A $1.4 trillion IPO should make the engineering and business numbers easy to understand.
The money is enormous. The ticker is new. And, for the first time, investors can publicly price the company many people have spent the last three years debating.
But beneath the headline valuation, Anthropic’s filing reveals a business with an unusual combination of traits: rapidly improving operating economics, massive infrastructure commitments, and investors that may also be its suppliers.
That is where the real story begins.
On Tuesday, Anthropic filed an amended registration statement announcing plans to issue 825 million shares at $70 to $78 each. The shares will trade on the Nasdaq Global Select Market under the ticker ANT, with pricing expected on November 18 and trading beginning on November 19.
Goldman Sachs and Morgan Stanley are leading the offering, with up to 20% of the shares allocated to retail investors.
At the top of the proposed range, Anthropic would raise $64.4 billion. That would make it the second-largest IPO in history, behind SpaceX’s $85.7 billion offering in June, while giving the company a fully diluted valuation of up to $1.4 trillion.
The trillion-dollar number dominated the conversation. It is exciting, memorable, and backed by a public filing.
It is also incomplete.
Three details in that filing explain what Anthropic is becoming—and why its biggest expense may have been getting a higher valuation.
The $104 billion loss is not an operating loss
Anthropic reported a net loss of $104.0 billion for the first nine months of 2026. At first glance, that suggests a company losing more than a hundred billion dollars while trying to scale.
Here is what the financial statements show instead:
| $B, except percentages | FY2023 | FY2024 | FY2025 | 9M 2025 | 9M 2026 |
|---|---|---|---|---|---|
| Revenue | 0.04 | 0.39 | 4.59 | 2.55 | 33.24 |
| Operating loss | (0.55) | (2.98) | (8.06) | (5.22) | (1.26) |
| GAAP net loss | (1.53) | (8.31) | (41.97) | (25.11) | (104.04) |
Revenue grew 1,206% year over year for the first nine months of 2026.
More importantly, the operating loss narrowed from $5.22 billion to $1.26 billion. That is the loss connected to running the business.
Anthropic’s compute and infrastructure expense fell from 172% of revenue to 57%. Operating cash flow improved from negative $2.02 billion to positive $6.57 billion.
Those numbers describe a business moving toward operating profitability at extraordinary scale.
Now compare them with the GAAP net loss, which increased from $25.11 billion to $104.04 billion.
The difference is almost entirely accounting.
According to the filing, $94.324 billion of the nine-month net loss came from an adverse fair-value change related to the company’s equity instruments. In other words, much of the reported loss was caused by the value assigned to those instruments rising, not by Anthropic spending $104 billion on its operations.
This is one of the oldest tricks in public-market accounting: the headline net loss can look catastrophic even while operating performance is improving, because a change in valuation flows through the income statement.
So the $104 billion number is technically accurate and economically misleading.
It is not the cost of running Anthropic’s AI models, training its systems, or serving its customers. It is primarily the accounting consequence of the company becoming more valuable.
Which is rather ironic.
Anthropic’s biggest reported expense was, in a sense, being worth more.
The infrastructure bill arrives before the revenue
The second important detail is Anthropic’s infrastructure commitment.
The company has agreed to roughly $900.9 billion of compute and infrastructure spending. About $70.7 billion of that commitment falls due within the next twelve months.
Anthropic says it currently has $131.8 billion in cash and short-term investments.
On paper, that means the company has enough liquidity to cover the near-term commitment several times over. But the scale and duration of the larger obligation change how investors should think about Anthropic.
This is no longer simply a software company that rents GPUs when it needs them. The contracts required to support its growth run into 2045.
That creates enormous visibility for future demand. It also creates enormous fixed obligations.
The model resembles a hyperscaler more than an ordinary AI startup: invest heavily in data centers, processors, networking, and power contracts; build capacity years ahead of demand; and recover those costs through long-term customer agreements.
That strategy can produce powerful advantages. Capacity is scarce. Once infrastructure is secured, it is difficult for competitors to replicate. Long-term contracts can also create a backlog that appears more durable than ordinary subscription revenue.
But it is a capital-intensive model with a different risk profile from software that can be delivered using relatively modest infrastructure.
The valuation assumes that today’s demand continues, that the hardware performs as expected, that power remains available, and that future AI products generate enough revenue to earn an acceptable return on hundreds of billions of dollars in committed spending.
The technology is the differentiator. The capital structure is the constraint.
The investors may also be the suppliers
The third footnote is arguably the most revealing.
Seven companies are buying $26 billion of Anthropic stock at the IPO price. Several of those companies are also leasing Anthropic the computing infrastructure it uses to operate its models.
That does not make the arrangement automatically improper. Large technology companies routinely finance suppliers, invest in partners, and enter commercial agreements with businesses they depend on. Sometimes the arrangements are strategic. Sometimes they create conflicts of interest. Sometimes they are both.
But investors should understand the circularity.
A company buys Anthropic stock, which increases the value of the infrastructure commitments made by Anthropic. That value can then support more spending, additional capacity, and more revenue for the companies providing the infrastructure.
It is not a Ponzi scheme. That characterization would be lazy and inaccurate.
It is a feedback loop.
And at a $1.4 trillion valuation, that feedback loop matters. The gap between Anthropic’s operating performance and the size of the capital surrounding it is large enough that financing relationships deserve more attention than a typical IPO receives.
What the footnotes reveal
Taken together, the filing describes something more complex than “another AI company going public.”
Anthropic’s revenue is accelerating faster than its operating expenses. Its operating loss is shrinking, and operating cash flow has turned positive.
At the same time, the company has accepted roughly $900.9 billion in infrastructure commitments, with hundreds of billions of dollars still to be paid over many years. And some of the companies funding the IPO are also the ones earning revenue from those commitments.
This is an AI business acquiring the financial characteristics of a capital-intensive infrastructure company while still being valued according to software-like expectations.
That may prove to be exactly the right strategy. Anthropic does not need to guess which models will win if it controls the infrastructure, distribution, and capacity required to serve the market.
But investors should not confuse the trillion-dollar headline with a simple measure of business quality.
The interesting part of this IPO is not that Anthropic finally put a price on itself. It is that the price exposes a business model in which valuation, financing, infrastructure, and suppliers are increasingly intertwined.
The market is paying a historic price for the possibility that Anthropic can turn extraordinary compute commitments into durable infrastructure advantages.
Whether that works will depend less on the IPO valuation than on what happens after the story reaches the footnotes.
This was first published on Sol AI — https://thesolai.github.io
Top comments (0)