On June 1, 2026, Anthropic confidentially submitted a draft registration statement to the SEC. It is the first formal step toward the most anticipated technology IPO since the AI cycle began. The headline number does the work the headline is supposed to do: a private valuation near $965 billion, set weeks earlier by a $65 billion Series H, on a revenue run rate that reportedly reached roughly $47 billion. Those figures are large enough to stop being information and start being scenery.
The useful question is not whether Anthropic is impressive. It obviously is. The useful question is what a disciplined operator or investor should actually check before deciding what any of it means. A confidential S-1 is not an audited prospectus, a run rate is not revenue, and a private mark is not a clearing price. This piece is a CFO-vantage reading of the filing: the facts as reported, the four numbers that carry the weight, and a framework for pricing an AI IPO without being anchored by the biggest figure in the room.
What Anthropic has actually filed
Stripped to what is on the record, here is the state of play. Every figure below is journalistic or self-reported; none of it is audited GAAP disclosure yet, which is the single most important caveat in this entire article.
| Item | Reported figure | Note |
|---|---|---|
| Filing | Confidential draft S-1, June 1, 2026 | Rule 135; share count and price not set |
| Last private valuation | ~$965 billion | Series H, closed late May 2026 |
| Series H raise | $65 billion | Led by Altimeter, Dragoneer, Greenoaks, Sequoia |
| Revenue run rate | ~$47 billion annualized | Up from roughly $9 to $10 billion a year earlier |
| Gross margin | ~40% today; ~77% targeted by 2028 | Cash-flow-positive targeted 2028 |
| Major backers | Amazon (~$13B invested), Alphabet (~14%, capped at 15%) | Both booked large paper gains in Q1 |
| Reported underwriters | Goldman Sachs, JPMorgan, Morgan Stanley | Per press reports, not confirmed in a public filing |
| Possible timing | A Nasdaq listing as early as Q4 2026 | Anthropic says timing depends on SEC review and market conditions |
A confidential filing buys optionality, not a date. It lets the SEC begin review while the company keeps its financials private until it chooses to flip the S-1 public, which by rule must happen at least fifteen days before any roadshow. So the interesting disclosures, the audited ones, are still ahead of us. What follows is how to be ready to read them.
The number that matters is not the valuation
Start with the trap. The $965 billion mark is a private-round post-money valuation, which is a negotiated number between a company and a small set of late-stage investors. It often carries structure such as liquidation preferences, ratchets, and information rights that a common share on a public exchange does not. It is a real data point about what sophisticated buyers paid. It is not the price the public market will set, and treating the two as interchangeable is the first mistake most coverage makes.
The number I would put first is the revenue quality, not its size. A ~$47 billion run rate that grew roughly fivefold in a year is genuinely extraordinary. But “run rate” annualizes a recent short window, so it flatters any business that is accelerating and punishes none. The audited S-1 will restate this as trailing GAAP revenue, and the gap between “run rate” and “trailing twelve months” is exactly where an operator earns their skepticism. The same discipline applies to any AI tool you evaluate for your own stack; I wrote the general version of that test in how founders should evaluate AI workflows before buying automation.
Three numbers under the run rate
The run rate tells you the company is growing. These three tell you whether the growth is a business.
Gross margin trajectory. Anthropic reportedly runs near a 40% gross margin today and targets 77% by 2028. That spread is the whole thesis. Traditional software prints 80%-plus gross margins because serving one more customer costs almost nothing; a frontier-model company pays for every token it generates in compute. Whether Anthropic is a software company or a capital-intensive utility with a software interface depends entirely on whether that margin curve bends the way the plan says. The public S-1 will show the actual trend, not the target.
Compute obligations. The reported commitments are the liability side that the run rate never mentions: well over $100 billion to AWS, roughly $200 billion to Google Cloud, plus other suppliers. These are the closest thing an AI company has to a fixed-cost base, and they are enormous relative to current revenue. A CFO reads the obligations table before the revenue table, because the obligations are contractual and the revenue is a forecast.
Customer concentration and durability. The encouraging signals here are real: Claude Code reportedly reached roughly $2.5 billion in annualized revenue, and enterprise accounts paying more than $1 million a year roughly doubled to over a thousand in a single quarter. Concentration among a small number of very large accounts is a strength and a fragility at once. The S-1’s customer-concentration disclosure is where you learn which.
How to read an AI IPO like an operator
The framework below is the checklist I would run against Anthropic’s public S-1 the day it drops, and against OpenAI’s or any large AI listing that follows. It generalizes, which is the point: the discipline outlives this specific deal.
| Check | What you are actually testing | Red flag |
|---|---|---|
| Run rate vs. trailing GAAP revenue | Whether growth is as clean as the annualized number implies | A wide, unexplained gap |
| Gross margin trend, not target | Software economics vs. compute-bound utility | Margin flat or falling as revenue scales |
| Compute obligations vs. revenue | The fixed-cost base the run rate hides | Multi-year commitments dwarfing revenue with no offsetting demand |
| Customer concentration | Durability of the revenue base | Revenue leaning on a handful of accounts or one cloud partner |
| Insider and backer selling | What the people with the most information are doing | Heavy early-investor exit at the IPO |
| Valuation vs. clearing price | Whether the private mark survives public scrutiny | IPO priced well below the last private round |
None of these require inside information. All of them will be answerable from the audited filing. Each one converts a headline into a question you can actually adjudicate. That is the entire job.
What I would actually watch
If I were advising an operator with a genuine reason to have a view, I would ignore the valuation debate almost entirely and watch two things. That covers an employee weighing equity, a founder benchmarking their own metrics, and an investor sizing a position.
First, the gross-margin line in the public S-1, quarter over quarter. If it is bending toward the 2028 target, the software thesis is intact and the compute obligations are an investment. If it is flat while revenue triples, the company is buying growth at a cost that the run rate is politely not mentioning, and the entire valuation rests on a curve that has not yet started to bend.
Second, what the best-informed holders do. Amazon and Alphabet booked billions in paper gains marking their stakes to the Series H price; the IPO converts paper to a public reference. Whether large early investors sell into the offering or hold is the most honest signal in any prospectus, because it is a decision made with better information than any outsider has. A private mark is an opinion. A clearing price is a verdict. An IPO is the moment the two are forced to meet. That is also why an IPO is a poor moment to form a view and a good moment to test one you already hold, using a system you built in calmer conditions. The case for building that system first is in why you build the research architecture before the investing decisions.
A sanity check: what would the market’s own multiple pay?
Here is a discipline that cuts through the whole valuation debate. Take Anthropic’s sales for 2026 and value them the way the public market values everyone else’s sales, on a price-to-sales multiple. A run rate is already an annualized figure, so the ~$47 billion reported in mid-2026 is a reasonable proxy for the sales the company is exiting the year on; if it keeps climbing, every number below scales up in proportion.
The S&P 500 currently trades around 3.0 to 3.6 times sales, which is historically elevated against a long-run average since 2000 of roughly 1.8x. Apply those to Anthropic’s run-rate sales:
| Basis for 2026 sales | Price-to-sales multiple | Implied valuation |
|---|---|---|
| ~$47B run rate | 1.8x (S&P 500 long-run average) | ~$85 billion |
| ~$47B run rate | 3.0x to 3.6x (S&P 500 today) | ~$140 to $170 billion |
| ~$47B run rate | 20.5x (implied by the ~$965B private mark) | ~$965 billion |
Read the bottom row carefully, because it is the point. The private valuation embeds a price-to-sales multiple of roughly 20x, which is about six to seven times what the broad market pays for a dollar of sales today and more than ten times the market’s long-run average. Valued on the S&P 500’s own current multiple, ~$47 billion of sales supports something closer to $140 to $170 billion; valued on the market’s historical multiple, closer to $85 billion. Either way it is a fraction of the last private mark.
That does not make the private investors wrong. It makes explicit what they are actually paying for: not this year’s sales, but a bet that sales keep compounding at a rate almost no public company has sustained, and that the ~40% gross margin climbs toward software economics. The gap between $140 billion and $965 billion is not a valuation. It is a growth and margin forecast wearing a valuation’s clothing. Price the sales you can see, then decide separately whether you believe the forecast. That is the only honest way to hold both numbers at once.
Common mistakes
Treating the private valuation as a price. $965 billion is what a handful of late-stage investors negotiated, structure included. The public market may pay more or less, and the gap is the actual news.
Reading run rate as revenue. Annualizing a recent window flatters every accelerating business. Wait for trailing GAAP revenue in the audited S-1 before drawing the growth line.
Ignoring the obligations table. Over $300 billion in reported compute commitments is the fixed-cost base the run rate never shows. Read it before the revenue.
Letting the model argue your side. If you use AI to research the IPO, it will fluently support whatever thesis you feed it. Make it build the bear case too; the discipline for that is in the minimum viable AI investing workflow.
FAQ
When is Anthropic’s IPO? There is no confirmed date. Anthropic filed a confidential draft S-1 with the SEC on June 1, 2026, which starts the review clock but sets no timeline. Press reports point to a possible Nasdaq listing as early as Q4 2026, but the company has said timing depends on SEC review and market conditions.
What is Anthropic’s valuation? Its last private valuation was roughly $965 billion, set by a $65 billion Series H that closed in late May 2026. That is a negotiated private mark, not a public-market price; the IPO will set the latter.
How much revenue does Anthropic make? Reports put its annualized revenue run rate near $47 billion in mid-2026, up from roughly $9 to $10 billion a year earlier. That is a run rate, not audited trailing revenue. The public S-1 will restate it on a GAAP basis.
Is Anthropic profitable? Not disclosed, and almost certainly not yet on a GAAP basis given heavy compute spending. Anthropic reportedly targets a ~77% gross margin and cash-flow-positive operations by 2028; the trajectory toward that target matters more than any single quarter.
What would Anthropic be worth on a normal sales multiple? On the S&P 500’s current price-to-sales multiple of roughly 3.0x to 3.6x, its ~$47 billion run-rate sales support a valuation near $140 to $170 billion; on the index’s long-run average of about 1.8x, closer to $85 billion. The ~$965 billion private mark implies a multiple around 20x sales, six to seven times what the broad market pays, so the private valuation is a bet on future growth and margin expansion, not a price for current sales.
Should I invest in the Anthropic IPO? This article is analysis, not advice. The disciplined move is to wait for the audited S-1 and read gross-margin trend, compute obligations, customer concentration, and insider selling before forming a view. It is the same checklist you would run on any AI listing.