Reducing the carbon footprint in spite of the increase in calculations
Societe Generale provides financial services to 31 million individuals and professionals worldwide. Its corporate and investment banking business - SG CIB - offers global access to markets through solutions in equities, bonds and currencies, commodities and alternative investments.
Qarnot, a concrete lever to reduce its environmental impact
We have estimated the carbon footprint of our risk calculations at around 800 tons of CO2 per year, spread across several data centers.
We also use machine learning and a 2019 study showed that, on the training phase, the latest language analysis models cost more than 300 tons of CO2. With the increase in the amount of computations, we decided to take concrete action to reduce our carbon footprint.
To achieve this, we have implemented various actions:
- a responsible code policy,
- the use of simpler models if they achieve the same objective or auto-ML,
- optimizing the rate of use of the internal fleet to limit energy loss,
- the use of a mix of private and public clouds with the choice of service providers having demonstrated their energy efficiency.
We chose Qarnot for:
- the simplicity of integrating its computing platform into the existing infrastructure, while respecting the security requirements of the banking sector,
- Access to a low-carbon infrastructure of several thousand cores,
- the transparency of the information provided on the carbon footprint of each calculation, detailing each line of expenditure and energy savings.
Measurable and significant results
By using the heat generated by its simulation calculations, the bank reduced its carbon footprint by 76%.
These indicators were also used to index top management bonuses to the carbon footprint reduction of their perimeter.