The winner of the planetarity stakes this week is 'Hyperscaler.' It's used today to describe the four or five companies rapidly building out data center infrastructure for AI. Amazon, Meta, Microsoft, Google, Oracle, IBM, xAI (and perhaps a couple more). They are investing eye-watering amounts of money ~ $400 billion this year. Which, after you account for inflation, is more than the cost of the entire Apollo mission, but spent in one year.
This may be the greatest investment boom in history. Some of us think it's a bubble, but before we judge it so, we want to understand the nature of the investment itself.
Why is it a planetary concept? Because:
it is drawing resources from across the planet: energy, rare earth minerals to build chips, supply chains for those chips, financing and deployment across the world.
And also the models so trained will, of course, be distributed to users across the world as well.
But hyperscaling is a bigger concept in the making than just its use for these five companies and their data center investments. It's an exceptionally rapid way to meet demand for a problem that is seen as super important. Today, of course, that is AI. Tomorrow, it could be carbon sequestration.
The first Daily Planet in the Hyperscalar series is a real estate industry report on Data Center investments.
The global data center industry in 2025 is poised for transformative growth, driven primarily by the relentless advancement of artificial intelligence. Billions of dollars invested in AI have fueled unprecedented demand for powerful, efficient data center infrastructure, with construction reaching record levels worldwide.
This surge is underpinned by rapid GPU advancements, enabling faster AI model training and iteration, which in turn accelerates innovation. However, power infrastructure bottlenecks, including lengthy transmission line build times and selective utility approvals, pose significant challenges to development.
To meet soaring energy demands sustainably, nuclear power, especially small modular reactors (SMRs), is gaining traction as a preferred clean energy solution, offering scalable and reliable power for data centers. Thermal management is also evolving, with liquid cooling becoming essential to handle increased power densities from advanced GPUs. New data centers are adopting liquid cooling as standard, while existing facilities retrofit to accommodate higher workloads.
Immersion cooling, though still emerging, promises efficient thermal solutions for the most power-dense environments. Financially, 2025 is set to be a record year for data center development financing, with an estimated $170 billion in assets requiring funding.