As artificial intelligence (AI) continues to evolve, investment trends are shifting from initial enthusiasm for software tools to a more focused approach on the underlying data centre infrastructure necessary to support AI systems. Goldman Sachs’ recent analysis reveals a “flight to quality,” where investors prioritize firms that own and operate extensive data centre facilities and computing resources over those specializing in niche AI applications or experimental software.
The Growing Role of Data Centres in AI Development
Goldman Sachs predicts a rapid increase in spending on AI infrastructure as enterprises expand their computing power for training and deploying AI models. Leading hyperscale cloud providers are investing tens of billions annually in new data centres and advanced computing hardware, alongside expanded networking systems to accommodate the surge in AI workloads.
AI Workloads Reshape Demand for Computing Capacity
According to Goldman Sachs Research, AI tasks could represent nearly 30% of global data centre capacity within two years. This surge stems from the intensive parallel processing required for training large AI models and the continuous, stable computing power needed for inference tasks—when AI systems generate predictions or responses.
This demand is driving an unprecedented expansion of data centre capacity, surpassing growth rates seen in earlier cloud computing phases. Besides hardware, energy supply has emerged as a critical factor in supporting AI infrastructure growth.
Energy Consumption and Environmental Considerations
Goldman Sachs estimates that global data centre power consumption could increase by approximately 175% by 2030 compared to 2023, largely fueled by AI workloads. This increase is comparable to adding the electricity demand of a major industrialized nation to the global grid. Consequently, utilities and governments are evaluating new investments in energy infrastructure to meet this rising demand.
Choosing locations for AI data centres now prioritizes access to stable energy sources and high-capacity fiber networks. Some companies are situating facilities in remote regions to secure land and electricity more easily. Research highlights that cooling systems and geographic factors significantly impact energy and water usage, influencing the environmental footprint of AI infrastructure.
Infrastructure Challenges Influence AI Strategies
The complexity of constructing large-scale data centres involves navigating supply chains, acquiring land, establishing grid connections, and securing long-term energy agreements. Delays in electrical equipment availability and grid expansion add further challenges. These constraints underscore why investors focus on companies with established data centre networks capable of supporting AI workloads reliably.
As the AI market matures, the initial phase—where companies gained value merely by associating with AI—is giving way to a more discerning investment approach. Firms with robust infrastructure and sustainable revenue models, such as data centre operators and chip manufacturers, are increasingly favored, as their services underpin all AI applications.
Long-Term Implications for the AI Economy
The expansion of AI infrastructure brings new considerations for energy demand, grid capacity, and environmental impact, which are drawing attention from industry leaders and policymakers alike. The future AI economy may rely as heavily on energy production and cooling technologies as on algorithmic innovation and software development.
This evolving landscape is shaping the next chapter of the AI race, where infrastructure quality and sustainability become critical factors in determining success.
(Photo credit: Lightsaber Collection)
Related Articles
Family Offices Embrace AI for Enhanced Financial Data Analysis, Ocorian Study Finds
Web3 Expo Dubai 2026 Set to Revolutionize Global Innovation with Massive Participation and $230,000 in Prizes
ChatGPT: Comprehensive Overview of the AI Chatbot Evolution and Updates
Google Advances AI Image Generation with Launch of Nano Banana Pro Model
