The Hidden Risk in Big Tech’s Trillion-Dollar AI Investment Surge
By Matthew Field, Senior Technology Reporter — 16 November 2025
In March, Nvidia CEO Jensen Huang took to the stage in his trademark leather jacket to unveil the company’s latest innovation: the Nvidia Blackwell Ultra, a supercharged graphics processor designed to power the next generation of artificial intelligence (AI) applications. Huang famously joked that he was Nvidia’s “chief revenue destroyer,” highlighting how the new chip’s immense power would render older models obsolete, leaving little demand even for prior high-performance GPUs.
This announcement came amid an unprecedented wave of investment in AI technology, with major tech giants such as Amazon, Meta, and OpenAI pouring trillions of dollars into acquiring Nvidia’s cutting-edge processors. This soaring demand propelled Nvidia’s revenue to an eye-watering $54 billion (£41 billion) in the most recent quarter and boosted the company’s market valuation to a staggering $4.7 trillion.
But are Nvidia’s chips really worth the hype?
Growing Concerns About an AI Bubble
Despite the enthusiasm surrounding AI hardware, concerns are mounting about a potential “AI bubble” — a scenario where the value of these advanced chips could depreciate faster than anticipated. This prospect is worrying both investors and industry analysts because it calls into question the accuracy of financial forecasts and the sustainability of current valuations.
Companies typically spread the cost of IT hardware like GPUs over several years by depreciating the asset’s value gradually. Recently, many tech firms have extended the assumed lifespan of Nvidia’s GPUs from three to five years, reflecting optimism about their ongoing utility and profitability. If these estimates prove true, the companies stand to report substantially higher profits. However, if the chips become obsolete more quickly — say, within two to three years — firms will be forced to write down their assets significantly, impacting earnings and valuations.
Echoes of Past Market Bubbles
Prominent investors wary of the AI boom have voiced sharp critiques. Jim Chanos, famed for foreseeing the collapse of energy giant Enron, pointed out after Meta’s July earnings that if GPU lifespans are closer to two or three years, then Meta’s profits could be “materially overstated.” Chanos likened the current enthusiasm for AI to the 1990s telecom bubble.
Michael Burry, well known from “The Big Short” for predicting the 2008 financial crisis, recently disclosed a $1 billion bet against the AI sector. Burry asserts that tech companies are inflating earnings by artificially extending asset lifespans and predicts the industry will understate depreciation by a colossal $176 billion by 2028. Further illustrating potential red flags, Amazon shortened its depreciation timeline for GPU assets from six years down to five earlier this year, costing the company about $700 million—a move seen by some as a precaution against overvaluation.
Risks for Emerging ‘Neocloud’ Start-Ups
While established tech giants might weather hits from accelerated depreciation, newer, debt-driven data center operators — often dubbed “neocloud” companies — face heightened risks. CoreWeave, a provider of data center space to major AI firms, announced it expects its GPUs to last up to six years, a notably ambitious projection. This estimate drew criticism from short-seller Kerrisdale Capital, which accused CoreWeave of employing “aggressive accounting” that “grossly flatters” its financial results.
In comparison, competitors like Nebius use more conservative four-year depreciation cycles. CoreWeave defends its approach, citing “deep experience” and “real-world usage data” showing that contracts for older GPUs have renewed at 95% of their original value—a sign that even aging chips hold significant market worth.
Investors remain cautious, however. CoreWeave’s shares plunged 16% following a recent earnings disappointment, with its valuation dropping one-third amid a broader tech sell-off since late October.
A More Nuanced View: Older GPUs Still Adding Value
Not all experts agree that the accelerated depreciation concerns indicate an impending crash. Richard Windsor, an industry analyst, points out that Google continues to use seven-year-old Tensor Processing Units (TPUs) effectively in its data centers. According to Windsor, as long as AI demand stays strong, older chips will find roles in handling less intensive tasks rather than becoming worthless.
David Harold, senior analyst at Jon Peddie Research, calls the “AI depreciation doom” narrative “overcooked.” He notes that older accelerators don’t simply vanish; they’re often reassigned to less demanding workloads or resold, continuing to generate revenue.
Although the most cutting-edge AI research may require Nvidia’s newest chips, a substantial share of everyday AI applications can function well on older hardware. This flexibility may soften the impact of accelerated depreciation fears.
The High Stakes of Nvidia’s GPUs in AI’s Future
Regardless, the valuation of Nvidia’s graphics processors extends beyond the company’s own balance sheet. These chips underpin the trillion-dollar AI investment spree that is reshaping the technology landscape. Should their value erode faster than expected, the ripple effects could spell trouble not only for Nvidia but also for the vast ecosystem of firms relying on this hardware to power AI innovation.
For tech behemoths like Meta and Amazon, the risk may be manageable, but for emerging companies and startups profoundly dependent on their GPU investments, the stakes are considerably higher. The final outcome could prove decisive in determining whether AI’s trillion-dollar boom stabilizes into sustainable growth or stumbles into an economic bust.
For more insights and news on artificial intelligence and the technology sector, stay tuned to The Telegraph.
