The Insatiable Appetite of AI
Artificial intelligence, particularly the generative models that have captured global attention, operates on a scale of complexity that dwarfs traditional computing. Training a single large language model requires sifting through enormous datasets, a process
that consumes vast amounts of computational power. But the demand doesn't stop there. As these models are deployed for everyday use—a process known as inference—they continue to require significant processing resources. This dual demand from training and inference is creating an insatiable appetite for specialised hardware, leading to what some are calling the largest infrastructure buildout in history. Projections show annual spending on AI infrastructure could reach $1.5 trillion by 2031, a figure that requires a massive expansion of the physical systems that underpin the digital world.
The Reign of Compute Hardware
At the heart of the AI boom are Graphics Processing Units, or GPUs. Originally designed for video games, these chips are uniquely suited for the parallel processing that AI workloads demand. This has turned companies like Nvidia into kingmakers, with their hardware becoming an essential and often bottlenecked resource. Hyperscale companies like Amazon, Google, Microsoft, and Meta are projected to spend around $800 billion on capital expenditures in 2026 alone, a huge portion of which is dedicated to acquiring these powerful chips. The demand is so intense that supply is expected to remain a constraint for the next few years. While GPUs currently dominate, the investment extends to the entire hardware ecosystem, including high-bandwidth memory and advanced networking equipment needed to connect massive clusters of chips. This hardware component is expected to account for the majority of the nearly $7 trillion companies are expected to invest in building and upgrading data centers by 2030.
Data Centres: A New Real Estate Boom
All this powerful hardware needs a home, and that has ignited a global construction boom for data centres. These are not your average office buildings; AI-specific data centres are highly specialised structures designed to handle extreme power densities and cooling requirements. Spending on data centre construction in the US has surged, with a record-breaking $22 billion spent in June 2026 alone. The global market is seeing a similar explosion, with demand predicted to nearly triple by 2030. This surge is turning data centres into a hot real estate and construction investment, creating thousands of jobs for skilled trades like electricians. However, this rapid growth is also putting a strain on land and labour resources in hub regions, driving up costs and creating competition with other industries.
Powering the Revolution
Perhaps the most significant challenge and investment area is power. AI data centres consume electricity on a scale that is straining electrical grids worldwide. Global data centre electricity consumption is projected to grow significantly, with some estimates suggesting it could double between 2025 and 2030. By 2030, data centres could consume up to 17% of total electricity in the United States. The issue is not just the amount of power but its volatility; AI workloads can cause massive, sudden spikes in demand that destabilise grids built for predictable loads. This has made reliable access to power the single most critical factor for data centre location and investment. As a result, tech giants are now making direct investments in energy generation, exploring everything from renewable sources to next-generation power solutions to ensure their multi-billion dollar AI factories can stay switched on.
