AI is neither naturally virtuous nor necessarily harmful. The real challenge is to go beyond beliefs and measure its real impact to arbitrate between innovation, performance and climate.
Artificial intelligence is now establishing itself as the strategic priority of many companies. Investments are increasing, use cases are being deployed at high speed and promises of productivity gains are fueling an unprecedented race for adoption.
However, in the midst of this acceleration, one question remains largely absent from management decisions: what will be the real carbon cost of this technological revolution?
Because behind the spectacular performance of AI models lies a much less visible reality: massive consumption of computing, energy and material resources. And as uses become more widespread, this subject could quickly become one of the main points of tension between technological ambitions and climate commitments.
But when it comes to AI and the environment, the debate often boils down to a clash of two beliefs. For some, AI is an ecological disaster in the making. For others, on the contrary, it constitutes a formidable accelerator of the transition. Between these two readings, one reality remains: in most organizations, certainties are often more numerous than evidence.
The BearingPoint study conducted among 510 executives shows that this concern is no longer theoretical. Nearly 40% of companies anticipate an increase of more than 30% in AI-related emissions in the coming years.
The paradox is striking.
On the one hand, AI is presented as an accelerator of the environmental transition. Energy optimization, predictive maintenance, waste reduction, improvement of logistics chains: the use cases are numerous and often relevant.
Two thirds of CIOs also consider that digital technologies could help reduce their organization’s overall emissions by 6% to 30%.
On the other hand, the infrastructures necessary for this transformation themselves generate a growing environmental footprint. Data centers, computing power, storage, equipment renewal: each new layer of artificial intelligence has an environmental cost that remains largely underestimated.
Should we therefore conclude that AI is incompatible with the climate ambitions of companies? No more. Above all, the data shows that organizations are still struggling to precisely measure what AI really brings or costs on the environmental level.
The subject is therefore no longer whether AI is good or bad for the climate.
The real question now is: how to move away from beliefs and enter into a logic of proof? Which uses of AI actually create more environmental value than they destroy?
However, the results of our study reveal a worrying management deficit.
Only 44% of companies report that a significant portion of their AI projects today generate a net positive environmental impact. Conversely, more than a quarter of organizations report that less than 10% of their initiatives achieve this objective.
In other words, AI is inherently unsustainable. Its impact depends on the decisions made upstream.
This is precisely where the challenge of the coming years lies.
For a long time, technological decisions have been guided by three main criteria: cost, performance and speed of deployment. From now on, a fourth indicator must be essential: the net carbon impact.
Each AI project should be evaluated according to a simple logic: are the emissions it generates lower than the emissions it avoids?
This approach seems obvious. However, only 35% of organizations today carry out a systematic assessment of the environmental impact of their technological projects. Even more worrying, only 9% carry out this analysis before and after deployment in order to measure the benefits actually obtained.
This is probably one of the most revealing takeaways from the study: there is no shortage of opinions from companies on the impact of AI. Above all, they lack the tools, data and processes necessary to confront these opinions with reality.
The consequence is clear: a large part of technological investments continue to be decided without complete visibility on their climate cost.
This situation profoundly redefines the role of IT departments.
The CIO is no longer only responsible for the performance of information systems. It is gradually becoming a key player in the company’s sustainability strategy.
Tomorrow, it will have to be able to arbitrate technological investments with regard to their environmental impact, to integrate climate objectives into digital roadmaps and, when necessary, to question certain projects that are nevertheless economically attractive.
However, organizations are still far from achieving this. Four in ten CIOs still do not participate in defining their company’s environmental objectives, and only 20% are actually involved in these discussions at executive committee level.
This lack of convergence between digital strategy and climate strategy constitutes one of the main blind spots in corporate governance today.
Added to this is an additional difficulty: the lack of transparency of the technological ecosystem.
More than 40% of companies indicate that they do not have the necessary emissions data from their suppliers, while only a minority considers this information sufficiently reliable to manage its commitments.
Without visibility on the emissions generated throughout the digital value chain, it becomes impossible to correctly assess its real footprint, particularly on scope 3.
How can we overcome beliefs when the information necessary to measure real impacts itself remains incomplete or difficult to access? Here again, the problem is not just one of corporate will. It is also that of access to reliable and usable data across the entire value chain.
Purchasing policies will therefore have to evolve. ESG criteria can no longer be considered as a simple element of compliance; they must become a factor in the selection and management of technological partners.
Finally, this transformation requires a change in maturity in the management of environmental performance. One in two companies today believe they do not have the necessary tools to effectively monitor their ESG indicators.
The challenge is no longer just to produce reporting. It is to create the conditions for real-time management of environmental impacts, in the same way that companies already monitor their financial or operational indicators.
AI represents a tremendous opportunity for economic acceleration and transformation. But it also raises a question of collective responsibility.
The decade ahead will not be one of adopting AI at all costs. It will be that of arbitrations.
For too long, the debate has pitted those who believe in AI against those who believe in sobriety. However, the issue is not about choosing a side. It is about giving ourselves the means to measure, compare and decide on the basis of facts rather than intuitions.
The organizations that truly benefit from this revolution will not be those that deploy the most artificial intelligence. They will be those who will be able to demonstrate that each technological investment contributes simultaneously to their economic performance and their climate trajectory.
Because tomorrow, the question will no longer be: “Are you using AI?”.
It will be: “What do you really know about its impact?”
Rémy Sergent,
Partner, BearingPoint
