AI’s Fast-Moving Chip Cycle Threatens Profitability and Exposes Fragile Financial Foundations

Gerik

Summary:

As tech firms pour billions into AI hardware, analysts warn that rapid chip obsolescence and inflated depreciation assumptions are masking unsustainable costs, potentially inflating profits and exposing the industry to deep financial risk....

Shorter Lifespan, Bigger Problem: AI’s Hardware Dilemma

The AI gold rush has triggered a $400 billion wave of investment in chips and data centers this year, but concerns are mounting over whether the financial infrastructure supporting this expansion is built on realistic assumptions. At the heart of the issue lies an uncomfortable truth: the expected lifespan of AI chips estimated at five to six years by most hyperscale cloud operators is proving far too optimistic.

Michael Burry, famed for predicting the 2008 subprime mortgage crisis, has called out what he describes as “accounting fraud” in how tech companies depreciate AI chips. According to his estimates, actual chip lifespans are closer to just two or three years. By maintaining inflated depreciation timelines, companies like Meta and Oracle are allegedly overstating their profitability, with potential underreporting of $176 billion in depreciation costs from 2026 to 2028. If corrected, Burry projects that Oracle and Meta’s net income could be overstated by 26.9% and 20.8%, respectively, by 2028.

Obsolescence Accelerates with Nvidia’s Relentless Roadmap

Mihir Kshirsagar of Princeton's Center for Information Technology Policy echoes these concerns, emphasizing the dual challenges of physical wear and technological redundancy. Nvidia, the dominant player in AI GPUs, illustrates this perfectly: less than a year after launching its Blackwell architecture, it has already unveiled the upcoming Rubin chips offering 7.5 times the performance and expected to ship in 2026.

Gil Luria, head of technology research at D.A. Davidson, notes that this pace of innovation causes AI chips to lose 85–90% of their market value within three to four years. Nvidia CEO Jensen Huang reinforced this dynamic in March, admitting that the moment Blackwell arrived, demand for its predecessor Hopper effectively vanished. Even Nvidia's own November 2025 defense of a 4–6 year chip lifespan did little to ease skepticism, as field data continues to contradict such projections.

Hardware Failure and Heat: Physical Decay Compounds Financial Risk

Beyond obsolescence, hardware reliability is under pressure. According to Luria, AI processors are running hotter and degrading faster than traditional chips. In one Meta study involving its Llama model, annual failure rates for GPU clusters reached 9%. Jon Peddie, founder of Jon Peddie Research, warns that if companies are forced to shorten depreciation schedules, the hit to net profits will be immediate and substantial.

This presents a clear causal sequence: overstated chip lifespans lower accounting costs and inflate profitability, but as the physical and functional lifespan proves shorter, firms will be forced to recognize previously hidden losses undermining their financial health.

Financial Engineering and the Risk of Asset-Backed AI Loans

Compounding the issue, many AI infrastructure firms are leveraging chip inventories as collateral to secure financing. Luria warns that companies like Oracle and CoreWeave both heavily indebted and reliant on rapid AI infrastructure growth are especially vulnerable. If chip value collapses sooner than expected, their debt structures may be exposed, making future borrowing prohibitively expensive.

The correlation here is direct: when underlying assets like chips depreciate faster than expected, the risk profile of the loans they back increases sharply. This could trigger a broader credit tightening for AI infrastructure players, slowing expansion or forcing distressed sales.

Patchwork Solutions and Secondary Use Markets

To mitigate risk, some companies are exploring ways to repurpose older chips. Jon Peddie suggests that GPUs from 2023 could be reassigned to lower-priority tasks or used as backup compute in less latency-sensitive environments. However, these secondary uses will only partially recoup lost value and are unlikely to change the larger economic reality.

The AI chip boom is thus revealing a structural misalignment between technological evolution and financial planning. As depreciation assumptions unravel and replacement cycles tighten, the sustainability of current AI infrastructure economics is called into question.

The AI industry’s rapid growth may be built on shaky ground. With hardware becoming obsolete in as little as two years and financial models lagging behind technological realities, companies face the risk of profit distortions, funding shocks, and balance sheet erosion. While giants like Amazon or Google can absorb these stresses, debt-laden infrastructure specialists may face an existential reckoning. As investors, policymakers, and CFOs awaken to the true costs behind AI’s breakthrough era, the next wave of innovation may be defined not by speed but by financial resilience.

To stay updated on all economic events of today, please check out our Economic calendar

News, historical chart data, and fundamental company data are provided by FastBull Ltd.

Risk Warnings and Disclaimers

You understand and acknowledge that there is a high degree of risk involved in trading. Following any strategies or investment methods may lead to potential losses. The content on the site is provided by our contributors and analysts for information purposes only. You are solely responsible for determining whether any trading assets, securities, strategy, or any other product is suitable for investing based on your own investment objectives and financial situation.