Universal Open Source AI Sustainability Operating System and SDK

No standard way to measure what AI fully consumes, and negatively impacts, with closed-loop reporting

AI's electricity demand is growing exponentially, creating problems and challenges we see every day. Yet no standard, shared way exists to measure what real and evolving AI workloads consume. Today's benchmarks measure peak hardware speed, not the energy real requests use on a real grid at a real time. Every provider reports differently, so buyers, regulators and researchers cannot compare claims or verify efficiency gains. Even organizations that know their footprint have no simple, standard path from measuring it to reducing it and addressing AI's disruptive concerns.

Every AI query has a price tag most people never see: tokens spent, dollars charged, and the electricity, water, and emissions emitted — behind the scenes. That footprint is invisible to most people generating it by using AI services, which means it's invisible to their decisions that could reduce it.

We've built and are improving codename Stardust, a free, open-source operating system and SDK that puts those numbers back in front of you. Today it runs as a browser extension alongside the AI chat apps people already use, including Claude, ChatGPT, and Gemini, and as software kits that developers add to their own AI applications. For teams running AI at scale, it plugs into the monitoring tools they already use through OpenTelemetry. Wherever a decision gets made, the same figures appear: tokens, cost, energy, water, and carbon, per prompt or per workload.

One glance gives you a compact score; one click shows the full breakdown in plain language. Every figure is labeled as measured or estimated, and our methods are published openly, so anyone can check our work and help improve it.

This is AIforE's ROI in practice: return on investment and return on impact. Stardust helps people and companies capture AI's real upside (productivity, speed, quality of life) while giving them what they need to right-size how they use it.

Connecting clients, providers, and the exchanges that make it work

At the center is Stardust Core. It measures a client's footprint across carbon, electricity, water, heat, and materials, matches it with the providers that can address each one, and connects to the exchanges and registries where credits for each are issued, traded, and retired. The value flowing through those markets, including recycling, helps offset the cost of sustainable operations.

Stardust never mints its own credits. Recognized programs verify and issue them; Stardust supplies the measurement data and retires credits in the client's name.

Ground rules for every credit

Like for like. A water credit never cancels carbon, and a renewable energy certificate is an electricity claim, not a carbon offset.

One benefit, one claim. Each credit is retired once, for one beneficiary and one claim period.

Match in time and place. Electricity is matched within a market boundary and, where possible, by the hour; water within the same or a stressed watershed; heat to a named local recipient.

Show the confidence. Every provider carries a confidence tier, so a verified registry credit is never presented the same way as an unverified attestation.

This is built in the open as Project Stardust, and the measurement standard behind it is AIforE's contribution to the field: one open schema for the negative and positive impact of computing, mapped to existing standards rather than competing with them. Many registries don't yet offer open write APIs, so early connectors will retire credits through an account holder and move to automation as registries support it.