UK Edition

Wednesday, 18 November 2026

Time Trade

Markets, trading & finance — British perspective

Markets

Why Companies Are Staying Private Longer, and What It Means for Investors

Time Trade flows-and-FX tape (2026-11-18): For three years, Anthropic has been one of the most consequential private companies in the American economy. This week, that company is expected to begin… Primary source: original at Investing.com UK Stocks (uk.investing.com).

· Investing.com UK Stocks

For three years, Anthropic has been one of the most consequential private companies in the American economy. This week, that company is expected to begin issuing shares to the public. In preparation for its initial public offering, the company opened its books to the public for the first time on October 26, when it filed a registration statement with the Securities and Exchange Commission. The company expects to begin trading on the Nasdaq under the ticker ANT.

At just under $65 billion, the offering would be the second-largest public offering on record behind SpaceX. Insiders’ remaining holdings are subject to lock-up agreements that release on a staggered schedule after listing, which means additional shares will reach the market at intervals most retail buyers never see coming. And because Nasdaq rewrote its listing rules this spring to let a company this large join the Nasdaq-100 after as few as 15 trading days, most Americans holding a 401(k) or a broad index fund will likely own Anthropic within weeks of its debut, whether or not they ever decide to buy a share.

Anthropic’s filing helps us better understand what Anthropic earns, what it owes, what it sees as its risks, and who controls it. Much of what was revealed confirmed the remarkable growth and size of the company, including revenue of over $33 billion in the first nine months of 2026 relative to $2.5 billion in the same period a year earlier. Some of the information disclosed is sobering, including almost $900 billion in compute and lease commitments, which is over twenty-five times the company’s revenue over those nine months.

The takeaway from this disclosure is that the path forward for Anthropic contains the potential for both tremendous gains and significant risks. This fact sheet focuses on four of them. First, the circular financing arrangements that allow the same dollars to be counted as demand at several stops around the AI industry. Second, the concentration of that industry in a handful of suppliers, one of which lost its chief executive days before this offering. Third, a policy environment in which investor protections are being narrowed rather than strengthened. And fourth, a governance structure that leaves public shareholders with almost no ability to act on what they learn.

The Market Anthropic Is Listing Into

On Thursday, November 5, the Nasdaq-100 fell roughly 6.4%, erasing about $2.3 trillion in market value on a single day. This was reported as the largest one-day dollar loss on record and the worst day for the index since March 2020.

This was not an isolated tremor. Expectations for Anthropic’s valuation fell with it. Potential buyers who had discussed initial valuation figures approaching $2 trillion in early October were now expecting closer to a $1.1 – $1.3 trillion valuation by November 6. The share price range that Anthropic eventually filed on November 10 was $70 – $78 per share.

Although the scale of what is being built has caused concerns over time, what changed this particular week was the political outlook surrounding it. Investors moved out of AI positions following the midterm elections amid fears that the results would slow data center construction and bring tougher scrutiny of new development. The six largest hyperscalers are expected to spend more than $850 billion on AI-related projects in 2026, an increase of more than 80% over 2025. That amount of spending is on the same order as the Interstate Highway System but compressed into a single year and financed substantially with debt. What the November selloff demonstrated is how quickly sentiment toward a build-out of that scale can turn.

The Circular Money Movement

The most important finding in the S-1 is about the structure of Anthropic’s money movement.

In this market, the companies that sell AI infrastructure also invest in the companies that buy it. The customer uses the investment to buy the supplier’s product, and the supplier books it as revenue. Money that never leaves the ecosystem is recorded, at each stop, as demand. The International Monetary Fund flagged this pattern in its October 2026 Global Financial Stability Report, identifying fourteen disclosed deals since 2023 that paired roughly $300 billion in equity and $650 billion in debt with approximately $1.4 trillion in future revenue commitments.

Anthropic’s filings show the companies sitting at the center of this web:

  • Amazon holds convertible notes and non-voting stock in Anthropic, supplies compute, and resells Anthropic’s models.
  • Google holds non-voting stock and supplies compute.
  • SpaceX is contracted for up to $120 billion in compute through 2029, with capacity expected to be fully delivered by December 2026.
  • Five-year master equipment leases with Broadcom carry about $160 billion in rent. This is the largest single supplier commitment in the filing.
  • Nscale, a compute partner, disclosed in its own IPO disclosure that its agreements to sell computing capacity to Anthropic provide for payments of up to approximately $45 billion, while noting it had not secured binding financing to build what it had promised.

The chart below shows what Anthropic has agreed to pay its largest suppliers over the life of each deal. These are the amounts that the company is committed to across the next several years, and it must generate the revenue to cover them. Taken together, they are the source of the pressure on Anthropic to hit revenue targets that its own filing admits it has difficulty forecasting.

On top of that, Anthropic’s November 10 amendment to its IPO documents disclosed a $26 billion concurrent private placement (or raising of private capital alongside the public offering), sold at the IPO price to seven companies that are also its suppliers:

Note that these buyers are getting a class of stock that carries no associated voting rights. Seven of the largest technology companies in the world are pouring $26 billion into Anthropic in exchange for shares with no associated say in governing the company. So instead of buying influence, they are buying a continuation of a customer relationship. And for the investors coming after them, they are setting a price.

That arrangement is only as durable as the people who make it. On November 12—six days before Anthropic is expected to price—Nvidia filed an 8-K announcing that Jensen Huang had resigned as president, chief executive, and board member, effective immediately. The executive who committed Nvidia’s $10 billion to this placement is no longer at the company, and an interim chief executive drawn from the finance organization now holds the pen on ecosystem investments. Nvidia reports earnings on November 17, the day before Anthropic is expected to set its price.

What Anthropic Has Promised To Pay

The company has committed to almost $900 billion in purchase obligations and leases, substantially all of it for compute and infrastructure. Roughly $70 billion falls due within twelve months, and almost $450 billion before the end of 2030, with the remainder running through 2045. These commitments alone are nearly the annualized revenue run-rate of just over $80 billion as of October 31, meaning Anthropic will still be underwater unless demand rises significantly.

One particularly striking feature of Anthropic’s IPO disclosure is that the company’s revenue is largely consumption-based, as customers pay for what they use, and this demand can clearly decline over time. Anthropic’s costs are fixed for years, and in some cases decades, while its revenue depends on continued demand. A drop in usage could put the company in a precarious position. Another feature to focus on is that Anthropic plainly states that it has underestimated its compute needs in the past and has difficulty forecasting revenue.

What the filing does not clearly say is how much of the $895 billion the company is actually obligated to pay. Press readings of the commitments footnote put almost $215 billion as subject to termination rights or to counterparty conditions such as financing, facility readiness or regulatory approval. That leaves the large majority binding. Of the terminable slice, only about $4.3 billion falls within the next twelve months.

The agreements with Microsoft, Google, Amazon and Broadcom—reported at just under $440 billion combined—are described as largely non-cancelable outside a material breach, and several require payment whether or not Anthropic uses the capacity it has reserved. This “take-or-pay” structure converts what would otherwise be a variable cost into a fixed one.

Anthropic’s Disclosed Compute and Lease Commitments  
Against Current Revenue Pace 

Source: Better Markets analysis of Anthropic’s S-1 and Amendments No. 1–2

A Product That Can Break Into Computer Systems

Anthropic’s filing describes a product that has been used as a weapon, and one that has, in testing, broken into systems on its own. The IPO disclosure says that in September 2025 a Chinese state-sponsored group used Claude Code in an attempt to penetrate roughly thirty targets and that some of those intrusions succeeded, in what the filing calls the first documented case of a large-scale cyberattack executed without substantial human intervention. Separately, misconfigured test environments during the company’s own evaluations let its models reach real third-party production systems.

Anthropic deserves credit for disclosing these incidents at all. But these incidents can also lead investors to question whether it can keep its product from being used to attack everyone else’s systems, and what follows in litigation, congressional scrutiny, federal contracting consequences, and customer flight when it cannot.

The Law Has Not Caught Up, and May Be Moving Backward

Beyond these deregulatory pushes at the SEC, there is also no comprehensive federal statute governing AI liability. In its absence, two things are happening at once.

On one hand, states and other jurisdictions are legislating. Eleven states had enacted AI chatbot laws by mid-2026. The European Union’s revised Product Liability Directive expressly covers all types of software, including AI systems, and applies to products placed on the market from December 9, 2026.

The federal government, on the other hand, is pushing the other way. Executive Order 14365, signed December 11, 2025, directs the Attorney General to stand up an AI Litigation Task Force to challenge state AI laws, makes states with “onerous” AI laws ineligible for certain federal broadband funds, and instructs the administration to draft preemptive federal legislation. Those recommendations arrived on March 20, 2026, as A National Policy Framework for Artificial Intelligence, a nonbinding set of recommendations that would preempt state AI laws imposing undue burdens, limit states’ ability to regulate model development, and restrict the liability of AI developers for unlawful conduct by third parties using their systems. Congress has not, to date, enacted preemption.

For investors, the direction of liability law matters in a way that is easy to get backwards. A shareholder might reasonably assume that weaker liability rules are good for the company they own, and therefore good for them. In the short run, they are. But liability rules are also how the cost of harm gets identified and priced into a company before the harm becomes systemic. Preemption without a federal replacement does not eliminate that cost, it defers it and leaves it unquantified on the company’s books. And the households that ultimately bear the harm are frequently the same households holding the stock through their retirement accounts.

What the SpaceX IPO Teaches Us About the Anthropic IPO

Anthropic’s IPO is far from the only IPO receiving a ton of hype. A good parallel is SpaceX’s IPO. Back in June, SpaceX raised a record $75 billion through its IPO. It priced its shares at $135, which meant that with over 13 billion shares outstanding, its valuation was $1.77 trillion, also a record for an initial offering. Some analysts warned that the company was overvalued and that investors would “have opportunities to buy the stock at more attractive levels after the IPO.”

Retail investors did not heed these warnings. Instead, they bought as much SpaceX stock in its first two days of trading as they bought across the entire U.S. stock market that week. They bought $405 million during the first five trading sessions. During this time, the stock soared from its IPO price of $135 to over $225. This, as well as the fact that much of the growth in companies like SpaceX has already occurred due to the ability of private companies to stay private for longer and go public at dizzying valuations, led to concerns that retail investors were buying high.

Retail investors also got exposure to SpaceX through index funds. Many indices changed their rules to eliminate the waiting periods that normally prevent newly public companies from entering the index shortly after their IPOs. These waiting periods are designed to prevent index funds “from buying a stock shortly after its IPO to avoid volatile or unbalanced trading in the weeks and months following the IPO.” But indexes such as the Nasdaq-100 made it so SpaceX would be included after 15 days of trading. This represented a return to Nasdaq’s roots in “promoting high-flying firms that in some cases have not generated strong financial results,” with the effect of SpaceX’s early inclusion being that passive investors in funds tracking the Nasdaq-100 “‘are going to find themselves arguably with riskier portfolios than they were historically.’”

For both retail investors who bought SpaceX at the IPO and index funds that purchased SpaceX shortly afterwards, the fact is that so far “it has clearly proven to be a bad investment.” As Robin Wigglesworth recently showed in the Financial Times, SpaceX has underperformed the S&P 500:

Investors should be aware of SpaceX’s performance as they consider the hype surrounding Anthropic.

SEC Rule Changes Both Past and Present Increase the Risks

Investors should also be aware of the possibility of post-IPO price declines due to the opacity that permeates private companies. Private companies are not required to file periodic reports or distribute audited financial statements. This makes it much more difficult to accurately gauge the value of the company. After a company goes public and is subject to SEC filing requirements, actual financial results show whether the hype surrounding the company was justified. This means price declines may occur once companies must divulge their financial performance.

The SEC’s deregulation of the private markets has exacerbated this problem by allowing companies to remain private longer. The intent of the securities laws is that companies raise money in the public markets through a public offering, which subjects companies to transparency. But, over time, the SEC has made it easier and easier for companies to avail themselves of exceptions to the public offering requirement. This has allowed companies to now raise as much or more money in the private markets as in the public markets. So companies are staying private longer, which is why companies are able to become huge before going public.

This increases the risk that the hype will not match the reality once these companies are subjected to the periodic reporting requirements that require them to report financial information to investors. It also means that there may simply be no way for investors to earn significant returns, because as “companies delay their IPO and go public later on in their life cycles, much of the growth phase (where most of the value is created) may have already passed, and the company is then either too expensive or not worth investing in for the majority of people.” Indeed, research shows that the stock prices of post-IPO technology companies tend to experience significant declines in the initial six months of public trading.

Although the securities laws allow shareholders to sue if the hype goes so far as companies including false or misleading statements in their registration statements as part of an IPO, the Supreme Court has made it harder for shareholders to sue successfully. Some companies go public by both allowing company insiders such as employees and early investors to sell their shares to the public and by selling new shares to the public pursuant to a registration statement. In Slack v. Pirani, the Supreme Court held that to sue the company for making false or misleading statements in a registration statement, a shareholder must be able to trace their shares and prove that the shares they purchased were newly issued shares and not pre-existing stock offloaded by early investors. Because tracing shares can be difficult, this makes it harder to sue. So even if companies conceal important information that would be material to their valuation as part of the process of going public, shareholders may have a hard time obtaining relief.

Although the SEC should be taking steps to increase transparency in the private markets, it is doing the opposite. The SEC is trying to make the public markets more like the private markets. Although public companies have been required to provide investors with quarterly financial reports for over 50 years, the SEC has proposed allowing companies to provide financial information to investors only twice a year. This would cut in half the amount of information investors receive about public companies and prevent investors in newly public companies from receiving the frequent and timely updates about a company’s performance that they desperately need to determine whether the valuation of the company while it was private matches reality.

The proposed shift to semiannual reporting could have especially serious ramifications for AI companies. Movement around AI is shifting rapidly. This means that investors in AI companies need more, and not less, frequent updates about what is happening at the companies they own.

Investors should be further aware of proposed rule changes that would diminish their rights as shareholders in Anthropic and other public companies. In September, the SEC proposed to rescind the rule that facilitates the ability of shareholders to obtain a vote on matters they consider important to the company. The SEC’s action would have the effect of silencing shareholders.

This rule change would be especially pernicious as companies themselves are increasingly limiting the ability of shareholders to have a say in how the companies they supposedly own are run. For example, although Anthropic’s shares sold as part of its IPO will have voting power, that power will be limited to 49.9%, meaning that investors in Anthropic’s IPO “must be comfortable knowing their investment is effectively solely financial, with little to no real control over decision-making.” Similarly, SpaceX structured its IPO to prevent shareholders from exercising control.

The SEC’s proposed rule change, along with Anthropic’s corporate governance structure, means that shareholders will have little say in the affairs of companies whose own executives admit that it has the potential to destroy humanity. This is a time for more shareholder democracy, not less.

The SEC Needs to Enforce Risk Disclosures as Part of the IPO

The hugely consequential impact that AI, and the companies promoting AI, may have on society at large means the SEC must ensure that investors in AI companies understand the risks. So the SEC should enforce meaningful disclosures with respect to Anthropic’s IPO. This would be consistent with the SEC’s normal process for reviewing a registration statement in connection with an IPO:

After a company files a registration statement on Form S-1 . . . the SEC staff will perform a cover-to-cover review of the document to monitor and enhance compliance with the applicable disclosure and accounting requirements under the Securities Act of 1933 . . . . The SEC staff does not evaluate the merits of an investment in an IPO but rather focuses on whether the disclosures provided in the registration statement provide investors with enough information to make an informed investment decision.

In this case, the SEC should require that Anthropic provide more detailed disclosures regarding what it admits are its own fears “that its tech could be used to develop chemical weapons,” that it could become “a target of expensive lawsuits,” and that “its own product could grow out of humanity’s control.” For example, the American Economic Liberties Project states that Anthropic’s registration statement warns of the risks from “‘any judicial or regulatory expansion of tort, product liability, or consumer protection law, or criminal liability, as applied to AI models’” and says Anthropic “should quantify exposure by legal theory” and “disclose what specific cases or proposed bills pose these risk and what changes they might implement as a result.”

It also says, with respect to Anthropic’s disclosures about the risks of a catastrophe, that the company should disclose what it believes about the probabilities of these risks coming to pass—as its own leaders and researchers have done elsewhere. It says that a “registration statement that describes them in generic terms, while executives describe them elsewhere in stark ones, raises questions about whether the filing omits material information.” There is no reason for the SEC to neglect “its authority in the public interest to request additional information.”

Of course, thus far the SEC has been asleep at the switch. Before Anthropic filed its registration statement, numerous reports emerged about its supposed contents. As Colorado Law School Professor Ann Lipton said, these leaks may have violated the securities laws. That is because during “the pre-filing period, the Company should not make any public statements—including through discussions with journalists—about the proposed IPO or its financing plans in general. . . . Accordingly, care should be taken to avoid any press release, website posting, marketing campaign, conference presentation, television or radio appearance, or other corporate communication that might be viewed as ‘conditioning the market’ for the IPO.”

The purpose of these restrictions is to “prevent issuers from conditioning the market through selective publicity before investors have access to the complete registration statement.” That is why in “past IPOs involving companies such as Google and Groupon, the SEC scrutinized public statements, media interviews, and leaked internal communications that appeared inconsistent with or supplemental to the prospectus.” Yet despite numerous events that could be interpreted as attempts to condition the market with respect to Anthropic’s IPO, the SEC has taken no action. While there is certainly time for the SEC to scrutinize the IPO-related leaks, the Commission’s dismal enforcement record over the last two years does not inspire confidence.

The SEC’s failure to act makes it challenging for investors when “substantial information about the offering circulates publicly before the prospectus becomes available.” That is because “by the time investors can compare media narratives against the actual filing, market perceptions may already be influenced by incomplete information distributed before the public registration statement becomes available.” This is no less important in the case of Anthropic than in the case of SpaceX, “given the scale of the transaction and the substantial public attention surrounding the offering.”

Conclusion: You Will Probably Own This Company. Make Sure You Own the Facts.

Anthropic has done something unusual and, in important respects, admirable. It has told the public that its product has been used in a state-sponsored cyberattack. It has told the public that its own models reached real companies during flawed tests. It has told the public that it has committed to nearly $900 billion in obligations. It has told the public that advanced AI may pose catastrophic risks. Although these disclosures provide investors with important information, they also raise the stakes of the IPO. The SEC must ensure that investors have as much material information as possible about these risks, and investors must understand that they are both buying into a company facing these risks and doing so with a limited ability to influence the company’s future.

The central fact of this offering is that the people who read these disclosures and buy the stock will have almost no ability to act on what they have read. Public shareholders can never hold more than half of the vote. Founders’ shares do. A trust elects a majority of the board. The seven suppliers putting $26 billion into the company at the offering price are taking shares with no vote at all.

So the information is real, and the powerlessness is real, and they arrive together.

That combination is why this fact sheet exists. If you are going to own this company—and if you hold an index fund, you probably will, whether you choose to or not—then you should know what you are getting:

  • A company growing faster than almost any in American corporate history.
  • Commitments more than ten times its annualized revenue.
  • A product that can be turned into a cyberweapon and has been.
  • And a governance structure that places every one of those issues beyond your reach.
  • Read the risk factors. They run roughly ninety-four pages in the amended filing, and they are unusually readable. Start there and read the section describing the governance structure.
  • Know what your funds hold. Check your index funds’ AI exposure, including the suppliers whose revenue depends on companies like this one.
  • Defend the rules that remain. Oppose the SEC’s proposed rescission of Rule 14a-8. Support a federal AI liability standard rather than preemption without replacement.