The rapid evolution of artificial intelligence (AI) is driving hyperscalers—tech giants like Microsoft, Amazon, Alphabet, and Meta—to invest heavily in hardware that quickly becomes obsolete, according to a recent report from Research Affiliates. As the AI arms race escalates, these companies are compelled to continually upgrade their data center infrastructure, transforming the landscape of capital expenditures (capex) in the tech industry. Research Affiliates’ CEO, Chris Brightman, illustrates this phenomenon as a shift from traditional asset management to a cycle of perpetual replenishment akin to supermarket inventory.
Brightman noted in a phone interview that the pace at which hyperscalers are churning through equipment is unprecedented. “They’re in an arms race where they need to replace their hardware very rapidly, in other words, restock their shelves in a hurry,” he explained. This urgency stems not only from the need to support large language models and advanced AI features but also from the financial strain it imposes. Despite their massive investments, these companies often incur losses on the AI products they sell, complicating their financial outlook. Brightman emphasized that the spending required to maintain competitive advantages could ultimately jeopardize profitability.
AI-related capex has surged dramatically, with estimates indicating an increase from $250 billion in 2024 to $650 billion this year, equating to approximately 2% of GDP. This seismic shift in capital allocation has led many to liken the AI boom to the industrial revolutions marked by steel and railroads. However, Brightman contrasts the longevity of traditional industrial assets with the fleeting nature of AI hardware, noting that while steel mills and rail tracks can depreciate over 40 to 45 years, hyperscalers are depreciating their GPUs and other essential equipment over a mere five to six years on their income statements. In reality, their useful lives are significantly shorter, with Brightman estimating that AI hardware loses its value in about three years.
One example highlighted in the report is the profitability of Nvidia‘s industry-standard H100 GPUs. In their second year, these GPUs generated $36,000 in annual profit, yielding a 137% return on investment. However, by year four, profits plummeted, with the GPUs incurring losses exceeding $4,400 and a negative ROI of 34%. Brightman concluded that the economic lifespan of AI hardware is vastly shorter than its accounting life, driven not by physical wear but by the rapid advancements in computing power offered by manufacturers like Nvidia and AMD.
According to Brightman, the craving for enhanced computing capabilities leads hyperscalers to invest in new technologies, as they face strict energy constraints. “Most of their spending isn’t growth capex; it’s ‘maintenance’ capex,” he stated. Despite this, the sheer scale of spending still allows for significant expansions in product and service offerings each year, albeit with only one-third directed towards actual growth initiatives.
The Hyperscalers’ Dilemma
The dilemma for these tech giants is stark: as they ramp up computing resources, their financial losses accumulate. Brightman pointed out that the Big Four have compelling reasons to continue down this path. For Amazon, the cloud business—a significant revenue driver—relies heavily on AI capabilities, which it struggles to recoup from customers. “If Amazon doesn’t stay in the arms race, they’ll lose the cloud business,” Brightman emphasized.
Microsoft faces similar pressures, particularly with its Microsoft 360 platform. As competition intensifies from Google‘s suite of applications, Microsoft recognizes that maintaining its customer base necessitates investment in AI features, even if those investments initially result in losses. Meanwhile, Alphabet must integrate AI advancements to protect its core search and advertising revenues amid fierce competition, exemplified by Microsoft’s challenge in the search engine market.
Meta, too, is not insulated from these market dynamics. The company must leverage AI to keep its social media platforms engaging and relevant, even as its advertising business faces competitive pressures. Brightman observed that while Meta’s current ad pricing does not fully cover the costs associated with its AI investments, the need to stay ahead remains paramount.
Despite the colossal influx of capital into AI, Brightman concludes that these investments may not translate into substantial profits for the tech giants. Rather, they function more as defensive maneuvers to safeguard existing market positions. “When capital turns over rapidly, and competition forces continuous reinvestment, extraordinary spending can sustain competitive position without creating value for shareholders,” he remarked.
The implications of this evolving landscape are significant. The rapid obsolescence of AI hardware highlights the challenges that accompany this new industrial era. While consumers and businesses may reap the benefits of enhanced AI capabilities, the organizations supplying this technology face a precarious balancing act between innovation and financial sustainability.
See also
Google Chrome and Microsoft Edge Introduce Local AI APIs for Enhanced Model Management
Perplexity AI’s Jesse Dwyer Declares Web Search ‘Primitive’ After 24 Years of Stagnation
Google DeepMind Unveils Gemini Robotics-ER 1.6, Enhancing Robot Spatial Reasoning by 10%
AI Chatbots Produce 49.6% Problematic Health Responses in New BMJ Open Study
Pollo AI Tops 2026 List of 5 Best AI UGC Video Generators for Scalable Ad Production
