Anthropic’s S-1 Turns the AI Race Into a Test of Business Fundamentals

Anthropic’s prospectus offers the clearest view yet of what it costs to build and sell frontier AI—and sets the bar OpenAI will be measured against if it lists in 2027.

For most of the past few years, the contest among AI labs has been scored on capability: whose model codes better, reasons further or handles longer tasks. Anthropic’s decision to go public adds a second scorecard. Once a company’s books are open, the question shifts from what its models can do to whether those models can pay for themselves.

The company has applied to list its Class A shares on the Nasdaq Global Select Market under the ticker ANT. The share count, price range and offering valuation have not yet been set.

The road to a public filing

Anthropic’s public S-1 landed on October 26, nearly five months after it confidentially submitted a draft to the SEC on June 1. That private filing followed a funding round that valued the company at $965 billion, more than double its $380 billion valuation in February.

The listing date has slipped as the process unfolded. An October debut was the early expectation; the target later moved to November, reportedly so investors could see third-quarter results before buying. Current expectations point to a roadshow starting around the week of November 9 and a first trade before Thanksgiving, though the company has committed to neither.

Several key terms remain open. Until an amended filing sets a price range and share count, and the SEC declares the registration effective, no shares can be sold.

Reading the numbers

Revenue is the headline. Anthropic booked $16.3 billion in the first six months of 2026, more than thirteen times the $1.2 billion it earned in the first half of 2025.

Profitability is more mixed. For the half as a whole, the company lost $693 million at the operating level. The second quarter on its own, however, swung to $974 million of operating income, an 8% GAAP operating margin.

The bottom line looks far worse: an $80.7 billion net loss for the half. Nearly all of it reflects a non-cash charge from revaluing financing instruments such as convertible notes, not the cost of serving customers. Investors who stop at the net-loss figure will misread the business; operating results and cash flow tell the more useful story.

The most telling ratio may be compute. Spending on compute and infrastructure fell from 201% of revenue in the first half of 2025 to 55% in the first half of 2026. A year ago, the company spent about two dollars on computing for every dollar it earned; now it spends a little over fifty cents.

That improvement sits against a very large forward bill. The filing describes commitments to data centers and cloud capacity reportedly exceeding $540 billion, including a compute arrangement with SpaceX worth roughly $3 billion a month. Those obligations are largely fixed. Revenue is not, and it must keep growing quickly for the cost ratio to keep falling.

An offering built for scale

Measured by the amount it hopes to raise, this could rank among the largest IPOs ever. Expectations have centered on a raise comparable to or larger than SpaceX’s June offering of more than $86 billion. Talk of a market value well above $1 trillion has circulated among investors, but the company has not put a number on it.

SpaceX’s experience offers a caution. Its stock jumped well above the offer price, then sank below it, and has since returned to roughly where it began. Strong early demand did not translate into a stable price.

Anthropic appears to be planning around that risk. Options under discussion include letting some existing holders sell in the offering in exchange for a longer-than-standard lock-up, and a large revolving credit line arranged alongside the listing. None of this is final.

From model race to business race

Benchmarks still matter, but they are no longer enough. A lab now has to hold onto customers, fit into the daily work of the people using it, control what each request costs to serve and secure the computing power to keep improving.

That shifts how the market will assign value. A model that leads on tests but loses customers, or costs too much to run, may be worth less than a slightly weaker one built into the workflows companies depend on. Labs that can show measurable gains in productivity or new revenue for their customers will have the stronger case.

The effects reach well beyond the labs. Chipmakers, data center operators and power suppliers are all riding the same demand. Eventually that demand has to rest on products customers will pay for, and public investors will test whether projected usage justifies the spending.

Price competition adds pressure. Cuts by rivals have raised doubts about how long premium pricing for top models can last, and local resistance to new data centers could slow construction across the industry.

The OpenAI question

OpenAI is the obvious point of comparison, and its timeline is clearer than it was a few months ago. It submitted its own confidential draft S-1 in early June, about a week after Anthropic. Since then, its leadership has reportedly told employees the company expects to be public in 2027, and it is said to be exploring another large private raise at a valuation well above $1 trillion in the meantime. No listing date has been announced.

Should OpenAI list next year, the two companies would give the market a rare side-by-side test. OpenAI brings enormous consumer reach; Anthropic brings deep enterprise adoption and a rapidly improving cost structure. How investors price each will reveal whether they reward audience size, business revenue, computing efficiency, steady margins or some mix of these.

Waiting cuts both ways. OpenAI gets more time to build revenue and a live public comparable to price against. Anthropic gets to define the metrics by which an AI lab is judged, and to tap investor demand before a second giant listing competes for it.

There is also the question of capacity. Even enthusiastic investors have limits, and a strong reception for one AI listing does not guarantee one for the next. Each company will be weighed on its own growth, spending, infrastructure obligations, competitive position, governance and risks.

Five things to watch

1. Who the revenue comes from. No single customer exceeded 10% of revenue through the first nine months of 2026, but a handful of large customers accounted for roughly a quarter of it. The company expects to keep depending on a relatively small group of big accounts.
2. Commitments against growth. The cost ratio is moving the right way, but long-term infrastructure contracts don’t flex if demand cools.
3. The third quarter. The listing was reportedly timed to include Q3 results. Whether the second quarter’s operating profit held up will color how investors read everything else.
4. Mission and governance. Anthropic is a public benefit corporation, and its structure includes a subsidiary dedicated to its stated purpose. Buyers will want clarity on how that purpose is weighed against shareholder returns.
5. The risk section. The prospectus devotes extensive space to risk factors, including the potential dangers of increasingly capable AI. Few companies describe the hazards of their own product so plainly.

What the filing changes

The prospectus doesn’t settle whether AI valuations are justified. What it does is make the economics visible. Its lasting effect may be a new yardstick: AI companies will be judged on whether technical progress becomes durable customer value, and whether that value covers the enormous cost of building and running frontier models.

More detail is coming soon: a price range, a share count, third-quarter figures and, eventually, a first day of trading. Each will be read as a signal not only about Anthropic but about every private AI company considering a listing.

If OpenAI follows in 2027, the comparison will sharpen. The winners of the next phase will be the companies that can show not just what their models are capable of, but how consistently they turn that capability into products people rely on and businesses that sustain themselves.

Disclaimer: This article is for information only and is not investment advice or a recommendation to buy or sell any security. It reflects public information as of October 27, 2026. Some details are based on unconfirmed reports and may change. Investing in a newly public company carries significant risk, so do your own research or consult a financial professional.