Have questions about battery systems or want personalised advice for your home?
Contact Us →Have questions about battery systems or want personalised advice for your home?
Contact Us →Full calculations, data sources, and references for every claim we make about our environmental impact.
We believe transparency is more valuable than expensive badges. This page documents our complete methodology for calculating BatteryIQ's environmental footprint, including all assumptions, data sources, and academic references.
If you spot an error or have a question, email hello@batteryiq.com.au.
Estimating AI energy consumption is challenging because providers don't publish per-query data. We've compiled estimates from multiple independent sources to arrive at our figures.
| Source | Estimate | Notes |
|---|---|---|
| Epoch AI | 0.3 Wh/query | Independent research organisation, pessimistic estimate (Feb 2025) |
| Google (Gemini) | 0.24 Wh/query | Official disclosure, median text query (Aug 2025) |
| Sam Altman (OpenAI) | 0.34 Wh/query | Public statement (2024) |
| MIT Technology Review | ~1 Wh/query | Independent reporting on Llama models (May 2025) |
We use 1 Wh per query as a conservative upper-bound estimate.
The "AI uses a bottle of water per query" claim is widely misrepresented. The academic research actually found approximately 500 mL per 20-50 responses, not per query.
| Source | Estimate | Methodology |
|---|---|---|
| Sam Altman (OpenAI) | 0.3 mL/query | Direct data centre cooling only |
| Li et al. (2023) | 10-25 mL/query | Including electricity generation water use |
| Li et al. (2023) | 500 mL per 20-50 queries | Widely cited academic study (often misrepresented as per-query) |
Key insight: Water consumption varies up to 6x depending on data centre location and cooling technology. Google Cloud uses air cooling in many facilities and has committed to being "water positive" by 2030.
We use 50 mL per query as a conservative estimate (double the Li et al. upper bound).
Reference: Li, P., Yang, J., Islam, M.A., & Ren, S. (2023). "Making AI Less 'Thirsty': Uncovering and Addressing the Secret Water Footprint of AI Models." University of California, Riverside.
We calculate avoided emissions based on typical battery dispatch patterns and Australian grid emission factors.
| Metric | Value | Source |
|---|---|---|
| Powerwall 3 usable capacity | 13.5 kWh | Tesla product documentation |
| Warranty period | 10-15 years | Tesla warranty terms |
| Annual coal displacement | ~5,000 kWh | Assumption: typical AU household dispatch |
| 15-year displacement | ~75,000 kWh | Calculated: 5,000 × 15 years |
| Grid carbon intensity (AU avg) | 0.5-0.7 kg CO₂/kWh | AEMO NEM data |
| Lifetime avoided emissions | 30,000-50,000 kg CO₂ | Calculated: 75,000 × 0.5-0.7 |
Coal-fired power plants require significant water for cooling.
Note: Actual avoided emissions vary by household consumption, solar generation, and local grid mix. These figures represent typical Australian conditions.
We calculate the environmental "return on investment" from our AI usage, assuming a conservative 1% conversion rate (1 in 100 calculator users installs a battery).
120,000 : 1
Calculation: 30,000 kg ÷ 0.25 kg = 120,000
(Worst-case at 10 Wh/query: 12,000 : 1)
37,500 : 1
Calculation: 187,500 L ÷ 5 L = 37,500
(Worst-case at 50 mL/query: 3,750 : 1)
We've verified each infrastructure provider's sustainability commitments against their official announcements.
| Provider | Commitment | Source |
|---|---|---|
| Amazon (AWS) | 100% renewable energy matched (2023) | Official announcement → |
| Google Cloud | Carbon neutral since 2007, targeting 24/7 CFE by 2030 | Official announcement → |
| Microsoft (Azure) | Carbon neutral since 2012, carbon negative by 2030 | Official announcement → |
Note: "100% renewable matched" means the provider purchases renewable energy certificates equal to their electricity consumption. "Carbon neutral" includes offsets for remaining emissions. Both represent verified commitments, though methodologies differ.
Our estimated annual operational footprint is approximately 1 tonne CO₂e. This includes a conservative buffer for any unmeasured emissions.
| Source | Status | Annual CO₂e |
|---|---|---|
| Office electricity | Solar + battery | 0 |
| Business travel | EV charged from renewables | 0 |
| Cloud infrastructure | Carbon-neutral providers | 0 |
| Network transmission | Negligible | ~0 |
| Conservative buffer | Unmeasured/embodied | ~1 tonne |
Offset: We've purchased Australian native reforestation credits via Greenfleet matching our estimated footprint.
[1] Amazon / AWS Renewable Energy
Amazon. "Amazon Meets 100% Renewable Energy Goal—Seven Years Early." About Amazon, 2024.
https://www.aboutamazon.com/news/sustainability/amazon-renewable-energy-goal[2] Google Cloud Carbon-Free Energy
Google Cloud. "Announcing Round-the-Clock Clean Energy for Cloud." Google Cloud Blog.
https://cloud.google.com/blog/topics/inside-google-cloud/announcing-round-the-clock-clean-energy-for-cloud[3] Microsoft Carbon Commitments
Microsoft. "Microsoft will be carbon negative by 2030." Official Microsoft Blog, January 2020.
https://blogs.microsoft.com/blog/2020/01/16/microsoft-will-be-carbon-negative-by-2030/[4] AI Water Footprint Research
Li, P., Yang, J., Islam, M.A., & Ren, S. (2023). "Making AI Less 'Thirsty': Uncovering and Addressing the Secret Water Footprint of AI Models." University of California, Riverside.
[5] Tesla Powerwall Specifications
Tesla. "Powerwall 3 Datasheet." Tesla Energy. (Product documentation for capacity and warranty specifications.)
[6] Australian Grid Emissions Data
AEMO. National Electricity Market emissions data and reports.
https://aemo.com.au/Last updated: December 2025
← Back to Our Carbon Commitment
Questions about our methodology? Email hello@batteryiq.com.au