The $517 Billion AI Dilemma: Are Chips Aging Too Quickly?
In the year 2025, the tech industry's relentless pursuit of AI excellence has led to a staggering investment of approximately $400 billion (S$517 billion) in specialized chips and data centers. However, as the dust settles, doubts are emerging about the wisdom of such an unprecedented financial commitment.
At the core of these concerns lies a critical question: Have we been too optimistic about the longevity of these specialized chips before they become obsolete? With persistent fears of an AI bubble and the US economy heavily reliant on the artificial intelligence boom, analysts are warning of a potential harsh and costly reality check.
Renowned investor Michael Burry, made famous by the movie The Big Short, described the situation as "fraud" on X in early November. Before the ChatGPT-induced AI wave, cloud computing giants typically assumed their chips and servers would last around six years.
But Mr. Mihir Kshirsagar from Princeton University's Center for Information Technology Policy challenges this assumption, arguing that the combination of wear and tear and technological obsolescence makes the six-year estimate difficult to sustain.
One of the key issues is the rapid release of new, more powerful processors by chip manufacturers, with Nvidia leading the way. Less than a year after launching its flagship Blackwell chip, Nvidia announced the upcoming Rubin chip for 2026, boasting performance 7.5 times greater.
At this pace, Mr. Gil Luria of financial advisory firm D.A. Davidson warns that chips can lose 85% to 90% of their market value within just three to four years. Nvidia CEO Jensen Huang himself acknowledged this, stating that when Blackwell was released, nobody wanted the previous generation of chips anymore.
AI processors are also experiencing higher failure rates, according to Mr. Luria. "They run so hot that sometimes the equipment just burns out," he explained. A recent Meta study on its Llama AI model found an annual failure rate of 9%.
Both Mr. Kshirsagar and Mr. Burry believe the realistic lifespan of these AI chips is only two to three years. Nvidia, however, pushed back in an unusual November statement, defending the industry's four-to-six-year estimate as based on real-world evidence and usage trends.
Mr. Kshirsagar believes these optimistic assumptions have led to "artificially low" costs, and consequences are inevitable. If companies were forced to shorten their depreciation timelines, it would have an immediate impact on their bottom line and profits, warns Mr. Jon Peddie of Jon Peddie Research.
"This is where companies can get into trouble with creative bookkeeping."
The potential fallout could have far-reaching effects on an economy increasingly dependent on AI, analysts caution. Mr. Luria is not concerned about tech giants like Amazon, Google, or Microsoft, which have diverse revenue streams. Instead, his focus is on AI specialists like Oracle and CoreWeave, which are already heavily indebted while racing to buy more chips to compete for cloud customers.
Building data centers requires significant capital, and if these companies appear less profitable due to more frequent equipment replacements, it will become more expensive for them to raise the necessary capital, Mr. Luria points out.
The situation is further complicated by some loans using the chips themselves as collateral. To mitigate the impact, some companies hope to resell older chips or use them for less demanding tasks than cutting-edge AI.
"A chip from 2023, if economically viable, can be used for second-tier problems and as a backup," Mr. Peddie suggested.
And this is the part most people miss: the potential consequences of an AI bubble bursting. Are we heading towards a scenario where the AI industry's rapid growth is built on a foundation of artificially low costs and optimistic assumptions? The debate is ongoing, and the future of AI investment remains a topic of intense discussion and speculation.
