Global AI Data Centre Investment to Reach $3 Trillion by 2029

Investment bank Morgan Stanley has projected that worldwide spending on AI-supporting data centres will reach approximately $3 trillion by 2029, representing an unprecedented infrastructure investment equivalent to France’s entire 2024 GDP. The UK alone is expected to construct another 100 data centres over the coming years to meet surging AI processing demands.

Context and Background

AI data centres require fundamentally different engineering approaches compared to traditional server facilities. The core difference lies in the need for parallel processing, where expensive Nvidia computer chips must operate in extremely close proximity to minimise latency. Each metre of distance between chips adds nanoseconds to processing time, creating performance bottlenecks that accumulate across warehouse-scale operations.

These AI processing cabinets, costing approximately $4 million each, must be densely packed to eliminate latency and enable Large Language Models to break language into minute elements of meaning. However, this density creates extreme power demands measured in gigawatts, with irregular electrical spikes equivalent to thousands of homes simultaneously switching kettles on and off every few seconds.

Daniel Bizo from The Uptime Institute describes AI workloads as presenting “singular problems” at unprecedented scale, comparing the engineering challenge to the Apollo programme. This erratic power consumption pattern requires careful grid management, unlike traditional data centres that maintain steady background demand.

Looking Forward

Technology giants are addressing energy challenges through diverse approaches. Microsoft has committed $30 billion to UK AI infrastructure and is investing in nuclear power projects, including restarting generation at Three Mile Island. Google is pursuing carbon-free energy by 2030 through nuclear investments, whilst Amazon Web Services claims to be the world’s largest corporate renewable energy buyer.

Industry experts acknowledge concerns about “bragawatts” - inflated projections of AI infrastructure scale - yet maintain that AI represents a more substantial technological shift than previous innovations, including the internet. The physical nature of AI data centres provides a concrete foundation unlike speculative tech bubbles, though the current spending trajectory remains unsustainable long-term.

Source Attribution:

Share this article