The AI Trade’s New Foe: Why Rapid GPU Depreciation Is Pummeling Tech Stocks

The Financial Shadow Over the AI Boom: Depreciation Becomes the Market’s Most-Hated Word

For months, the stock market has been fueled by the seemingly limitless potential of artificial intelligence, driving valuations of chipmakers and cloud providers to historic highs. However, a silent, non-cash expense—depreciation—has emerged as the new source of investor anxiety, threatening to deflate the high-flying AI trade.

While the initial focus of market skepticism centered on the complex web of financial arrangements dubbed “circularity” in Big Tech—where investments are immediately recycled into cloud purchases—the underlying, more fundamental problem is the accounting reality of massive capital expenditure (CapEx) on specialized hardware. The speed at which cutting-edge Graphics Processing Units (GPUs) become obsolete is forcing a reckoning on corporate balance sheets, leading analysts to question the true profitability of the AI infrastructure race.

Understanding the Depreciation Dilemma: CapEx vs. Useful Life

The AI infrastructure buildout requires unprecedented levels of investment. A single high-end AI chip, such as Nvidia’s H100 or the newer Blackwell series, can cost tens of thousands of dollars. Cloud giants like Microsoft, Amazon, and Google are spending billions annually to acquire these chips and build the necessary data centers.

In accounting terms, these purchases are treated as capital expenditures (CapEx) and are not expensed immediately. Instead, their cost is spread out over their estimated “useful life” through depreciation. This non-cash expense reduces reported net income over time.

The Cost of Technological Obsolescence

The core problem facing the AI sector is the unprecedented pace of innovation. Historically, standard server equipment might be depreciated over five to seven years. However, in the hyper-competitive AI race, a GPU purchased today may be functionally obsolete in terms of performance per watt within 18 to 36 months.

Financial analysts are increasingly demanding that companies shorten the depreciation schedules for AI-specific hardware to reflect this rapid obsolescence. If a company shortens the useful life of a $30,000 GPU from five years to three years, the annual depreciation expense increases by 67%. This change directly impacts reported earnings, even if the company’s cash flow remains robust in the short term.

“The market is realizing that the massive CapEx fueling the AI boom isn’t just a one-time investment; it’s a continuous, accelerating treadmill. If the useful life of a $30,000 chip is effectively two years, the resulting earnings hit is substantial and permanent,” noted one analyst covering the cloud sector.

This shift in accounting perspective is crucial because investors primarily value growth companies based on future earnings potential. Higher depreciation expenses lead to lower reported net income, which can significantly reduce the price-to-earnings (P/E) multiples that investors are willing to pay for these stocks.

The Circularity Concern: Inflated Demand and Risk

While depreciation is the accounting mechanism causing the pain, the concept of “circularity” initially raised the alarm about the underlying demand structure for AI services. This term describes a pattern of dealmaking that can potentially inflate reported revenue figures for cloud providers.

How Circularity Works

1. Investment: A major cloud provider (e.g., Microsoft or Google) invests a significant sum in a promising AI startup.
2. Spending: The AI startup immediately commits to spending a large portion, often 70% or more, of that investment capital on GPU compute capacity from the investing cloud provider.
3. Revenue Recognition: The cloud provider recognizes the startup’s spending as revenue, effectively recycling its own investment back into its top-line figures.

This circular flow of cash raises questions about the genuine, organic demand for AI compute power. If a substantial portion of a cloud provider’s AI revenue relies on funding its own customers, the revenue quality is perceived as lower and less sustainable than revenue generated from customers using their own external capital.

Market Implications of Circular Deals

Analysts are concerned that these circular deals mask the true utilization rates and profitability of the newly deployed GPU fleets. If the startups fail or the AI boom slows, the cloud providers are left with billions of dollars in rapidly depreciating hardware that is not generating sufficient external revenue.

• Risk to Cloud Providers: Companies like Microsoft Azure and Google Cloud face the twin risks of accelerated depreciation hitting earnings and potential write-downs if the demand for their funded AI capacity fails to materialize long-term.
• Risk to Chipmakers: While Nvidia is insulated from the depreciation expense itself (as they sell the chips), any perceived slowdown in organic demand from their largest customers (the cloud providers) could lead to a sharp correction in their stock price, which is priced for near-perfect growth.

The Broader Impact on the Technology Sector

The market’s sudden focus on depreciation and circularity signals a shift from a purely growth-driven narrative to one prioritizing earnings quality and return on invested capital (ROIC). This scrutiny is affecting the entire ecosystem built around the AI thesis.

Precedent and Historical Context

This situation echoes the dot-com bubble era, where telecom companies engaged in massive, often unnecessary, CapEx to lay fiber optic cables. When the expected demand failed to materialize, the resulting depreciation and asset write-downs led to widespread bankruptcies and a prolonged market slump in that sector. While the current AI demand is far more tangible, the financial discipline required remains the same.

The GAAP vs. Cash Flow Debate

Cloud providers often argue that the depreciation hit is merely an accounting adjustment and that their underlying cash flow remains strong. They point to metrics like EBITDA (Earnings Before Interest, Taxes, Depreciation, and Amortization) to show operational health.

However, investors are increasingly skeptical of relying solely on EBITDA when the depreciation expense represents a very real, continuous need for replacement CapEx. The cash flow generated must be sufficient not only to cover operating costs but also to fund the next generation of hardware required to stay competitive—the very definition of the depreciation cost.

Key Takeaways for Investors

The market is recalibrating its expectations for AI infrastructure companies, moving away from pure revenue growth at any cost toward sustainable profitability. For investors, the following points are crucial:

• Focus on True Earnings: Look beyond EBITDA and scrutinize net income and free cash flow, factoring in realistic depreciation schedules for AI hardware.
• Analyze Revenue Quality: Demand transparency regarding the source of cloud revenue. Revenue derived from external, non-funded customers is inherently more valuable than revenue generated through circular investment schemes.
• Monitor Obsolescence: Keep a close eye on the pace of chip innovation. Each new major GPU release (e.g., from H100 to B200) accelerates the obsolescence of the previous generation, shortening the effective useful life of those assets.
• Valuation Risk: Companies with high P/E ratios that rely heavily on the AI narrative are most vulnerable to downward revisions as depreciation expenses weigh on reported earnings.

Conclusion: The Path to Sustainable AI Profitability

The emergence of depreciation as the stock market’s “most-hated word” is not a sign that the AI revolution is ending, but rather that it is maturing. The market is transitioning from the initial euphoric phase of massive investment to the disciplined phase of assessing returns on that investment.

For the AI trade to maintain its momentum, Big Tech companies must demonstrate that the revenue generated by their GPU fleets is sufficient to cover the accelerated depreciation costs and still deliver attractive net profits. Until then, the shadow cast by the short useful life of cutting-edge AI hardware will continue to dampen investor enthusiasm and pressure valuations across the sector.

What’s Next

Expect increased pressure on cloud providers throughout late 2025 and 2026 to provide more granular detail in their financial reporting regarding the depreciation schedules used for AI-specific assets. Regulators and accounting bodies may also step in to provide clearer guidance on the useful life of rapidly evolving digital infrastructure, potentially standardizing shorter depreciation periods across the industry.