Has the AI Bubble Burst?

As an Amazon Associate we earn from qualifying purchases.

Key Takeaways

• Recent company results combine strong AI-related demand with substantial spending and cash-flow pressure.
• Falling share prices and missed expectations do not establish that demand for AI infrastructure has collapsed.
• Revenue, contracts, capital spending, and inventories measure different parts of the investment cycle.

The AI Bubble Debate Meets Company Results

Amazon reported $46.0 billion in quarterly cloud sales on October 29, 2026, alongside a $42.4 billion free-cash-flow outflow over the preceding 12 months. Those figures capture the tension behind the AI bubble debate: customers are spending more on computing services, but supplying that demand is absorbing enormous amounts of capital. Strong sales and financial strain can appear in the same business at the same time.

An October 30 Seeking Alpha commentary by Damir Tokic presents a more decisive interpretation. Its public summary argues that capital expenditure disappointments, rising Caterpillar inventories, and ServiceNow’s revenue outlook confirm a bursting artificial intelligence (AI) infrastructure bubble. It also points to large share-price declines as evidence that the market adjustment is underway.

The underlying company disclosures support a more differentiated assessment. Alphabet reported sharply higher cloud revenue and increased spending on property and equipment. Caterpillar reported rising inventories, but also higher sales and expanding operating margins. ServiceNow reported subscription growth and raised its annual AI contract-value target. Amazon combined accelerating cloud revenue with a substantial cash deficit after infrastructure investment.

These findings concern different economic questions. Revenue describes purchases already recognized in a company’s accounts. Capital expenditure represents investment in assets expected to support future activity. Inventories record goods and materials held before sale or use. Share prices reflect expectations about future returns, including expectations that may have become difficult to satisfy.

The distinction matters because “bubble burst” can describe falling valuations without describing collapsing demand. A stock can lose value when a profitable company grows less quickly than investors expected. Infrastructure can also remain heavily used even after its owners discover that construction costs or financing expenses leave inadequate returns.

The evidence does establish financial pressure. It does not establish a uniform contraction across the businesses named in the argument. A useful interpretation separates the repricing of expectations from changes in customer spending, and separates both from the eventual profitability of the assets being built.

Alphabet’s Spending Increased Even as Expectations Became Harder to Meet

Alphabet’s quarterly earnings release reported $127.6 billion in revenue for the quarter ended September 30, 2026, an increase of 25% from the comparable 2025 period. Google Cloud revenue reached $30.7 billion, up 103%. The company attributed cloud growth to enterprise AI infrastructure and applications, together with its core cloud services and sales of its specialized computing systems.

Those results describe substantial commercial activity. They also show why revenue growth alone cannot settle the investment question: the company is spending heavily to support that activity, and the cash consequences differ from its operating-profit results.

Purchases of property and equipment reached $51.355 billion during the quarter. That compares with $44.924 billion in the preceding quarter and $23.953 billion in the comparable 2025 quarter. Spending increased both sequentially and year over year.

A claim that expenditure fell short of market expectations describes a different comparison. Analysts may have expected an even larger number, but an expectations shortfall cannot be substituted for an actual spending decline. The reported figures show that Alphabet continued expanding its investment at a substantial pace.

The cash-flow statement provides the sharper financial concern. Alphabet generated $42.290 billion in operating cash during the quarter, less than its spending on property and equipment. Its reported quarterly free cash flow was negative $9.065 billion, following negative $5.855 billion in the preceding quarter. Free cash flow measures the cash remaining after the capital expenditure included in the company’s definition.

That does not mean Alphabet’s operating businesses stopped producing cash. It means investment exceeded the cash generated by those operations during those periods. The distinction changes the interpretation from business collapse to the financing and timing of an exceptionally large expansion.

Profitability also depends on which part of the accounts is being considered. Google Cloud reported $11.121 billion in operating income, compared with $3.594 billion a year earlier. Cloud operating income demonstrates that the segment produces an accounting profit, but it does not show that every new facility will earn an adequate return over its working life.

Buildings and computing equipment enter the accounts differently from routine operating expenses. Their cost is generally allocated over time through depreciation, rather than charged entirely against profit when purchased. Cash leaves the business before that full expense appears in earnings.

Alphabet’s results consequently support two observations at once. Demand and operating earnings expanded, and infrastructure spending placed pressure on near-term cash generation. Neither observation cancels the other, and neither converts an analyst expectations miss into evidence that the company reduced its physical investment.

Caterpillar’s Inventory Increase Needs an Industrial Explanation

Caterpillar’s September balance sheet reported inventories of $22.214 billion, compared with $18.135 billion at December 31, 2025. That represents an increase of approximately 22.5%. The comparison covers nine months, rather than a year-over-year comparison between two September quarter ends.

Inventory growth can indicate that production is running ahead of sales. It can also reflect materials purchased for future manufacturing, equipment moving through production, or goods awaiting delivery. The aggregate balance does not identify which explanation accounts for the increase.

Caterpillar also serves markets extending beyond data centers. Construction equipment and mining machinery contribute to its business, as do engines and power systems. A companywide inventory increase cannot automatically be assigned to AI-related equipment or interpreted as unsold data-center power capacity.

The operating figures provide additional information. Caterpillar’s quarterly results announcement reported $20.7 billion in sales and revenues, up 17% from the comparable 2025 quarter. Its operating margin increased to 20.4%, compared with 17.3% a year earlier.

Within Power & Energy, total segment sales rose 18%. Power-generation sales increased 30%, reaching $3.421 billion. That category is relevant to electricity demand, but it is broader than AI data centers and cannot be treated as an isolated measure of spending on AI infrastructure.

The inventory increase still has financial consequences. Money committed to materials and equipment remains unavailable for other purposes until the company sells the products or otherwise converts that investment into cash. A growing inventory balance can increase financing needs even when sales are expanding.

The composition of inventory also affects risk. Materials that can serve several products differ from completed equipment built for a particular order. Goods with established customers differ from stock manufactured in anticipation of demand that has not yet become contractual.

Caterpillar’s distribution model adds another distinction. The company explains on its quarterly results page that sales to independent dealers and sales by those dealers to end users occur at different times. Manufacturer inventory, dealer inventory, and equipment already delivered to customers describe separate stages of the commercial process.

The reported inventory build supports concern about working capital, meaning funds tied up in day-to-day operations. The accompanying sales growth prevents that concern from becoming automatic proof of a demand collapse. Interpreting the increase as overproduction specifically caused by weakening AI expenditure goes beyond what the consolidated balance establishes.

ServiceNow Shows the Difference Between Contracts and Revenue

ServiceNow’s October earnings announcement reported $3.988 billion in subscription revenue for the September quarter, up 21% year over year. Total revenue reached $4.106 billion, an increase of 20.5%. The company also raised its year-end AI annual contract value target to $1.7 billion.

Annual contract value (ACV) expresses the annualized value associated with contracts under the company’s reporting approach. It is not interchangeable with revenue already recognized during a quarter. A contract can support future business without immediately producing an equivalent amount of reported revenue.

That timing difference is central to subscription software. Customers enter agreements covering service over a period, and revenue is recognized as the service is provided. Contract activity can strengthen before its full effect appears in quarterly sales.

ServiceNow also reported current remaining performance obligations (cRPO) of $13.65 billion, up 20.5% year over year. This measure represents contracted revenue expected to be recognized over the next 12 months. It describes business already under contract, rather than all sales the company might secure during that period.

The company’s fourth-quarter guidance included cRPO growth of 18.5%. That was below the growth rate reported for the September quarter. A slower expected rate can reasonably concern investors who paid for continuing acceleration, even though the forecast still describes expansion.

Subscription revenue guidance pointed in a different direction. ServiceNow projected fourth-quarter subscription revenue of $4.245 billion to $4.255 billion, representing reported growth of 22.5% to 23%. Calling the overall outlook weak without identifying the metric and comparison would obscure this difference.

Currency adjustments create another source of apparent inconsistency. Reported growth reflects changes in exchange rates when international revenue is translated into dollars. Constant-currency growth attempts to remove those effects. Comparisons become misleading when an actual reported growth rate is placed beside an adjusted forecast without explanation.

AI adoption and AI monetization also measure different activities. Adoption concerns customers using a product or capability. Monetization concerns the revenue and profit generated from that use. Increased usage can precede higher revenue, but it can also increase computing costs or remain included within an existing subscription.

ServiceNow’s figures show a software business still expanding, with an increased AI contract-value target and a slower expected growth rate in one forward-looking contract measure. They support analysis of the pace and economics of monetization. They do not establish that enterprise customers have broadly stopped purchasing AI-related software.

Amazon Combines Cloud Expansion With a Large Cash Deficit

Amazon’s third-quarter results reported that Amazon Web Services (AWS) revenue increased approximately 39% year over year to $46.0 billion. AWS operating income reached $17.8 billion, compared with $11.4 billion in the comparable 2025 quarter. These figures describe both higher sales and higher operating profit in its cloud segment.

The financing picture is less comfortable. Amazon reported $168.3 billion in operating cash flow over the preceding 12 months, but free cash flow was negative $42.4 billion. The company attributed the deterioration primarily to increased purchases of property and equipment, largely reflecting AI investment.

Amazon also maintained its expectation of approximately $220 billion in cash capital expenditure for 2026. That is a company forecast, rather than a completed full-year spending total. It indicates a continuing expansion program, not an announced retreat from infrastructure investment.

AWS revenue includes cloud services beyond AI. Its total cannot be labeled AI revenue, and its operating income cannot be treated as profit generated exclusively by AI products. Equally, consolidated Amazon capital expenditure covers a business with activities beyond cloud computing.

These boundaries matter when judging payback. Dividing companywide expenditure by an assumed AI revenue figure would mix categories that do not share a consistent definition. An estimate produced that way could look precise without describing the return on any identifiable investment.

The cash deficit remains meaningful without that calculation. Amazon invested more in the property and equipment covered by its free-cash-flow measure than it generated in operating cash over the same period. Continued spending at that relationship would require existing cash resources, financing, or changes elsewhere in the business.

A profitable cloud segment can coexist with this deficit because capital investment and operating profit follow different accounting schedules. Infrastructure can require payment before it produces revenue. Depreciation then spreads the expense over the estimated useful life of the assets.

The economic question is whether those assets generate sufficient future cash after electricity, maintenance, and replacement costs. Current growth provides evidence of customer demand, but it does not guarantee that every dollar invested will receive an acceptable return.

Amazon’s release also discussed progress in its satellite communications business. That illustrates another reason to preserve segment boundaries: a diversified company can fund cloud infrastructure and space systems during the same period. The financial performance of one activity does not automatically establish the commercial prospects of the other.

Cash Generation and Valuation Follow Different Timetables

A company’s share price can fall before its revenue does. Investors price expected future cash flows, so a reduction in expected growth can change valuation even when current sales continue rising. Higher financing costs can produce a similar result by reducing the present value assigned to future earnings.

That is why a market correction and an operating contraction should be described separately. A correction concerns the price investors will pay for ownership. An operating contraction concerns declining activity within the business. Either can occur without the other, although financial stress can eventually connect them.

For Alphabet and Amazon, recent results show that infrastructure investment is absorbing cash faster than some operating measures suggest. For ServiceNow, the more relevant issue is how contract growth becomes recognized revenue and profitable service delivery. Caterpillar’s exposure passes through industrial orders and the movement of goods through production and distribution.

The following comparison identifies the financial distinction illustrated by each company. It does not treat their business models or accounting measures as interchangeable.

The timing of asset replacement adds another complication. Computing equipment can remain functional after newer equipment offers better economics. The owner may face pressure to replace usable systems because customers demand lower prices or improved performance.

An accounting useful life is an estimate used to allocate expense. It does not guarantee that an asset will remain commercially competitive for that entire period. Conversely, the introduction of newer hardware does not automatically make all existing equipment worthless.

The distinction between useful infrastructure and attractive investment returns extends to data-center valuation debates. A project can serve paying customers and still disappoint the investors who financed it at an excessive price. The relevant outcome depends on the purchase price and the cash eventually available to its owners.

Financing Relationships Can Connect Otherwise Separate Risks

Revenue recorded by one company can represent expenditure by another company in the same investment chain. Caterpillar sells equipment into industrial projects. Cloud providers purchase infrastructure and sell computing services. Software businesses purchase computing capacity and charge customers for applications.

Adding revenue from every stage would count some economic activity more than once. It would not measure the amount ultimately paid by businesses and consumers outside the supply chain. This distinction becomes important when comparing infrastructure investment with claims about the size of the AI market.

Financing can create further connections. An equipment supplier may also hold an ownership stake in a customer. A cloud provider can invest in a software company that purchases its services. These arrangements may support deployment, but they also connect commercial exposure with investment exposure.

New Space Economy’s coverage of circular AI financing examines this overlap between supplier, customer, and investor relationships. The economic concern is not that every linked transaction is invalid. It is that a customer’s ability to continue purchasing may depend partly on continued access to capital from businesses that benefit from those purchases.

Contractual revenue and financing proceeds remain different things. A customer can satisfy its obligations using money raised from investors, but that does not establish that its own operations generate enough cash to support the same spending indefinitely. The financing source affects how long the arrangement can continue before operating results must carry more of the burden.

Equity financing and borrowing also distribute risk differently. Equity investors accept uncertain returns and possible losses in ownership value. Borrowing generally creates scheduled obligations that remain payable even when revenue arrives later than expected.

This helps explain why the same change in sentiment can affect businesses differently. A company with established operating cash flow has different options from a developer whose main assets are still under construction. A canceled funding transaction can delay a project without indicating that all customers have stopped wanting the service.

The publication’s discussion of Firmus’s withdrawn listing provides context for this distinction between demand for computing and demand for shares at a proposed valuation. Investors can reject an offering price without rejecting the usefulness of the underlying technology.

Physical Constraints Change the Meaning of Slower Spending

A data center does not produce commercial computing capacity simply because its equipment has been ordered. Electricity must reach the site, cooling systems must operate, and the installed hardware must be connected and commissioned. Spending and service availability can move on different schedules.

This creates ambiguity around a quarterly expenditure shortfall. Lower spending than expected can reflect weaker demand, but it can also reflect delayed equipment delivery or construction timing. A financial total records when expenditure occurs; it does not, by itself, explain why a payment moved between periods.

The same distinction applies to inventory. Equipment waiting for a site to become ready is economically different from equipment manufactured without a customer. Both can appear as goods awaiting deployment somewhere in the supply chain, but their prospects for conversion into revenue differ.

New Space Economy’s coverage of data-center infrastructure constraints connects financial performance with electricity access and construction execution. Those physical dependencies explain how a project can encounter revenue delays even when potential customers still need computing capacity.

Constraints can also change pricing. A shortage of usable capacity may support the prices charged by facilities already operating. The same shortage can raise costs for developers trying to build new capacity. Suppliers and buyers experience the same bottleneck from different commercial positions.

Relocating computing changes the cost structure rather than removing it. Proposals for orbital data centers extend this discussion into the space economy. Such concepts exchange terrestrial site dependencies for requirements involving launch and spacecraft operation.

An orbital computing service would still need paying customers and sufficient revenue to recover its costs. Launch expenditure and hardware replacement would affect those economics, as would the system’s ability to transfer data. A shortage of terrestrial electricity does not independently establish that space-based computing can provide a cheaper service.

These relationships also run in the opposite direction. More efficient or less expensive terrestrial computing could affect the prices available to future orbital providers. The financial prospects of space infrastructure depend partly on competition from the ground, not simply on whether AI usage continues increasing.

The Evidence Supports Financial Pressure Without a Uniform Demand Collapse

The company results describe expanding businesses facing different financial pressures. Alphabet increased capital expenditure and reported negative quarterly free cash flow. Amazon reported strong cloud growth alongside a substantial trailing cash deficit after investment. Caterpillar carried more inventory, and ServiceNow forecast slower growth in one measure of contracted future revenue.

None of those developments is trivial. Together, they show that revenue expansion does not eliminate the costs of financing, construction, and service delivery. They also show why a broad AI bubble label can conceal differences that materially affect the interpretation.

A market valuation can depend on several favorable assumptions holding simultaneously. Revenue may need to grow quickly, prices may need to remain attractive, and infrastructure may need to enter service on schedule. A modest change in one assumption can produce a large valuation adjustment when much of the expected value lies years ahead.

That adjustment does not require revenue to turn negative or customers to abandon the technology. It requires investors to assign a lower value to future cash flows. The business can continue growing from a lower valuation, potentially with a different financing plan.

The reverse is also possible. Strong current revenue can coexist with poor investment decisions whose effects become visible later. Long-lived commitments can outlast the conditions that originally justified them, and profitable existing operations can temporarily absorb losses from new projects.

The available disclosures support neither a claim that every AI investment is economically justified nor a claim that the named companies confirm a general collapse in demand. They show substantial demand alongside increasingly visible differences between operating growth and cash available after investment.

The strongest finding concerns the terms of the debate. Capital expenditure below expectations is not the same as falling capital expenditure. Inventory growth is not automatically excess AI equipment. Contract growth is not recognized revenue, and a falling share price is not a direct measure of customer usage.

Describing those distinctions accurately produces a narrower but more defensible assessment: the economics of AI expansion are under pressure, and some valuations may depend on outcomes that remain uncertain. The reported operating results still show significant commercial growth.

Summary

The AI bubble debate concerns who receives the financial benefits of the technology as much as whether the technology is useful. Customers can gain from lower computing prices or more capable software even when infrastructure owners receive disappointing returns. Competition can transfer value away from producers and toward users.

Alphabet, Caterpillar, ServiceNow, and Amazon illustrate separate parts of that process. Their results show growing activity, substantial investment, and financial tradeoffs that differ by business model. They do not provide a single, consistent measure confirming that an industrywide bubble has burst.

The distinction leaves room for an outcome that simple boom-or-collapse descriptions miss. AI services can become more widely used as some suppliers face lower margins, some projects are delayed, and some investors accept losses. Commercial adoption and investment performance do not have to move together.

Appendix: Useful Books Available on Amazon

• Irrational Exuberance
• Prediction Machines: The Simple Economics of Artificial Intelligence
• Power and Prediction: The Disruptive Economics of Artificial Intelligence
• Chip War: The Fight for the World’s Most Critical Technology
• Technological Revolutions and Financial Capital: The Dynamics of Bubbles and Golden Ages

Appendix: Top Questions Answered in This Article

Have the latest company results confirmed that the AI bubble has burst?

The results do not establish a uniform collapse in AI demand. Alphabet and Amazon reported strong cloud growth, Caterpillar reported higher sales, and ServiceNow reported growing subscription revenue. The disclosures also show financial pressure, including heavy infrastructure spending and cash deficits after investment, which can support lower valuations without proving an industrywide contraction.

Did Alphabet reduce its capital expenditure during the September 2026 quarter?

Alphabet reported $51.355 billion in purchases of property and equipment during the quarter. That exceeded both the preceding quarter’s $44.924 billion and the comparable 2025 quarter’s $23.953 billion. Spending below an analyst forecast would represent an expectations shortfall, not an actual reduction in expenditure.

Why can a profitable company report negative free cash flow?

Operating profit and free cash flow measure different aspects of a business. A company can earn an accounting profit but spend more cash on new infrastructure than its operations generate. Capital assets are generally expensed over time through depreciation, even though payment can occur much earlier.

Does Caterpillar’s inventory increase prove that AI equipment demand is weakening?

Caterpillar’s inventory increase does not isolate AI-related equipment or establish its cause. The company serves construction and mining markets as well as power applications. Its September 2026 inventory balance rose approximately 22.5% from December 2025, but quarterly sales and revenues also increased from the comparable 2025 period.

What does ServiceNow’s annual contract value measure?

Annual contract value expresses the annualized value associated with contracts under the company’s reporting approach. It differs from revenue recognized during a particular quarter. ServiceNow’s increased AI contract-value target describes expected commercial commitments, rather than an equivalent amount of AI revenue already recorded in its financial statements.

Why does Amazon’s negative free cash flow matter despite AWS growth?

Amazon’s cash deficit shows that investment exceeded operating cash generation under its reported free-cash-flow measure. AWS revenue and operating income increased, but those gains did not eliminate the cash demands of infrastructure expansion. Continued deficits would require existing cash resources, financing, or changes in investment and operating performance.

Can AI adoption increase during a stock-market decline?

AI adoption can increase even when related stocks fall. Share prices reflect expectations about future profits and the price investors will pay for those profits. Customers may use more AI services at the same time that competition, higher costs, or excessive starting valuations reduce shareholder returns.

What is circular financing in the AI industry?

Circular financing describes arrangements in which businesses have overlapping relationships as investors, suppliers, and customers. These relationships can help fund infrastructure and support purchases. They can also concentrate risk when a customer’s spending depends on continued funding from companies that benefit from selling to that customer.

Do electricity constraints always indicate weak data-center demand?

Electricity constraints concern the ability to bring infrastructure into operation, rather than necessarily indicating weak customer demand. A project can face delayed revenue because its power connection or construction schedule has slipped. The associated spending and inventory patterns can resemble demand weakness unless the physical circumstances are considered.

Would an AI investment correction automatically eliminate orbital data centers?

An investment correction would not automatically eliminate orbital computing projects, but it could affect financing and expected selling prices. Space-based providers would still need to recover launch, spacecraft, and operating costs. Their commercial prospects would depend on customer demand and competition from terrestrial computing services, rather than AI eniasm alone.

Appendix: Glossary of Key Terms

Free Cash Flow

Cash remaining after the investment spending included in a company’s stated calculation. Companies may define this measure differently. Negative free cash flow means the relevant investment outflows exceeded operating cash generation during the measured period, rather than necessarily indicating an operating loss.

Artificial Intelligence

Computer systems designed to perform tasks involving capabilities such as prediction or language processing. In this discussion, the term includes the software services customers purchase and the computing infrastructure used to develop and operate those services.

Capital Expenditure

Spending on assets expected to serve a business beyond the current accounting period. Data-center buildings and computing equipment are examples. Capital expenditure affects cash when payment occurs, but its associated accounting expense is generally allocated over time.

Inventory

Materials, unfinished products, and completed goods held by a business before sale or use. A rising balance can reflect production plans or delivery delays. It can also indicate weak sales, so the balance alone does not establish the cause.

Operating Margin

Operating income expressed as a percentage of revenue. It indicates how much revenue remains after the operating expenses included in the calculation. It does not directly measure cash remaining after capital expenditure or the eventual return on newly constructed infrastructure.

Depreciation

The accounting allocation of a physical asset’s cost over its estimated useful life. Depreciation affects reported profit without necessarily matching cash payments during the same period. The estimated accounting life can differ from the period during which the asset remains commercially competitive.

Working Capital

Funds associated with a business’s short-term operating assets and liabilities. Inventory and customer receivables can tie up cash before it is collected. Supplier payment terms and customer advances can reduce the amount a company must finance itself.

Annual Contract Value

An annualized measure of contractual business, calculated according to the reporting company’s methodology. It helps describe the scale of customer agreements. It should not be treated as identical to cash collected or revenue recognized during a particular quarter.

Current Remaining Performance Obligations

Contracted revenue that a company expects to recognize within the next 12 months. This measure describes existing contractual business. It excludes potential future agreements that have not yet been signed and is different from revenue already recognized.

Constant-Currency Growth

A calculation intended to show business growth without the effect of changing exchange rates. Companies translate comparable financial results using a consistent currency basis. The resulting growth rate can differ from the rate reported in their financial statements.

Circular Financing

Financing relationships in which companies also transact with one another as customers or suppliers. Such arrangements can support business expansion, but they can connect investment risk with commercial risk when spending depends on continued access to related financing.