The AI-Driven Global Data Center Boom
How AI is Driving Infrastructure Expansion & Energy Challenges
According to an Amazon Web Services (AWS) executive quoted in LinkedIn News on 4/25/24, “a new data center appears somewhere in the world every three days.”
The surge in AI projects has dramatically increased the demand for reliable and affordable power in data centers, leading to their establishment of data centers in far flung and unexpected locations (for example, South America). This rapid expansion is causing shortages of essential components like backup generators, which now face delivery delays of up to two years. More alarming is the fact that AI's substantial power requirements are straining the capacity of America's electrical grid.
AI datacenters are built specifically to handle the needs of AI workloads which are more demanding than their predecessors, requiring advanced infrastructure with high-performance servers, storage, networking, and specialized accelerators for efficient AI task processing. These centers require robust power connections and innovative cooling solutions to manage their substantial power use and heat generation. Because they are connected to the network by large data “pipes”, however, these facilities do offer location flexibility, which can reduce real estate costs, and entail a complex design to ensure reliability and manage risks.
Financially, hardware investment in AI datacenters, which topped $8 billion in 2023 (Source: Pitchbook), currently dwarfs the revenue of software the applications developed for these platforms. In fact, it may take up to three years to recover these capital costs and about five years to achieve substantial SaaS-style profit margins. While spending on cloud services and applications is expected to surpass hardware investments in 2024, the initial low profitability period poses significant challenges for smaller datacenter operators in maintaining competitive pricing and retaining customers. As a result, we do expect to see churn in these early days of AI driven data centers.
Also, startup cloud providers are undercutting hyperscalers on cost by offering 50%-70% savings on GPU hours using advanced Nvidia chips. These players have carved out a $4.6 billion niche in the $150 billion internet-as-a-service (aka “IaaS”) market, providing unique access to the latest AI technologies and fostering an ecosystem that includes colocation services and private cloud offerings. (Source: Pitchbook, April 2024)
As the technology continues to evolve, infrastructure and data center professionals will face “relentless” pressure to improve data center infrastructure technologies, focusing on advances that enhancement business technology platforms that will help them improve performance and control operating expenses. (Source: Gartner)
It’s certainly clear that the future of data centers is intertwined with emerging technologies like AI which, despite its early stage, promises to revolutionize complex computations and deliver new use cases and capabilities.