AI is reshaping the global economy, but its rapid expansion depends on energy-hungry infrastructure that raises urgent questions about sustainability, equity, and access, forcing policymakers and industry leaders to confront the physical costs behind digital innovation.
Often presented as a purely digital revolution, AI is in fact supported by vast networks of data centers, servers, electricity grids, cooling systems, and water resources. As adoption accelerates, the debate can no longer focus only on innovation, competitiveness, or productivity. It must also examine the infrastructure that makes AI possible, and the environmental and social pressures that infrastructure creates.
AI Is Not as Digital as We Think
When people interact with AI systems, they rarely consider the thousands of servers operating behind every query. Large language models, cloud computing platforms, and advanced AI applications require enormous computational power. Training and operating these systems depends on vast networks of data centers distributed across the world.
These facilities consume significant amounts of electricity and require sophisticated cooling systems, often involving substantial water use. As AI adoption accelerates, so does the demand for additional computing capacity: according to the IEA AI report, in the base scenario for 2030, the total stock of data centers will rise by more than 60%. Yet electricity demand is expected to grow faster, with data centers electricity consumption projected to more than double. This is due to the increase of the share of accelerated facilities, which are used for computational intensive tasks, such as powering Large Language Models (LLM) or deep learning process, compared to standard facilities, used for lighter tasks, such as cloud computing or web hosting.
This reality challenges a common misconception: that digital technologies exist independently from physical constraints. In practice, every AI-generated answer relies on infrastructure that consumes resources, occupies land, and requires continuous energy inputs.
The AI revolution is therefore not only a software revolution. It is also an infrastructure revolution.
Can Technology Escape the Laws of Physics?
Much of the public narrative surrounding artificial intelligence is built on technological optimism. AI is expected to increase productivity, accelerate scientific discovery, improve healthcare, and support economic growth. These expectations are not unfounded. The potential benefits are enormous.
However, technological progress does not eliminate physical limits.
The second law of thermodynamics remains valid regardless of technological advances. Every economic activity requires energy. Every transformation generates losses. No machine is perfectly efficient, and no production process is entirely free from environmental costs. Moreover, an increase in technology efficiency does not translate directly into the reduction of its environmental footprint. As technologies become cheaper, people are willing to increase their consumption: this is also known as the Jevons Paradox.
More efficient data centers might reduces the costs to run AI models. Company providing this service could decrease their costs and more people and firms could now use better AI models or they could use it more. Eventually, digital companies will be required to expand their facilities to follow the demand and this increases the carbon footprint of the sector.
Recently, a Chinese study has confirmed this trend between 2010-2022, however, this does not necessarily imply this trend will continue.
This does not mean that innovation is futile. On the contrary, efficiency improvements are essential. But it does mean that sustainability cannot be reduced to marketing slogans or corporate communication campaigns.
The debate should not focus on whether AI is beneficial—it clearly can be. The question is how to maximize its benefits while minimizing its environmental footprint. Ignoring this challenge would be a mistake. Physics is not negotiable.
Who Benefits and Who Bears the Costs?
One of the most important questions concerns distribution.
The benefits of artificial intelligence are often global. Businesses improve productivity, researchers gain access to powerful tools, and consumers enjoy new services. Yet many of the costs associated with AI infrastructure remain concentrated in specific locations.
Communities hosting large data centers may experience increased pressure on electricity grids, water resources, and local infrastructure. In some regions, concerns have emerged regarding resource allocation, environmental impacts, and long-term sustainability.
Nowadays, the electric grid is recognized as a bottleneck for further data center development. The IEA in its grid report, shows a mismatch between the time required to expand the grid capacity and the time to expand energy generation and server computational power. While the first process can take between 5 and 15 years, the second up to 5 years.
The slow development of new grid infrastructures combined with the faster demand for connecting new loads in the network creates a saturation issue. The demand for electricity grows faster than the capacity to supply it and this may lead to an increase in electricity prices for families and enterprises. This trend has been confirmed by an analysis conducted by the Federal Reserve, division of Dallas. The report argues that electricity prices have been increasing in the US in the last years, after a period of stability. The trend is mostly concentrated in areas where bigger data centers are concentrated.
This creates a familiar economic dilemma: benefits are broadly distributed, while some costs remain local.
The challenge for policymakers is not to stop technological development but to ensure that communities hosting critical digital infrastructure are not left carrying a disproportionate share of the burden.
Transparency, environmental reporting, and meaningful engagement with local stakeholders will become increasingly important as demand for AI infrastructure continues to grow.
Can AI Remain Accessible to Everyone?
While environmental sustainability is a critical issue, accessibility is equally important.
Artificial intelligence is rapidly becoming a tool for education, research, entrepreneurship, and professional development. Access to AI systems may soon become as important as access to the internet itself.
This raises an important policy question: who will benefit from AI, and who risks being left behind?
Competition between companies and countries is driving remarkable innovation. However, the long-term success of AI should not be measured solely by the power of the most advanced models. It should also be measured by the extent to which these technologies remain accessible to society as a whole.
Universities offer an interesting example. For decades, public institutions have provided students and researchers with access to libraries, scientific databases, and educational resources. As artificial intelligence becomes an increasingly important research tool, ensuring broad access to advanced AI systems may become part of the infrastructure required for modern education.
If access remains concentrated among a small number of organizations or wealthy institutions, the technology could deepen existing inequalities rather than reduce them.
What Kind of AI Ecosystem Do We Want?
The future of AI will depend not only on technical capabilities but also on governance choices.
The challenge is to balance three objectives simultaneously: encouraging innovation, managing environmental impacts, and ensuring broad accessibility. These goals should not be viewed as competing priorities. A sustainable AI ecosystem requires these all three.
To develop AI as a widely available and productive tool for firms and people, while limiting its environmental footprint, policymakers can adopt various policies. Two examples are discussed below:
A first policy that can decrease the saturation of the electric grid is requiring companies to adopt autonomous electricity sources while building new energy intensive facilities. By doing so, big tech companies would decrease negative externalities on local communities, which are seeing electricity prices growing. Such measures are already being discussed: the European Commission, in its paper Strategic Roadmap for Digitalisation and AI in the Energy Sector directly cited the co-location of energy sources in proximity to data centers.
A second potentially helpful policy is the carbon tax. Treating CO2 as a cost can incentivize Big Tech companies to invest in the research for more efficient hardware and to adopt cleaner energy generation methods. However, due to highly fragmented frameworks between countries, an excessive tax could discourage new investments in the country where it is applied, in favor of less regulated ones. This is the result the German Environmental Agency discussed in its report: Carbon leakage in AI-driven data center growth?
Nevertheless, data centers will remain essential. Investments in infrastructures will continue. Competition between companies and countries will intensify. Yet the discussion should extend beyond processing power and market valuations.
The future debate on artificial intelligence should focus not only on who builds the most powerful systems, but also on who has access to them, who bears their environmental costs, and how their benefits are distributed across society.
Conclusion
AI has the potential to become one of the most transformative technologies of the twenty-first century. But, behind every breakthrough lies a growing network of physical infrastructure that consumes energy, water, and resources. Recognizing these realities does not mean opposing innovation. It means ensuring that innovation develops in a way that is environmentally sustainable, socially inclusive, and economically responsible. The success of AI will ultimately depend not only on what it can do, but also on how fairly and sustainably we choose to build it.
Inside Telecom provides you with an extensive list of content covering all aspects of the Tech industry. Keep an eye on our Press Releases section to stay informed and updated with our daily articles.
