The AI-driven capital expenditure boom sweeping through tech giants, especially hyperscalers, may be entering a perilous phase, Deutsche Bank analysts Adrian Cox and Stefan Abrudan report.
The bank’s research highlights a growing disconnect between sky-high AI infrastructure investments and the financial returns these companies can realistically achieve, which also raises concerns about overvaluation and the risk of a valuation reset in the tech sector.
The emerging tension, according to the analysts, highlights the challenging economics behind the race to build AI data centers and the sustainability of market enthusiasm around AI.
Hyperscalers like Microsoft, Amazon, Meta, and Alphabet’s Google have gone into overdrive with the investing of hundreds of billions of dollars in specialized and custom AI data centers, computing hardware, and software to power large-scale machine learning models.
Deutsche Bank estimates that spending on data centers alone for 2025 could reach an eye-watering $400 billion, a figure Praetorian Capital says is comparable to the GDP of mid-sized nations like Malaysia or Egypt. These investments include vast outlays of land, buildings, massive power infrastructure, and the GPUs vital for AI model training and inference.
And while the spending spree fuels rapid advancements in AI capabilities and consumer adoption, the German bank cautions that the financial realities of such investments are far more sobering than the hype suggests. The pace at which these highly specialized AI assets often depreciate means the costs to maintain and replace infrastructure are mounting faster than the revenue generated from AI-related businesses.
Again drawing from insights from hedge fund Praetorian Capital, Deutsche Bank highlights a stark imbalance in hyperscalers’ AI economics. Annual depreciation for data centers built in 2025 is expected to be over $40 billion, yet the revenues attributed to AI from these assets are estimated at just $15 to $20 billion. This means that the cost of capital consumption is roughly double the revenue they generate.
Historical echoes and risks
Deutsche Bank draws uncomfortable parallels to past bubbles, such as the dotcom boom, which despite laying the foundation for the internet era, resulted in catastrophic investment losses for many.
The bank warns that while current spending by these tech giants is funded mostly through operating cash flow rather than debt, the capital intensity poses risks to profitability and valuation. A “fear of missing out” on AI’s transformative promise pushes companies to continue investing aggressively, even at the risk of destroying shareholder value.
What’s next for investors and hyperscalers?
Despite these warnings, Deutsche Bank notes that leading hyperscalers remain relatively resilient, with healthier cash flows and more prudent valuations than previous tech bubbles. However, the bank emphasizes the need for investors and management teams to critically assess returns from AI investment versus its hype.
The bank also points to potential technological innovations that could disrupt the current trajectory, such as more efficient AI models developed on less costly infrastructure, which could upend the massive capital requirements today. An example in this case might be how Deepseek shook the world earlier in the year upon its competitive, low-cost release.
For investors watching the AI boom closely, Deutsche Bank’s analysis is a timely reminder of the importance of financial scrutiny amid the excitement. While AI offers big promises, the fundamental economics of hyperscaler investments reveal structural vulnerabilities that must be navigated carefully in the coming years.
