As AI’s energy consumption rises, so do sustainability concerns. This report explores why this technology is so power-intensive, how to mitigate its impact and whether it can help reduce emissions. It also examines four possible scenarios for future energy consumption – from unchecked growth to greater efficiency – and their sustainability implications.

AI’s energy dilemma

The rapid growth of AI and particularly generative AI (GenAI) has triggered a boom in adoption across industries, including telecom. Enterprises are embedding the technology into their operations, products and services, while consumers integrate it into their daily lives. Meanwhile, data centres are scaling up to meet the rising demand for AI-driven workloads. However, this surge in adoption comes at a significant cost. AI systems, especially those powered by advanced machine learning (ML) models, are extraordinarily resource-intensive, consuming vast amounts of computational power and energy. Training and running these models often require high-performance graphics processing units (GPUs) and large-scale data processing, leading to increased energy consumption.

If you are not a subscriber, enter your details below to download an extract of the report

For data centre operators and certain industries (e.g., telecom), this presents a dual challenge. On the one hand, they must meet rising customer expectations for AI-driven services, maintain a high-quality user experience and performance, stay competitive and ensure their infrastructure can handle the growing demand. On the other hand, they are drawing ever closer to deadlines for net-zero targets. Data centres, which are the backbone of AI operations, are already energy-hungry facilities – and the proliferation of AI workloads could exacerbate their environmental impact. Hyperscalers such as Amazon, Microsoft, Meta and Google face mounting challenges due to the rising energy demands of data centres. Their emissions are primarily concentrated in scope 2 and scope 3, both of which continue to rise grow (see figure below).

Total CO2e emissions between 2019 – 2023

Source: STL Partners

The situation is further complicated by the global push for net-zero emissions. Regulators, driven by initiatives such as the Corporate Sustainability Reporting Directive (CSRD) by the European Union (EU), are mandating stricter climate disclosures, while investors increasingly tie financial incentives to measurable sustainability outcomes. Some customers now factor in environmental performance into decisions, making it a competitive differentiator in certain areas, while shareholders and employees demand both cost efficiency and meaningful climate action. Without clear strategies for sustainable growth, the AI boom risks escalating energy demands and undermining efforts to align technological progress with sustainability goals. In this report, we explore four potential scenarios for future energy consumption.

Future energy consumption scenarios in the AI age

Source: STL Partners

After assessing AI’s power demands versus its role in driving energy efficiency in data centres, chip manufacturing and telecom, this report takes a broader view to evaluate whether AI, when extrapolated to other industries, is likely to have a more positive or negative impact on global energy consumption.

Table of contents:

• Executive summary
  • Recommendations for key stakeholders
  • AI and sustainability: A delicate balance
• Introduction
• The energy-consuming reality of AI
  • AI’s growing carbon footprint
  • How much energy does AI consume?
  • The other costs of AI
  • AI investment is not slowing down
• Efforts to reduce AI’s environmental impact
  • Hyperscalers and data centre operators
  • Semiconductor chip manufacturers
  • Governments and regulatory bodies
  • Future opportunities and emerging technologies
  • End-users: Consumers and enterprises
• What does this mean for telecoms?
  • How are telcos offsetting AI-driven emissions?
• AI’s expanding energy footprint—what’s next?
  • Future energy consumption scenarios
  • Conclusion

Related research:

• Telco generative AI adoption tracker
• External pressures shaping the edge AI landscape
• The telecom techco: The role of network cloud, automation and AI
• Real-time automation: The future of transport & logistics

If you are not a subscriber, enter your details below to download an extract of the report

Kerina Naran

Kerina Naran

Research Analyst

Kerina is a research analyst at STL Partners and is now applying her advanced research and analytical skills from an extensive background in biomedical sciences to the dynamic telecoms and technology market. Kerina joined STL Partners after obtaining a PhD in Virology and Immunology from Imperial College London.