Environment
Every AI product has an environmental footprint, and it’s fair to ask about ours. Here’s a straight account of where Sorca’s energy goes and how we keep it small — and, just as honestly, what we can’t yet claim.
We haven’t had our footprint independently audited, so we won’t quote you a precise tonnage as if we had. What we can say is grounded in how the product is actually built: a small app on efficient EU cloud, one text inference per note, no training, no owned hardware. As we grow, we’ll pull measured figures from our cloud providers’ carbon tools and publish them here — measured, not modelled.
Sorca runs on Google Cloud in Frankfurt (eur3) and Vercel functions pinned to the EU. Google matches its data-centre electricity with renewable purchases and reports one of the lower-carbon major clouds. We own no servers and no data centre of our own.
Generating a note is a single text-only inference call — no images, no video, no re-runs. Transcription happens in your browser and the audio is discarded, so there is no heavy audio pipeline burning energy on our side.
Model training is the energy-hungry part of AI, and we do none of it. Your sessions, notes and letters are excluded from any training corpus (Anthropic zero-retention). We use models other people already built — we don’t spend energy building our own.
Lightweight models for the high-volume internal work, no duplicate processing, and caching where we safely can. A tool that does one focused thing per note has a naturally smaller footprint than a heavy, always-on product.
A single text inference to draft a clinical note is on the order of a short web search in energy terms — small, and far smaller than generating an image or a video. It’s the training of large models that’s expensive, and we do none of that. This is an illustrative comparison, not a measured figure — we’d rather be honest about the uncertainty than dress an estimate up as fact.
The bigger lever is what we don’t do: no model training, no stored audio, no always-on heavy compute, no data centre of our own.