The Greentech Meetup 2025 highlighted how artificial intelligence is becoming a strategic lever for climate risk anticipation, territorial resilience, and economic adaptation. The discussions converged on one core reality: climate risk is no longer marginal, and AI is shifting from exploratory technology to critical infrastructure.
1. Climate risk is now a systemic economic risk
A central message was the scale of economic exposure to weather and climate:
- 25–30% of national GDP is directly affected by weather conditions.
- Climate hazards now influence industrial location strategy, agricultural yields, infrastructure planning, and insurance models.
- Artificial intelligence is increasingly seen as the only tool capable of handling this complexity at operational speed.
This reframes climate risk from an environmental issue into a macroeconomic and industrial risk.
2. Météo-France: AI as critical forecasting infrastructure
Nicolas Trift, Director of Strategy at Météo-France, presented a clear evolution of weather prediction:
Key facts:
- AI has been used in meteorology for decades.
- 35 million weather observations are ingested daily.
- Massive open datasets are available through data.gouv.fr.
- AI-based models generate forecasts up to 1,000 times faster than traditional physics-based models.
Historical evolution:
- 1990s: statistical post-processing and correction models.
- 2023: acceleration with large-scale deep learning weather models.
- 2024: launch of new European-scale climate projection models.
Operational AI use cases already deployed:
- Fog prediction in the Seine Valley:
Detection performance improved from 68% to 90% using AI. - Extreme rainfall and cyclone prediction in radar-scarce territories (e.g., Mayotte).
Météo-France has created:
- An internal AI Lab with multidisciplinary teams (data scientists + meteorologists).
- MeteoPhare, its internal incubator.
- A clear principle: AI supports decision-making, but human experts remain the final authority.
An open hackathon on public weather data will be held in Toulouse (December 2–4).
3. From prediction to decision support for crisis management
Servane Kerene-Vénière , CEO of Keyros, emphasized AI’s role in transitioning from forecasting to operational resilience:
AI is now used to:
- Support crisis cells in real time
- Filter signal from noise
- Identify blind spots
- Estimate how many people will be affected
- Simulate external shock scenarios on supply chains
With tools like Power BI France, preparation time for crisis simulations has been reduced from several days to only a few hours.
4. Financial sector: climate risk is becoming uninsurable without AI
Aldrick Zappelini, Chief Data & AI Officer at Crédit Agricole, highlighted structural disruptions:
Concrete impacts:
- 30% of French territory is affected by soil shrink–swell risk (clay-related damages).
- Agricultural yields will shift significantly per IPCC scenarios.
- Traditional insurance models are breaking.
Key shifts underway:
- Creation of new climate-risk-focused business units
- Use of generative AI to analyze client sustainability reports
- ESG risk anticipation becoming a standard credit process
Strategic transformation:
- AI-driven advice to clients on:
- Where to build factories
- How to secure long-term logistics
- How to adapt agricultural practices
- Massive upcoming challenges in farm transmission due to retirement waves.
5. AI models: where real climate value sits
Etienne Grass, Global Chief AI Officer – Capgemini Invent, drew a clear technical boundary:
Two AI families matter when it comes to climate change’ impact :
- Low-impact for climate
- Pure generative AI (LLMs) → limited direct value for physical climate problems according to him.
- High-impact AI
- Computer Vision:
- Satellite image analysis
- Parcel-level land mapping
- Detection of pesticide-intensive farming
- Ontological models:
- Material sustainability modelling
- Sustainable product design
- Computer Vision:
A major trend:
- Open-source simulation platforms built on scientific and public datasets to:
- Stress-test supply chains
- Simulate portfolio climate exposure
- Redesign industrial ecosystems
6. The hidden cost of AI: energy, carbon, and resource pressure
A rare level of transparency was shared on AI’s environmental footprint:
Real cost of AI inference:
- ~€2.6 per kWh on average
- ~€4 per kWh on hyperscaler cloud platforms
- ~€0.4 per kWh on emerging “neo-cloud” infrastructures
Major warnings:
- Current AI cost structure is artificially low.
- Long-term prices will rise sharply.
- Energy and rare-earth material constraints are becoming structural.
Key facts presented:
- LLMs consume 2,600× more energy than traditional models.
- AI’s carbon footprint could grow from 2% to 7.5% of global emissions by 2030.
- Next-generation NVIDIA chips could reach 600 kW per unit.
Open-source tools now exist to:
- Calculate carbon impact per algorithm (GitHub-based)
- Track full AI lifecycle:
- Data centers
- Cooling systems
- Chip manufacturing
- Scope 3 emissions
A strong concept gained traction: “Frugal AI”
Less brute force, more efficiency.
7. Sovereignty and infrastructure: strategic divergence from hyperscalers
A strong French and European stance emerged:
Organizations like Météo-France and Crédit Agricole are adopting hybrid infrastructure strategies:
- Internal private clouds
- Sovereign infrastructure first
- External cloud capacity only when necessary
Core objective:
- Avoid long-term dependency on hyperscalers
- Preserve operational and data sovereignty
- Maintain critical infrastructure autonomy
8. Strategic synthesis
Key structural conclusions from the meetup:
- AI is shifting from innovation tool to critical climate risk infrastructure.
- Fast, AI-driven climate forecasting is now operational reality.
- The real battle is moving toward:
- Decision intelligence
- Infrastructure sovereignty
- Environmental accountability of AI itself.
- The paradox is now explicit:
AI is both a solution to climate risk and a growing contributor to it.
The next competitive advantage will not be raw AI power, but:
- Energy-efficient AI
- Sovereign infrastructure
- Traceable environmental cost
Straight assessment
This conference marked a clear inflection point:
- Climate risk is now treated as a first-order business risk.
- AI is no longer optional.
- But uncontrolled AI growth is not sustainable.
The industries that will dominate will be the ones that master frugal, transparent, and sovereign AI rather than brute-force scale.
