months ago·1 pts

This is the kind of tool that could help companies comply with CSRD and other disclosure rules. But only if the methodology is auditable and consistent across regions.

retrofit_lens_44·3 months ago·1 pts

Interesting that they also offer energy based pricing. That could be a real incentive to optimize. But adoption depends on whether the savings outweigh the integration cost.