The AI Data Centre Potential: The Urgent Need for Balance
AI data centres have been touted as the future of the industry, but these innovative campuses are set to face technical and financial challenges.
According to a new report by Moody’s Ratings, while the surge in hyperscale data centre capacity continues – on account of the global AI boom – identifying when this growth will level off has become increasingly difficult.
In fact, if these power assets are delayed or insufficient, Moody’s suggests that data centre growth will likely fall back to the moderate (5%) or fast (10%) growth scenarios.
“Surging growth in hyperscale data centre capacity will eventually level off, but identifying that inflection point has become increasingly difficult as AI data centre campuses emerge as another key growth driver,” the report states.
“Data centre developers and large tech companies, or hyperscalers, continue to invest heavily in large data centre projects. But the computing needs they will serve in the future remain uncertain given the constant flow of innovative technological breakthroughs, posing long-term credit risks to developers, landlords and investors.”
Driving long-term growth
Data centres that support AI applications are now having to account for increased hyperscale data centre capacity. Moody’s explains that this increase in computing power to train AI models is ultimately changing how data centre infrastructure is developed.
AI data centres consume as much electricity as a medium or large city, with many built in remote locations to access electricity that is reliable and cost-effective. However, Moody’s explains that these remote locations face greater construction risks on account of maintaining a skilled workforce throughout the year, in addition to increased transport costs.
New infrastructure requirements are different from traditional data centres, as they generate greater levels of heat that current facilities cannot remove fast enough. Legacy data centres were not necessarily built to support the high-power requirements of AI.
“The industry is now facing unprecedented demand for new infrastructure solutions to efficiently power, cool and support this next generation of compute and as a result, AI is fundamentally reshaping the architecture of IT infrastructure,” Rajesh Sennik, Head of Data Centre Advisory at KPMG UK, told Data Centre Magazine at the end of 2024.
Hyperscalers are increasingly developing AI data centre campuses that centralise gigawatts of computing capacity at single locations where reliable, low-cost power can be constructed.
For instance, Moody’s suggests the 5.6GW Wonder Valley project in Alberta, Meta’s 2GW data centre campus in Louisiana and OpenAI's Stargate AI joint venture will add multiple gigawatts of capacity by approximately 2029.
“As investment continues to pour into new data centres, some level of capital reallocation and retrenchment is inevitable,” the report says. “The hyperscalers that have been spurring the market's expansion continuously right-size their newly leased and owned capacity under development because much of this new capacity is being built in anticipation of future needs.
“As it becomes available, it may exceed or fall short of a hyperscaler’s current needs.”
Creating power for Gen AI is not without challenges
Providing computing power for generative AI (Gen AI) requires significantly more capital investment than previous technological breakthroughs. According to Moody’s, such a financial risk has increased as AI has accelerated the pace of innovation and now requires larger amounts of upfront capital.
Matching a data centre's design to its expected computing load has become crucial, with more turnkey facilities built to single-tenant specifications. These tenants may not need this space beyond their original lease term or may require significant capital investments at renewal to adapt to new computing needs, Moody’s explains.
With this in mind, rack densities are continuing to climb roughly 12 kilowatts per rack on average in 2024.
Critically, Nvidia’s rapid development of new GPU racks has driven more power usage within smaller spaces with each generation. As AI innovation continues, Moody’s suggests that racks could reach 1-5MW over time.
The report adds: “The emergence of Chinese startup DeepSeek, which startled the tech sector earlier this year by announcing that it had trained its competitive AI chatbot with fewer and less-advanced chips than those used by American rivals, illustrates how innovative breakthroughs can put capital investments at risk.”
Likewise, trade tensions are inevitably impacting the data centre supply chain, with rising tariffs in the US creating market uncertainty that wasn’t anticipated. Tariffs on essential materials and components increase costs and could potentially create financial pressures for existing data centres and delays for those under development.
Moody’s highlights that new data centre projects could face delays as developers reassess tariff impacts on costs and timelines.
Crucially, rising costs of structural steel and critical electrical components, coupled with delays or limits on rare earth mineral exports, may further complicate bringing new capacity online. The report says that while cash-rich hyperscalers will continue to ensure new developments, tariffs could pressure service margins over time.
It explains: “Abrupt changes in tariff policies increase uncertainty, making it challenging for data centre developers to forecast costs and timelines accurately.”
Explore the latest edition of Data Centre Magazine and be part of the conversation at our global conference series, Tech & AI LIVE.
Discover all our upcoming events and secure your tickets today.
Subscribe to our Weekly Newsletter for the latest insights, trends and expert analysis.
Data Centre Magazine is a BizClik brand