top of page

The market financing its own optimism

2 minutes ago
6 min read

By Matthew Parish


Sunday 22 November 2026


Anthropic’s turbulent debut and OpenAI’s postponed listing raise an uncomfortable question: how much of the artificial intelligence boom can survive without another round of financing?


Financial bubbles rarely begin with something entirely imaginary. More often they begin with a genuine innovation, followed by an exaggerated estimate of how quickly it will make everybody rich. Artificial intelligence may transform the economy. That does not establish that the prices attached to its leading companies are sensible, or that the financial arrangements supporting their expansion are sustainable.


The distinction has become rather urgent. Anthropic’s arrival on the stock market was supposed to demonstrate that investors remained willing to finance the next stage of the artificial intelligence revolution. Its reception instead revealed a market capable of enthusiasm and alarm almost simultaneously.


On Thursday 19 November, Anthropic shares finished their first trading day approximately 17 per cent above their offer price. Yet the wider market fell sharply: the Dow lost around 1,100 points, while the S&P 500 surrendered an earlier gain. By midday on Friday, Anthropic itself was down roughly 13.5 per cent, trading just above $84 and approaching its $83 flotation price. The Friday figure was an intraday reading, but the reversal was already striking.


By Friday’s close, the damage extended well beyond one newly listed company. The S&P 500 had fallen 3.7 per cent that day, leaving it more than 10 per cent below its 6 October record. The Nasdaq Composite had lost its gains for the year. A correction was therefore already under way. Whether it becomes a crash remains an open question.


It would be too simple to say that Anthropic caused these falls. A flotation takes place within a market already responding to earnings, interest rates, politics and changing expectations. Nevertheless, its debut supplied a useful test. Investors could welcome a particular share offering while becoming less willing to accept the valuations and spending assumptions of the industry surrounding it.


The central difficulty lies in the way parts of that industry finance themselves.


Consider a simplified arrangement. A chip manufacturer invests in an artificial intelligence developer. The developer uses its new capital to purchase computing capacity from a cloud provider. The cloud provider buys the manufacturer’s chips. The manufacturer records sales, the developer obtains infrastructure, and the cloud provider announces a substantial customer commitment.


There may be real equipment, real services and properly recorded revenue at every stage. Circular financing is not, by itself, evidence of fraud. Suppliers have good reasons to help promising customers expand. But the arrangement creates a question that impressive revenue figures can obscure: how much demand comes from customers able to pay out of their own earnings, and how much depends upon money supplied by the sellers themselves?


The Congressional Research Service describes an instructive example. Amazon invested $5 billion in Anthropic in April, with provision for further investment, while Anthropic committed to spending $100 billion on Amazon chips and cloud services over ten years. The amounts and timescales differ; this is not a simple exchange of identical sums. It nevertheless illustrates how investment relationships and commercial demand can become intertwined.


Such arrangements can accelerate the development of a viable industry. They can also postpone the discovery that an industry has built more capacity than its independent customers will profitably support.


A supplier investing in its customer acquires two exposures to the same commercial proposition. If that customer succeeds, the supplier benefits from sales and from its investment. If the customer struggles, orders may weaken just as the investment loses value. Add lenders, long leases and guarantees, and the consequences can spread beyond the original transaction.


The European Central Bank has drawn attention to precisely this combination of debt, opaque commitments and circular relationships in artificial intelligence financing. It also highlights the danger that computing assets may have a shorter useful economic life than those lending against them expect. Equipment can remain physically operational while becoming commercially unattractive beside a newer generation of machines.


This is where a disappointing investment cycle could become a financial crisis. Shareholders can absorb losses, however painfully, without every failed prediction becoming a default. Debt imposes a timetable. Interest and principal fall due whether the promised productivity revolution arrives next quarter or next decade.


A plausible crash would therefore develop through several connected disappointments. Paying customers might adopt artificial intelligence more slowly than expected, or demand lower prices. Developers would then face a widening gap between receipts and computing commitments. New investors might insist on lower valuations. Infrastructure providers could encounter postponed orders or renegotiated contracts. Creditors might become reluctant to refinance them.


Each individual decision could be prudent. Taken together, those decisions could withdraw the financing on which the whole expansion depends.


The stock market would amplify that process if falling share prices made fresh capital harder to obtain. Expectations of growth support valuations; valuations help companies raise money; that money finances spending which appears to confirm the original expectations. When confidence weakens, the same relationship can work backwards.


OpenAI’s reluctance to proceed immediately with a flotation deserves attention in this context, although its stated explanation must be treated accurately. In September, Sam Altman ruled out a 2026 listing, citing artificial intelligence safety concerns.


That is not evidence that management privately expects a stock market collapse. Nor does postponement establish that a company cannot withstand public scrutiny. A decision about listing can reflect governance, product development, regulatory uncertainty and access to private capital, as well as valuation.


Nevertheless, the financial implications are significant. Remaining private avoids an immediate, continuously visible market verdict. Anthropic now receives that verdict every trading day. Its performance offers investors a reference point against which to assess other developers, including OpenAI.


Waiting also carries a risk. If public investors become more sceptical of artificial intelligence, a later flotation could take place on less favourable terms. Conversely, a company that uses the interval to demonstrate durable customer demand and improving cash generation may emerge stronger. Delay buys time; it does not determine what that time will be worth.


There remains a substantial argument against assuming catastrophe. Artificial intelligence can generate useful products and measurable savings. Strategic investors may understand their customers better than outside lenders do. Infrastructure built during an investment boom can find productive uses even after the original financial projections prove extravagant.


Nor does every company described as a technology stock share the same financial condition. A profitable, diversified business funding investment from existing operations differs considerably from a developer dependent upon repeated fundraising. Treating them as one homogeneous bubble conceals the distinctions that will determine where losses fall.


The historical warning is subtler than the claim that new technology is worthless. Railways and telecommunications networks changed economic life while inflicting severe losses upon some of the people who financed them. A technology can fulfil its social promise without fulfilling the promises embedded in its share price.


For Ukraine, these distinctions have practical importance. A sustained international retreat from risk could make private reconstruction investment harder to secure and encourage foreign customers to defer technology spending. Financial pressure on partner countries could also complicate political arguments about overseas commitments. These are possible transmission channels, rather than automatic consequences of a bad week on Wall Street.


The appropriate response is therefore close attention to evidence. Are receipts from independent customers catching up with infrastructure expenditure? Are operating improvements translating into cash? Can borrowers meet obligations without repeated refinancing? Are guarantees and purchase commitments sufficiently transparent for investors to understand who ultimately bears the risk?


A recovery in share prices would not, by itself, answer those questions. Neither does this week’s fall prove that the most alarming answers are correct.


Anthropic’s turbulent debut has exposed the limits of a reassuring narrative in which each new financing round validates the last. OpenAI’s decision to wait leaves another important valuation test unresolved. Between them stands an industry attempting to finance an extraordinary expansion before the ultimate returns are clear.


A crash is possible, not inevitable. The danger is that too many participants have mistaken one another’s willingness to invest for independent confirmation that the investment will pay. Eventually, the money circulating amongst suppliers, developers and financiers must be replenished by customers earning enough value from the technology to sustain it. That is the test no flotation can evade.

 
 

Note from Matthew Parish, Editor-in-Chief. The Lviv Herald is a unique and independent source of analytical journalism about the war in Ukraine and its aftermath, and all the geopolitical and diplomatic consequences of the war as well as the tremendous advances in military technology the war has yielded. To achieve this independence, we rely exclusively on donations. Please donate if you can, with the buttons at the top of this page.

Copyright (c) Lviv Herald 2024-26. All rights reserved.  Accredited by the Armed Forces of Ukraine after approval by the State Security Service of Ukraine. To view our policy on the anonymity of authors, please click the "About" page.

bottom of page