TLDR: A new report by The Shift Project highlights the alarming and unsustainable energy consumption trajectory of Artificial Intelligence, particularly generative AI, and data centers, projecting a significant increase in global carbon emissions by 2030.
A recent report by the French think tank, The Shift Project, has issued a stark warning regarding the escalating energy demands and environmental impact of Artificial Intelligence and its underlying data center infrastructure. Presented after 15 months of extensive research, the 2025 report, which DirectIndustry e-Magazine covered, underscores a critical challenge: reconciling the rapid deployment of AI with global sustainability goals.
The report identifies data centers as the ‘true backbone of the digital economy,’ hosting a wide array of services from traditional web hosting to advanced AI and generative AI applications. These facilities rely on substantial physical infrastructure, including electrical hardware, GPUs, processors, and cooling systems, all of which are significant energy consumers. Maxime Efoui-Hess, coordinator of the Digital Program at The Shift Project, emphasized that ‘the new structuring dynamics of digital technology are linked to the physical computing infrastructures for artificial intelligence,’ necessitating this in-depth analysis.
The findings paint a concerning picture of an ‘unsustainable climate trajectory.’ According to Marlène de Bank, a digital research engineer at The Shift Project, the global carbon footprint of data centers could quadruple by 2030, soaring from 250 MtCOâ‚‚e/year in 2020 to an estimated 920 MtCOâ‚‚e/year. This projected increase is equivalent to ‘twice France’s annual emissions across all sectors,’ highlighting the immense scale of the problem.
This warning aligns with broader industry concerns. The International Energy Agency (IEA) reported that electricity consumed by data centers could more than double by 2030, potentially exceeding 1,000 terawatt-hours (TWh). Goldman Sachs offers an even more aggressive forecast, predicting a 50% increase in global power demand from data centers by 2027, and up to 165% by the end of the decade. Dr. Fatih Birol, Executive Director of the IEA, cautioned that ‘There is no AI without electricity,’ and that ‘The energy industry is caught unprepared for the speed and demands of the AI industry.’
The surge in energy consumption is largely driven by resource-intensive AI workloads, including large language models, which require powerful CPUs and GPUs and advanced liquid cooling systems. Digital Realty projects a 12% increase in data center energy consumption by 2030 due to these demands. The World Economic Forum estimates that training a single model like GPT-3 required approximately 1,300 megawatt-hours of electricity.
Also Read:
- Hitachi and OpenAI Partner to Advance Energy-Efficient Generative AI Solutions
- AI Pioneer Geoffrey Hinton Issues Urgent Warning on Unchecked Artificial Intelligence Development
While AI holds potential for reducing overall greenhouse gas emissions in sectors like power, transport, and food (estimated at 3.3 to 5.4 GtCOâ‚‚e annually by 2035 by the London School of Economics and Systemiq), the report stresses that the environmental costs of powering AI must be addressed. The Shift Project’s recommendations aim to inform public debate and promote strategies to reconcile AI development with environmental sustainability, urging a focus on energy efficiency in data center design and operation, smarter cooling systems, and AI-driven power management.
