Will Salesforce’s AI Cut Emissions or Raise Energy Use?

Salesforce, the cloud-based customer relationship management (CRM) platform, is examining the relationship between AI and sustainability.

In its AI sustainability outlook, the company outlines both the environmental challenges tied to growing AI adoption and the opportunities AI presents for improving energy and resource efficiency.

Sunya Norman, SVP Impact at Salesforce, said on Linkedin: “As a leader in agentic AI, it is Salesforce's imperative to ensure that AI is trusted, reliable and sustainable. 

“In this outlook, we explore the current landscape, our efforts and evolving insights so far and our preliminary path forward. We’re early in the journey, but the future is being shaped now. 

“By sharing our progress openly, we aim to spark transparency and inspire collective action. A sustainable future with AI is within reach and I’m optimistic about what we can achieve together.”

AI's energy use and environmental toll

The report says that by 2030, AI data centres could consume around 3% of the world’s electricity.

These facilities use large volumes of power to operate high-performance computing hardware and cooling systems, which prevent overheating during complex computations.

Gartner, a technology research firm, warns that up to 40% of AI data-centre projects may encounter power constraints as early as 2027.

Such constraints could hinder AI expansion, increase costs and disrupt the reliability of power supplies.

Today, around 56% of the energy used to power global data centres comes from fossil fuels.

As AI demand grows, this could lead to higher emissions if data centres do not shift towards clean energy.

The International Energy Agency (IEA) notes that without this transition, data centres could become the fastest-growing source of global greenhouse gas emissions.

AI systems also require large volumes of water.

Data centre cooling processes use water sourced from watersheds that are already under stress.

Around one fifth of these centres draw water from areas facing moderate to high levels of water scarcity.

In addition to energy and water use, building AI compute infrastructure involves mining critical minerals such as lithium and copper.

These processes are both energy- and water-intensive, while also producing pollutants that contribute to environmental damage.

Using AI to support sustainability

While AI places strain on energy systems, it can also offer tools for climate action and energy efficiency.

According to the IEA, AI could help reduce energy-related emissions by 5% by 2035.

Salesforce outlines key areas where AI can assist:

• Simplifying and improving complex system design

• Speeding up research and innovation

• Supporting behavioural shifts for sustainability

• Enhancing climate modelling and policy analysis

• Supporting adaptation and resilience planning

In energy systems, AI can make forecasting and maintenance more precise, optimise integration of renewables and deliver personalised insights to help consumers reduce usage.

In water management, AI can improve irrigation systems, boost wastewater treatment and monitor water quality more effectively.

AI agents in action

Salesforce highlights a number of projects using its AI innovation platform, Agentforce, to improve sustainability outcomes.

For instance, the charity Good360 is developing a resource matching agent that routes donations efficiently during natural disasters.

The system prioritises regions most in need.

Stephane Moulec, Chief Technology Officer at Good360, says: “Globally, a significant amount of goods that could be matched to disaster survivors end up going to the landfill. Good360 is here to change that.”

Meanwhile, Rare is piloting an AI-powered regenerative agriculture coach that gives real-time, localised advice to farmers based on crop type, weather data and soil conditions.

The system’s first phase supports 5,000 farmers and aims to reduce staff time by 40%.

Another example is community organisation Groundswell, which is using AI to expand access to community solar.

The initiative will help about 30,000 households reduce energy costs by up to 50%.

To manage AI's environmental impact, Salesforce sets out a framework based on three key areas: smart demand, efficiency and clean supply.

Smart demand
Not every AI model fits every task.

Salesforce urges companies to assess how much and what type of AI is truly needed.

Leaner models with simpler algorithms may be more appropriate for some use cases.

Transparency in how and when power is consumed, along with incentives for efficient use, are central to this approach.

Efficiency
Salesforce encourages the development of smaller, task-specific AI models.

Techniques like quantisation (reducing the precision of calculations), distillation (streamlining models by training smaller ones from larger ones) and pruning (removing redundant neural connections) are highlighted as ways to lower compute demand.

These methods allow companies to run AI on smaller devices, easing pressure on large data centres.

Clean supply
Companies should embed sustainability in procurement processes and favour suppliers that use renewable energy.

Investment in clean power and water sources for AI operations is also necessary.

Salesforce calls on businesses to support policies and infrastructure developments that reduce AI’s environmental footprint.

The company's outlook balances a warning about the environmental costs of AI with a message that it can play a part in tackling the same challenges it contributes to.