Artificial intelligence (AI) is reshaping how companies operate and how much energy they consume. As hyperscalers announce multibillion-dollar data centre pipelines and Southeast Asia emerges as one of the world’s fastest growing hosts for AI infrastructure, tension has moved to the centre of investor conversations. Can companies pursue aggressive AI adoption and still hit their net zero commitments?

For Amundi, Europe’s largest asset manager, the answer is yes, but only if governance keeps pace with adoption.

“We believe the two are not mutually exclusive. Companies can do AI responsibly and still achieve their net zero targets and commitments,” says Sylvia Chen, head of ESG Research Asia at Amundi. “It’s how they do it. They have to really consider how AI will impact their net zero commitments, including higher energy usage, and how that energy is going to be sourced.”

Chen is quick to point out that AI is not merely a climate liability. Across sectors, Amundi is seeing credible applications that deliver genuine environmental gains.

In utilities, AI is being deployed for grid optimisation, assessing utilisation and cutting downtime on what has become one of the biggest bottlenecks to the energy transition.

In renewables, machine-learning forecasting of wind patterns and solar irradiance is helping operators manage intermittency, long cited as clean energy’s Achilles’ heel. In agriculture, precision farming that combines drones, AI chips and satellite imagery is optimising yields and fertiliser use across large acreages.

Perhaps the most underappreciated application, Chen says, is climate adaptation.

“Early detection, early warnings for flooding, monitoring of extreme weather. AI-enhanced satellite imagery is more accurate and gives us more time to prepare,” she notes.

That translates into avoided losses and damage from climate hazards, a benefit that accrues to companies and investors alike, says Chen.

None of that, however, resolves the footprint of AI itself as data centres are notoriously hungry for both energy and water, and the industry’s traditional efficiency yardstick, power usage effectiveness (PUE), no longer tells the whole story, Chen explains.

“You can run power from fossil fuel very efficiently. The PUE will still be low, but there will be huge consumption and emissions on the carbon side.”

That is why a newer metric, carbon usage effectiveness (CUE), is increasingly entering Amundi’s engagement conversations, alongside water metrics.

In one case Amundi can verify, roughly a third of a company’s data centres sat in water-stressed regions, prompting a shift from water cooling towards air cooling and alternative sources to build resilience.

The bigger constraint, however, may be the grid, says Chen. “You’d be amazed. Some of the bottlenecks we’re seeing in terms of grid access could be up to 10 years in some places.”

That backlog is driving the rise of behind-the-meter (BTM) solutions, meaning on-site or near-site power generation, increasingly paired with battery storage to deliver firm power around the clock, she says.

For data centre operators, years of idle waiting for grid connection is simply uneconomic. For investors, it raises pointed questions about where alternative energy will come from, how firm it is, and what the plan is to transition away from coal or gas dependence.

These questions land with particular force in Southeast Asia as Malaysia has become one of the major recipients of data centre investment, and with it comes growing scrutiny of the demands placed on energy, water and communities.

Chen argues the answer lies in system-wide planning rather than asset-by-asset thinking. “You can’t just select the site. Access to energy is one thing, water another, but also community engagement, the impact on the wider network and communities, alongside job creation. You have to consider all of that in totality.”

She points to real cases of communities protesting data centres near their homes over fears of rising electricity costs, a reminder that the social licence for AI infrastructure cannot be taken for granted. The discipline required, she notes, is no different from any major infrastructure project. It calls for rigorous environmental impact assessment and better communication with the people on the ground.

Amundi engages thousands of companies each year, escalating concerns through voting where necessary, and its AI conversations have evolved rapidly. What began several years ago as a technology-sector dialogue is now broad-based, spanning energy sourcing and water resilience through to product safety, online harms and even child rights, as AI applications reach everyday consumers.

“If you look at cybersecurity a couple of years back, it emerged as an operational risk, but now it’s strategic and integrated. We’d imagine AI will follow the same path, just a lot faster.”

In practice, Amundi’s expectations of investee companies follow a consistent framework covering board oversight of AI governance, integration with net zero strategy, clear policies and implementation, key performance indicators (KPIs) and monitoring, and public disclosure.

Human oversight remains non-negotiable, whether in bank credit scoring, where institutions must be able to explain to a customer how an algorithm arrived at a decision, or in AI-assisted recruitment screening, where final decisions should rest with people.

Regulation is also reinforcing that direction, says Chen. The risk-based framework of the European Union AI Act reaches beyond Europe’s borders, as companies selling products or services to EU-based clients fall within its scope.

“I would make it a bare minimum in terms of the assessments that need to be done,” Chen says.

Looking ahead three to five years, Chen wants to see companies move beyond headline AI commitments to demonstrating how those commitments will actually be supported.

“Not just how wide or how fast they adopt AI, but how realistic those commitments are, and what is needed on the ground in terms of energy and resources to support them,” she says.

Equally, responsible AI, from workforce impact to human oversight and governance, should sit at the core of strategy for any heavy adopter, she says.