AI is no longer optional for businesses; energy constraints are becoming a central business issue.
Artificial intelligence is now a cross-functional performance driver for all company functions. In the financial sector, AI is increasingly used for everything from estimating credit risk and combating payment fraud to setting insurance rates and assessing asset volatility. Failing to integrate AI inevitably leads to losses in productivity, service quality, user experience, and competitiveness. But a new factor has suddenly entered the economic equation: energy (the bill depends largely on the joules expended to run AI software).
The most visible AI models—particularly large general-purpose language models—were built by a few players with exceptional financial, material, and energy resources. For most companies, value creation no longer lies in the race for ever-larger systems, but in the operational use of AI, that is, in inference, as close as possible to business processes.
However, inference is a continuous, large-scale activity, directly correlated to usage volumes. Without proper management, it quickly becomes an uncontrollable cost center (energy, cloud, infrastructure, latency, supplier dependence). For equivalent business performance, choosing an oversized model can multiply energy costs by 10 or 20, without any real operational gain. Therefore, frugality is not a peripheral environmental issue: it determines the profitability, scalability, and economic sustainability of AI applications.
For Denis Beau, Premier sous-gouverneur de la Banque de France, Read more on link