TL;DR
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
A Kiplinger analysis argues that rising AI-related earnings may be less durable than current stock valuations suggest. It points to hyperscalers’ capital spending, disputed assumptions about hardware depreciation and supplier-linked financing, while noting that the scale of future AI revenue remains uncertain.
A Kiplinger analysis has identified three risks to the durability of earnings supporting the AI infrastructure boom: hyperscalers’ spending and free cash flow, questions about how long companies depreciate AI hardware, and financing links between suppliers and customers. The concerns do not establish that an AI bubble exists, but they challenge the argument that recent earnings growth alone proves current valuations are sustainable.
The analysis distinguishes a potential “earnings bubble” from the price-led bubble associated with the dot-com era. In its account, dot-com share prices rose faster than profits, pushing price-to-earnings ratios sharply higher. The article says current ratios are not near early-2000s levels because earnings have also grown. Its concern is different: if those earnings prove overstated or temporary, profits and share prices could both fall.
One warning sign is the cost of the data-center buildout. The report says large cloud providers, or hyperscalers, have been spending heavily on AI infrastructure and have gone free-cash-flow negative. The spending benefits chipmakers that sell the equipment. If cloud providers do not generate enough AI revenue to support their investments, the report argues, a slowdown in purchases could reduce semiconductor companies’ earnings as well as their stock prices.
A second concern is accounting. The report says hyperscalers have been criticized for extending estimated useful lives for servers and networking equipment from three or four years to five or six. A longer depreciation schedule spreads costs across more years and can reduce annual depreciation expense, raising reported operating profit in the near term. These are allegations about accounting effects, not proof that companies have misreported results. Nvidia has defended the treatment of its GPUs: it said customers depreciate them over four to six years based on real-world longevity and use.
How AI Spending Could Reach Earnings
The issue matters because investors use earnings to judge stock valuations. If reported profits are supported by recurring customer demand, large infrastructure investments may be economically justified. If they instead depend on spending that slows, cost assumptions that understate near-term expenses, or financing that cycles through suppliers and customers, the earnings picture may be less durable than it appears.
The report’s third warning sign is circular financing: a supplier invests in a company that then buys the supplier’s products. It cites Nvidia’s investment in cloud provider Nebius as an example, describing a sequence in which Nvidia takes an equity stake and Nebius uses funding to purchase Nvidia GPUs. The analysis argues that this can make it harder to distinguish sales driven by independent customer demand from sales enabled by supplier capital. The existence of an investment and a customer relationship does not, by itself, show that revenue is improper.
The report also gives a mid-2026 estimate of $25 billion in AI revenue against $660 billion in capital spending, a 4% revenue-to-capex ratio. It presents this as an estimate of AI revenue relative to spending, not profit or a measure of eventual returns. The figure highlights the gap the analysis says investors should watch, but it does not establish that future revenue will fail to catch up.
Products Worth Considering
FlexiSpot E6 Electric Standing Desk, 55 × 28 in, Bamboo Top, White Frame
Logitech MX Mechanical Wireless Illuminated Full-Size Keyboard, Clicky
Logitech MX Master 4 for Mac Wireless Ergonomic Mouse, White Silver
As an affiliate, we earn on qualifying purchases.
The Dot-Com Comparison Has Limits
The debate described by Kiplinger pits investors who see unsustainable AI valuations against those who argue that growing earnings support current prices. The dot-com comparison is common, but the report says the present cycle differs in one important respect: earnings have risen alongside prices rather than lagging them in the same way as during the earlier price bubble.
That distinction does not settle whether AI stocks are fairly valued. The analysis argues that a new technology can be transformative and still be accompanied by a bubble, pointing to railroads and the internet as examples of major innovations associated with speculative cycles. Its central question is whether AI revenue can justify the capital invested, rather than whether the technology has long-term uses.
“Customers depreciate GPUs (graphics processing units) over four to six years based on real-world longevity and utilization patterns.”
— Nvidia, as quoted in the Kiplinger report
The Revenue Gap Is Still Unsettled
The report’s warning signs describe risks and allegations, not a finding that AI earnings are false or that a market bubble is certain. The source does not provide company-by-company cash-flow figures, detailed evidence for the depreciation criticisms, or a full account of the financing arrangements it characterizes as circular. Nvidia’s stated rationale for GPU depreciation is included, but the material does not give responses from every company discussed.
It is also unclear whether the cited mid-2026 revenue and spending estimates will hold, how the estimate defines AI revenue, or how quickly customers may turn infrastructure into paying services. Revenue-to-capex is not profit, and a low ratio at one point does not by itself determine whether investments will earn adequate returns over time. The source does not specify a publication date, so the estimate’s status beyond the stated period cannot be assessed from the material provided.
Watch Revenue and Spending Trends
The key indicators to follow are whether hyperscalers generate AI revenue at a pace that supports their capital spending, whether their free cash flow improves or remains under pressure, and whether infrastructure purchases continue to grow. A sustained gap between spending and monetized demand could increase the risk that investment is scaled back, affecting suppliers’ sales and earnings.
Investors can also track company disclosures about depreciation assumptions and equipment use, alongside information about supplier investments in customers. Those measures may help clarify how much earnings growth comes from customer demand and how much is tied to accounting estimates or financing structures. The available report does not set a date for a resolution: whether AI earnings validate current valuations remains dependent on future results.
Key Questions
Does the analysis say an AI earnings bubble has been proven?
No. It presents three warning signs and argues that earnings may be less durable than some investors expect. It does not establish that a bubble exists or that a downturn is certain.
What are the three warning signs?
The report points to hyperscalers’ heavy capital spending and weak free cash flow, disputed assumptions about the useful lives of servers and networking equipment, and financing links in which suppliers invest in companies that buy their products.
What does the 4% AI revenue-to-capex figure mean?
The report cites a mid-2026 estimate of $25 billion in AI revenue and $660 billion in capital spending, describing the relationship as 4%. It is revenue relative to capital spending, not profit, and it does not forecast the eventual return on that spending.
Why does the report compare AI with the dot-com era?
It uses the comparison to distinguish a price bubble, where stock prices outpace earnings, from a possible earnings bubble, where both profits and prices could fall if earnings expectations prove unsustainable. The report says current price-to-earnings ratios are not at early-2000s levels.
Source: kiplinger.com
