The energy consumption of global data centres is set to double by 2030, driven by the increasing power needs of generative AI (Gen AI), suggests new research. According to a new report from Deloitte, this growth is expected to result in data centres accounting for 4% of global electricity use by the end of the decade. The study additionally indicates that the energy required for essential infrastructure such as storage systems, cooling mechanisms, and networking switches is projected to reach 96 gigawatts worldwide by 2026, with AI operations alone consuming over 40% of this total.
“Sustainability is as important as transparency for building trust, for ensuring robust governance, for transforming talent and for managing mature data life cycle, which are all instrumental in defining the Gen AI roadmap,” said Deloitte Romania consulting market leader Andrei Ionescu. “The report points out that, on average, a Gen AI prompt request uses 10 to 100 times more electricity than a standard internet search query. Given the use of this amount of power, hyperscalers and data centre operators must consider alternative energy sources, innovative cooling methods, and more energy-efficient solutions when designing data centres.”
Agentic AI set to transform organisational operations by 2027
A significant rise in agentic AI, which refers to software designed to perform tasks with minimal human input, is also anticipated. By 2025, about a quarter of organisations using Gen AI are expected to launch agentic AI pilots or proofs of concept, with this number predicted to increase to half by 2027. Compared to traditional machine learning techniques, these AI agents, based on large language models, offer a broader range of applications.
As Gen AI solutions become more prevalent, the issue of deepfake content, which involves AI-generated media that mimics reality, becomes increasingly important. Organisations are encouraged to enhance digital content security by adopting tools that utilise diverse data sets and promote cross-industry collaboration. The Deloitte report advises users to verify information with trusted sources and to use multi-factor authentication to safeguard against deepfake threats.
The rise of on-device Gen AI is another trend highlighted in the report. By 2025, more than 30% of smartphones shipped are expected to be equipped with GenAI capabilities. About half of personal computers will feature local GenAI processing power, up from an estimated 30% in 2024.
In the streaming services sector, the report notes that streaming video on demand (SVOD) has reached its saturation point, with US households averaging four subscriptions and European households at 2.5 as of 2024. A decline in standalone subscriptions is anticipated in 2025. However, SVOD revenues might still grow due to price increases, stricter password-sharing controls, and improved bundling strategies.
The market is likely to stabilise around two or three standalone direct-to-consumer SVOD providers per region, complemented by aggregators. This shift mirrors traditional pay TV models and is already evident in markets like the US, UK, France, and Central Europe. In these areas, about a quarter of all SVOD subscriptions are indirect, accessed via pay TV or telecom services. This trend towards aggregation could lead to cost reductions and a more sustainable streaming landscape, as per the findings of the report.
Separately, a Gartner study published in November 2024, showed that 40% of current AI data centres are expected to encounter operational limitations due to power shortages. The consulting firm noted that these issues will be exacerbated by the increasing demand for AI and GenAI technologies, which are significantly straining the power resources of data centres.
Read more: 40% of AI data centres to face power constraints by 2027
