Los resúmenes, navegación y metadatos de políticas están localizados. El cuerpo legal completo usa como respaldo el texto canónico en inglés salvo cuando ya mantenemos una traducción específica de jurisdicción.
01Summary (Q1 2026)
| Metric | Q1 2026 | vs. Q1 2025 |
|---|---|---|
| Total emissions (tCO₂e) | 184 | −12% |
| Emissions per employee | 0.71 tCO₂e | −18% |
| Compute energy (MWh) | 312 | −6% |
| Renewable share | 74% | +9 pp |
| Water used (kL) | 1,840 | −4% |
02Methodology
We follow the GHG Protocol Corporate Standard and ISO 14064-1:2018. Operational control boundary. Cloud emissions calculated via AWS Customer Carbon Footprint Tool and Azure Emissions Impact Dashboard, cross-checked with the SCI specification (greensoftware.foundation). Numbers are independently verified annually by DNV.
03Scopes 1, 2, 3
- Scope 1 (direct) — small. Our offices use district heating; we operate no fleet.
- Scope 2 (purchased energy) — offices in 8 cities. 74% of office grid electricity is matched with renewables; HQ Gurugram is on a green-power tariff.
- Scope 3 (value chain) — the largest share. Dominated by cloud compute (74% of Scope 3) and business travel (18%).
04Compute & data
We publish a per-product energy intensity (kWh per million inferences):
| Product | kWh / 1M inferences | Trend |
|---|---|---|
| VIEW (vision detection) | 0.84 | ↓ |
| SAFER (crowd density) | 1.12 | ↓ |
| HyperFUSE (biometric match) | 0.31 | ↓ |
| LENS LM (language model) | 4.62 | flat |
| NMS (telemetry classification) | 0.06 | ↓ |
05How we reduce
- Smaller models. We benchmark distilled / quantised variants for every release. Default to the smallest that hits accuracy targets.
- Right-sized inference. Edge deployment where the workload allows; cloud only where centralisation actually saves energy.
- Region-aware scheduling. Batch jobs are routed to regions with the cleanest grid mix that hour.
- Hardware reuse. Min. 4-year refresh cycle on laptops; refurbished where possible.
06Offsets
We offset what we cannot reduce, but only with verified high-quality removals — currently biochar and enhanced rock weathering, sourced via Frontier and Puro.earth. We don’t buy avoidance credits. Retired-credits ledger: lenscorp.ai/sustainability/credits.
07Annual report
Our full disclosure is published in line with TCFD and ISSB IFRS S2. Latest report: LENS_Carbon_Disclosure.pdf.
¿Necesita un aviso regional?
Publicamos avisos complementarios para India, UE/Reino Unido, salud en EE. UU., California y Brasil.