#6 of 30 · Enterprise AI + cloud
IBM (Granite / watsonx)
Small open models, 80% renewable electricity, net zero by 2030.
In plain words
How scores workIBM scores 63 out of 100 — middle of the pack. Some disclosure or clean-energy commitments, offset by rising emissions or blind spots. Confidence in this score is good confidence (B).
What helps
- Efficiency · Granite models are 1B–34B parameters, a fraction of frontier sizes; Granite 4.0's hybrid Mamba architecture cuts memory 70%.
- Carbon · 79.6% renewable electricity in 2024 (1.48 TWh); operational emissions 265 kt, down 77% against 2010.
- Transparency · Annual Impact Report with energy, renewables and PUE; open weights allow independent measurement.
What hurts
- Efficiency · No per-prompt measurement published.
- Carbon · Weighted-average data-centre PUE of 1.41, versus 1.1–1.15 at hyperscalers.
- Water · No WUE or water-consumption figures for data centres.
What nobody knows
- Per-prompt energy disclosed
- Per-prompt water disclosed
- Training emissions in model cards
- Water-use effectiveness reported
IBM's Granite models are small (1B–34B), open-weight and trained on its Vela and Blue Vela clusters. IBM procured 79.6% renewable electricity in 2024, cut operational emissions 77% from 2010 and targets net zero by 2030 with under 350 kt of residual emissions — but its co-location data centres average a PUE of 1.41, well behind hyperscalers, and it publishes nothing per model or per prompt.
Key figures
Per-prompt and fleet metrics. Hover a title for what it means; open “How we got this number” for the working.
ecogpt.co estimate; no company figure.
How we got this number
Granite models are 1B–34B parameters — the sub-10B class whose measured median is 0.042 Wh per query; Granite 4.0's hybrid design cuts memory a further 70%. Uncertainty roughly ±3×.
On-site cooling only, so comparable with Google and OpenAI's figures.
How we got this number
0.05 Wh per prompt × 1 L/kWh (estimated WUE for co-location sites with PUE 1.41). Counting the water evaporated to generate the electricity adds ≈0.09 mL.
Location-based: what the grid actually emitted. Renewable certificates would bring the market-based figure near zero.
How we got this number
0.05 Wh × 390 gCO₂/kWh (US grid, location-based).
How we got this number
No disclosure. IBM leases co-location capacity with a fleet PUE of 1.41, which usually means conventional evaporative cooling; 1.0 L/kWh is a mid-range assumption.
Target 90% by 2030.
Hardware & materials
The chips, power, copper and future e-waste behind the fleet — the part sustainability reports leave out
GPUs, TPUs or custom accelerators the company runs. Analyst figures are in H100-equivalents, company figures in physical chips.
How we got this number
Vela is an A100 cluster on IBM Cloud; Blue Vela is an on-premises H100 system IBM describes as 'thousands of GPUs'. Uncertainty ±2×.
Megawatts the AI fleet pulls. One megawatt runs roughly 750 US homes.
How we got this number
1.86 TWh a year ÷ 8,760 hours. Average draw implied by annual electricity; peak capacity is higher.
Terawatt-hours per year. One TWh is a year of electricity for about 95,000 US homes.
How we got this number
IBM reports 1,480,000 MWh of renewable electricity as 79.6% of consumption, which implies 1.86 TWh in total.
CO₂e emitted making the chips and servers before they answer a single prompt.
How we got this number
20,000 accelerators × 514 kg CO₂e each (chip plus share of the host server). Emitted in manufacturing, before the first prompt is served.
Tonnes of server hardware that will be retired in three to five years.
How we got this number
20,000 accelerators × 17 kg of server hardware each. Tomorrow's e-waste unless reused; generative AI could produce 2.5 Mt of it a year by 2030.
Tonnes of copper in cabling, busbars and transformers.
How we got this number
212 MW × 27 t of copper per MW (cabling, busbars, transformers).
Diesel backup units or gas turbines at the data centres, permitted or observed.
How we got this number
212 MW ÷ 2.6 MW per generator, the average size permitted at US data centres. Permit counts are not published fleet-wide; this is the number a fleet this size normally carries.
Share of retired servers and parts given a second life.
Kilograms of antimony-equivalent, the life-cycle unit for scarce minerals consumed.
Score breakdown
Each pillar is 0–100; weights are fixed across the index
- Granite models are 1B–34B parameters, a fraction of frontier sizes; Granite 4.0's hybrid Mamba architecture cuts memory 70%. IBMHugging Face, Salesforce, Cohere, Carnegie Mellon
- No per-prompt measurement published. IBM Community
- No WUE or water-consumption figures for data centres. IBM
- Annual Impact Report with energy, renewables and PUE; open weights allow independent measurement. IBM
- No model-level emissions or water. IBM Community
- Net zero by 2030 with residual emissions below 350 kt; 90% renewable electricity by 2030; 2025 conservation target already exceeded. Data Center DynamicsIBM
- No enforcement actions; prioritises co-location sites where renewable electricity is available. IBM
Models
What this company ships, and what each prompt costs
| Model | Weights | Energy / prompt | Water | Training CO₂e |
|---|---|---|---|---|
| Granite 4.0 (Micro / Small) 3B–32B hybrid Mamba/transformer | Open | 0.05 Wh Estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement. | 0.050 mL Estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement. | 300 t Estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement. |
Where the compute lives
Named sites from company disclosures and press reporting
- Armonk HQNew York, USA
- Poughkeepsie (Blue Vela)New York, USA · On-premises H100 cluster for Granite.
- IBM Cloud DallasTexas, USA
- IBM Cloud FrankfurtGermany
Timeline
Incidents and developments involving this company
