TL;DR
A Noahpinion report says major cloud and technology companies are spending more on AI infrastructure and increasingly borrowing to fund the buildout. Revenue growth from AI coding tools has so far helped offset concern about the spending, but the size of the investment and the consequences of any slowdown remain uncertain.
Major technology companies are accelerating AI data-center spending and increasingly turning to borrowing to fund it, while revenue from AI products has also grown, according to a recent Noahpinion analysis. The combination has revived comparisons with the railroad investment boom and its 1873 financial crash, though the report says a near-term AI bust is not clearly underway.
The report describes a change from the earlier phase of the AI buildout, when hyperscalers such as Amazon, Google, Meta and Oracle mainly used cash generated by their existing businesses to pay for infrastructure. It says spending has since grown enough that some companies are borrowing to continue expanding. For all the hyperscalers other than Microsoft, the report says free cash flow either recently turned negative or is expected to do so later in 2026. The report does not provide a company-by-company breakdown in the supplied material.
At the same time, revenue growth has accelerated. Noahpinion credits the release of Claude Code, Codex and other agentic coding tools with making software development a strong commercial use for AI, generating substantial revenue for Anthropic and, to a lesser extent, OpenAI. That increase has eased immediate concerns that infrastructure spending is simply outrunning sales, but it does not remove the risks linked to rising borrowing.
The analysis argues that debt changes the potential consequences of a slowdown. When infrastructure is funded from cash, disappointing AI revenue can hurt share prices and profitability; with more debt, companies also need enough future income to meet repayment obligations. The report says a sharp slowdown could create a risk of default, but it does not claim that any company is currently unable to pay its debts.
Borrowing Raises the Cost of a Slowdown
The issue matters because AI infrastructure now involves large commitments across data centers, chips, power and related construction. If demand and revenue keep growing, borrowing may support continued expansion. If revenue growth falls short of expectations, companies with more debt could face pressure to cut spending, seek new financing or sell assets, with potential effects on suppliers and investors.
The comparison with railroads is a warning about scale and financing, not proof that the same outcome will recur. The 1873 railroad crash was followed by a prolonged U.S. depression, according to the source report. But the industries, financial structures and economic conditions differ, and the available figures do not establish that AI infrastructure is headed for an equivalent collapse.
AI data center infrastructure equipment
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Why the Railroad Comparison Returned
More than a year before this report, Noahpinion discussed whether heavy borrowing to build AI data centers could lead to an investment bust. At that stage, the concern was prospective: spending was rising faster than industry revenue, but hyperscalers were largely financing projects from their own cash. The current analysis says both sides of that equation have changed: spending has accelerated, and so has revenue.
The source also reports disagreement over the size of the AI investment boom. Economist Stijn Van Nieuwerburgh estimates AI investment at 3.63% of GDP, compared with 2.24% for railroads, while other estimates differ substantially. The analysis says some estimates count announced spending, while Van Nieuwerburgh’s method starts with physical construction projects and assumes higher costs. Dylan Patel is cited as estimating more than 6% of GDP when a wider range of AI-related infrastructure is included. The estimates use different methods and scopes, so they are not directly comparable.
Railroad investment figures are also uncertain. The report cites estimates that railroad spending peaked around 5.4% of GDP, while other historical sources put the buildout higher. It cautions that 19th-century records are incomplete and that the railroad boom unfolded over a different period. The comparison is best understood as evidence of the scale of investment, rather than a precise ranking.
As an affiliate, we earn on qualifying purchases.
Investment Estimates Still Diverge
The total cost of the AI buildout remains unsettled. Estimates differ over which spending to count, how to value planned construction and whether to include items such as chip fabrication equipment, semiconductors and energy infrastructure. The source says these methodological differences explain at least part of the gap, but the supplied material does not resolve it.
It is also unclear how quickly AI revenue will continue to grow, how much debt each company is taking on specifically for data centers, and whether future demand will generate enough cash to service that debt. The reported negative or expected-negative free cash flow figures are not accompanied here by detailed company disclosures or a consistent time period. No default or sector-wide financial crisis is established in the report.
enterprise AI development hardware
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Revenue and Debt Will Set the Test
The next test is whether AI revenue can keep pace with infrastructure costs as companies continue building and borrowing. Investors and readers will need to follow company earnings reports, cash-flow updates, debt disclosures and spending plans to see whether the trend described by Noahpinion holds across the hyperscalers.
Further estimates may also clarify the scale of investment, but comparisons will remain limited unless they use similar definitions and time windows. For now, the source’s conclusion is measured: faster revenue growth has reduced immediate bubble concerns, while rising borrowing makes the industry more exposed if that growth slows.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the current concern about AI and the railroads?
The concern is that large AI infrastructure investments, increasingly funded with debt, could become financially risky if revenue does not keep growing. The comparison refers to the railroad investment boom and the crash of 1873; it does not establish that AI will follow the same path.
Are AI companies already facing a financial crisis?
The source report does not say that. It says revenue from AI coding tools has accelerated and that a near-term bust is not clearly underway. It warns that increased borrowing could make a future slowdown more dangerous.
Which AI products are cited as revenue drivers?
The analysis points to Claude Code, Codex and other agentic coding tools, saying they helped accelerate revenue for Anthropic and, to a lesser extent, OpenAI.
How large is AI infrastructure investment compared with railroad spending?
There is no agreed comparison. The source cites estimates ranging from 3.63% of GDP for AI investment to more than 6% when a broader set of infrastructure is counted. Railroad estimates also vary, and the methods and time periods are not directly comparable.
What should readers watch for next?
Company disclosures on cash flow, borrowing, capital spending and AI revenue will help show whether income is keeping pace with infrastructure costs. The source does not specify a single upcoming milestone.
Source: rss
