Artificial Intelligence (AI) is rapidly becoming a core part of modern business. Organisations are using AI to analyse data, automate processes, forecast demand, and improve decision-making. While these technologies are often framed as digital and intangible, their environmental impact is very real.
As AI adoption accelerates, so does the energy required to power it. For businesses with climate commitments, understanding the carbon cost of AI is now an essential part of responsible growth and a credible sustainability strategy.
Why AI Has a Carbon Footprint
AI systems depend on large-scale computing infrastructure. The emissions associated with AI largely come from the electricity required to train models, run them at scale, and support the data centres they rely on.
Training advanced AI models is particularly energy intensive. Research from the University of Massachusetts Amherst found that training a single large natural language processing model can generate carbon emissions equivalent to the lifetime emissions of five average cars, including fuel consumption. While model efficiency has improved since then, overall demand has increased sharply, offsetting many of those gains.
Once deployed, AI models continue to consume energy every time they are used. Generative AI tools, recommendation systems, and predictive analytics all require constant computation. As organisations embed AI across multiple functions, energy demand grows steadily rather than remaining a one-off cost.
This activity is underpinned by data centres, which already account for an estimated 1 to 2 percent of global electricity consumption, according to the International Energy Agency. The IEA also notes that AI-driven demand could significantly increase energy use from data centres in the coming decade, particularly in regions where electricity grids are still heavily reliant on fossil fuels.
Why the Carbon Cost of AI Matters for Businesses
For many organisations, emissions linked to AI and digital services fall into Scope 2 and Scope 3 categories. These emissions are often poorly understood, under-measured or excluded entirely from carbon reporting. As a result, they can quietly undermine progress towards net zero targets and science-based commitments.
Regulatory expectations are also rising. Frameworks such as the Corporate Sustainability Reporting Directive (CSRD) and the Task Force on Climate-related Financial Disclosures (TCFD) place increasing emphasis on transparency, governance, and full value chain emissions. Digital infrastructure is becoming harder to justify as an unexamined blind spot.
There is also a reputational dimension. Customers, investors, and employees are paying closer attention to how emerging technologies are used. Responsible AI is no longer only about ethics and data privacy. It is also about environmental impact and long-term resilience.
Can AI Still Support Climate Action?
Despite its footprint, AI can play a valuable role in reducing emissions when applied carefully. The World Economic Forum has highlighted how AI can support energy efficiency, climate risk modelling, smarter logistics and more accurate carbon measurement. In these cases, AI has the potential to enable emissions reductions that outweigh its own energy use.
The challenge for businesses is to distinguish between high-impact, purpose-driven use of AI and low-value applications that add complexity and emissions without delivering meaningful benefits.
Reducing the Carbon Impact of AI in Practice
Managing the carbon cost of AI starts with visibility. Organisations should treat cloud computing, data storage, and AI workloads as part of their core operational footprint, rather than as abstract IT services. This includes engaging suppliers for better data and incorporating digital emissions into wider carbon accounting.
Technology choices also matter. Smaller and more efficient models often achieve comparable outcomes with significantly lower energy use. According to research published by Google and DeepMind, improvements in model efficiency and hardware design can reduce energy intensity by orders of magnitude when efficiency is prioritised from the outset.
Supplier selection is another key lever. Major cloud providers vary widely in their use of renewable energy and transparency around emissions.
Most importantly, AI strategy should be aligned with sustainability strategy. Decisions about automation, data use and innovation should be assessed in terms of long-term emissions impact, not just short-term productivity gains.
What This Means for the Future of Carbon Reporting
As AI becomes embedded across business operations, its carbon impact will attract greater scrutiny. The Greenhouse Gas Protocol is already considering how digital and cloud-based services are treated within Scope 3 accounting, and further guidance is expected as digital emissions grow.
Organisations that begin integrating AI into their carbon management approach now will be better placed to respond to future regulation and stakeholder expectations.
Making AI a Climate Asset Rather Than a Liability
AI is neither inherently good nor bad for the climate. Its impact depends on how, where, and why it is used. For businesses, the goal is not to avoid AI, but to use it intentionally, transparently and in a way that supports credible climate action.
By measuring digital emissions, choosing responsible providers, and aligning innovation with sustainability goals, organisations can ensure that AI strengthens rather than weakens their net zero journey.
Positive Planet supports businesses in understanding their full carbon footprint, including the often-overlooked impact of digital infrastructure and emerging technologies. This enables organisations to innovate with confidence while staying aligned with science-based climate targets.