On Thursday, May 28, Anthropic announced a Series H funding round that raised $65 billion at a post-money valuation of $965 billion. The figure sits a rounding error below the trillion-dollar mark, and it arrives with a vocabulary problem. Venture capital has a word for a private company worth $1 billion, the unicorn, coined by the investor Aileen Lee in 2013. It has words for the rarer animals above it, the decacorn at $10 billion and the hectocorn at $100 billion. It has no word at all for what Anthropic has become. The lexicon ran out before the valuations did.
A Round That Should Not Exist
Most companies never reach Series C. A Series H is so rare that the short list of firms that have raised one reads like a roll call of the defining consumer technologies of the past two decades: Facebook, Lyft, Discord, Slack. Anthropic’s round dwarfs all of them. The $965 billion valuation is roughly triple the $380 billion the company commanded as recently as February, a tripling achieved in a single quarter. Its run-rate revenue crossed $47 billion earlier this month, against roughly $10 billion a year earlier, the kind of curve that explains, if it does not entirely justify, the price. By Fortune’s reckoning the valuation now rivals the annual military budget of the United States and exceeds the combined market capitalization of every American airline. OpenAI, its closest rival, is not far behind, last valued near $852 billion and reportedly preparing to file a confidential listing prospectus within weeks.
The Capital the Machines Demand
These are not ordinary equity rounds. They are infrastructure programs wearing the costume of venture finance. Training and serving frontier models consumes capital on a scale with no precedent in software. The four largest American cloud operators, Amazon, Microsoft, Alphabet and Meta, spent a combined sum approaching $380 billion on capital expenditure in 2025, much of it on AI data centers, and their guidance for 2026 points toward roughly $700 billion, close to double in the space of a single year. Nvidia, the company selling the shovels, booked a record quarter of data-center revenue above $60 billion. Anthropic’s own round reflects this physical reality: alongside the financial backers sit Micron, Samsung and SK Hynix, the memory and chip suppliers whose silicon the models run on. When the makers of your raw materials are also your shareholders, the line dividing a customer, an investor and a vendor has effectively dissolved. Roughly $15 billion of the round was previously committed money from the hyperscalers, including the $5 billion Amazon pledged in April.
The Law That Made It Possible
The more interesting question is not why these companies are worth so much, but why they can reach such valuations without ever selling a share to the public. The answer is regulatory, and it is largely forgotten. For most of the modern era, American securities law contained a tripwire. Section 12(g) of the Securities Exchange Act forced any company with more than $10 million in assets and 500 or more shareholders of record to register with the Securities and Exchange Commission and begin reporting as though it were public. The rule was written for an age when 500 owners signaled that a firm had become a matter of public interest. In practice, it also functioned as a deadline.
Google is the cleanest example. By the end of 2003 the company had crossed the 500-shareholder threshold, largely through the stock options it had granted employees. Its own IPO registration statement conceded the point, noting that the law obliged certain private companies to report as if public, a requirement that accelerated its decision to list. Google filed in April 2004 and went public that August, not because it needed the money, but because the rule left it little choice. Facebook would run into the same wall, “the Facebook problem,” in the years that followed.
Two Thousand Shareholders, and the Dam Breaks
The Jumpstart Our Business Startups Act, signed in April 2012, dismantled the tripwire. It raised the shareholder ceiling from 500 to 2,000, and, decisively, it stopped counting employees who held shares through compensation plans. A company could now grant equity to thousands of staff, raise from a long roster of institutions, and still sit comfortably below the threshold that once forced disclosure. The deadline was gone. McKinsey notes plainly that Google’s 2004 listing was triggered by the old 500-shareholder limit, and that the new ceiling is precisely what now permits companies to stay private. The median technology company that does eventually list is now roughly eleven years old, against under seven a decade ago.
The Logic of Staying Private
Freed from the tripwire, the largest private companies have found that public markets offer them little they cannot get elsewhere, and much they would rather avoid. Late-stage private capital is abundant: sovereign wealth funds, crossover investors and the hyperscalers themselves will write checks of a size once available only on an exchange. Employee liquidity, historically the strongest argument for an IPO, is now handled through company-sponsored tender offers and a deepening secondary market. What staying private buys is control. There is no quarterly earnings theater, no activist shareholder building a position, no obligation to justify a multi-year, multi-hundred-billion-dollar capital plan to investors who measure time in three-month increments. For a company spending at Anthropic’s rate on a wager that may not resolve for years, that insulation is not a luxury. It is strategically essential.
The Thousand-Fold Unicorn
Put the three forces together and the new species comes into focus. The AI revolution supplies the demand and the revenue growth. The capital expenditure super-cycle supplies the appetite for sums no earlier startup could have absorbed. And the post-2012 regulatory settlement supplies the permission to grow to nearly a trillion dollars in private hands. The unicorn was named in 2013 for a company worth $1 billion. Anthropic is, in the most literal sense, a thousand-fold unicorn, and it is not alone. OpenAI sits beside it; SpaceX and ByteDance occupy the same rarefied tier; and the global population of unicorns has swollen from fewer than 40 in 2013 to well over a thousand today. The exclusive club has become a crowded one, even as its ceiling has risen beyond anything its founders imagined.
The Reckoning That Waits
None of this is without hazard, and the detached observer should resist the urge to call it either a miracle or a bubble. Some of the capital is visibly circular. Nvidia invests in the AI labs that buy its chips; the labs commit to cloud providers that buy more of the same chips; the value travels in a loop that flatters everyone inside it. OpenAI is reported to have generated on the order of $13 billion in revenue last year against capital commitments measured in the trillions over the coming decade. The asset manager GMO has called the arrangement reminiscent of the circular financing of the internet bubble. Whether this proves the dawn of the most valuable industry in history or the most expensive misallocation of capital since the fiber-optic glut, the structure is now set. The largest concentration of private wealth creation in the Western world is taking place almost entirely outside the public markets, visible to ordinary investors only when, and if, these companies finally choose to list. Anthropic says that moment may arrive within months. For now the trillion-dollar company exists, it simply has no name, and the public owns none of it.
Read our full Report Disclaimer.
Report Disclaimer
This report is provided for informational purposes only and does not constitute financial, legal, or investment advice. The views expressed are those of Bretalon Ltd and are based on information believed to be reliable at the time of publication. Past performance is not indicative of future results. Recipients should conduct their own due diligence before making any decisions based on this material. For full terms, see our Report Disclaimer.