Moody's: The Challenges Hyperscalers Face with the AI Boom

As the AI race intensifies, it is driving significant requirements for hyperscalers.

Such technological advancements, particularly in AI, prompt an increased need for hyperscaler infrastructure integration.

These technology enterprises are confronting challenges linked to data centre capacity as innovative AI models, like the latest large language models (LLMs), push infrastructure capacities to new limits.

Notably, AI models demand exponentially more computational power compared to earlier models.

Technology giant Nvidia reports that responses from reasoning models now require more than 100 times the computing power than before.

Moody's recent analysis highlights the expansion imperative in data centre capacity, noting that "growth in reasoning models creates a challenge for AI service providers to balance high throughput volumes with quick query response times, substantially raising AI data centre capacity demand."

This heightened demand is further fuelled as open-source AI models simplify access, prompting a surge in new startup ventures.

Moody’s notes that “all major AI labs serving popular models are short of capacity as demand for inferencing tokens has exploded, requiring them to cap the usage”.

Tracking the hyperscaler spending spree

Capital investment by key US hyperscalers such as Amazon Web Services (AWS), Microsoft, Alphabet, Meta and Oracle increased by 66% to US$211bn in 2024, driven chiefly by investments focusing on AI infrastructure.
This excludes finance leases, suggesting that the total investment is indeed much larger.

For example, Microsoft's future operating and finance lease commitments for data centres that are still underway have escalated to US$105bn by year-end 2024.

Amazon's AI revenue stream is estimated to have a significant annual revenue rate, with AWS reporting a 19% increase in revenue to US$108bn for 2024, supported by the ongoing enhancement of its AI capabilities.Mood's' analysis demonstrates that US tech companies accounted for 44% of global installed data centre capacity as of 2023, prior to the exponential growth in AI demands.

Meanwhile, the emergence of neo-cloud startups, such as Coreweave and Lambda, indicates a shift as capital investments rise to match the surging demand for AI-driven services.

The sovereign AI boom

Countries are allocating extensive resources to establish national AI infrastructure, targeting native datasets for training.

Notably, China has directed US$138bn towards emerging technologies, while the European Union has set aside €200bn (US$235.7bn) for its InvestAI schemes, aimed at building AI data centres.

Moody’s foresees most investments will benefit local developers, promoting the development of domestic AI models over reliance on US-based technologies.

Examples of this trend include Korea's HyperCLOVA X and Italy’s Colosseum, showcasing the global push towards domestic AI infrastructure.

Additionally, Microsoft plans significant investments exceeding US$35bn over three years to bolster AI and cloud data centre resources.

Data centre risks to evolve, Moody's says

Moody’s outlines that data centres, as long-term investments, will face challenges due to the volatile nature of AI development and market conditions.

The lack of adequate data centre capacity is anticipated to persist, impacting revenue expectations despite contractual commitments.

Moreover, the ongoing uncertainty concerning tariffs on semiconductors and IT hardware poses further risks to this burgeoning sector.

The capital intensity of US hyperscalers has seen a steep increase due to AI investments, reflecting in heightened capital expenditure relative to revenues.

This dynamic environment, characterised by the rapid evolution of AI technologies and competitive pressures, demands that hyperscalers maintain strategic foresight to assess the return on their extensive infrastructure investments.