How much is Anthropic targeting in its Nasdaq IPO valuation?

Anthropic has told its IPO investor group it is targeting a Nasdaq listing at a valuation of $2 trillion or more, per the FT report carried by The Decoder on September 14, 2026. The same report cites SemiAnalysis analyst Joey Brookhart saying investors expect $120 billion in annualized revenue by year's end and nearly triple that by the close of 2027. Anthropic was valued at just under $1 trillion after its last funding round of $65 billion.

What does Anthropic's $11.5 billion quarterly revenue figure actually cover?

The $11.5 billion figure is the most recent quarterly revenue Anthropic reported in its September 14, 2026 investor briefing, representing a 14-fold year-over-year jump. The revenue mix is dominated by the Claude API and the Claude consumer subscription product. The corresponding annualized run rate hit $65 billion at the end of July 2026. For an API buyer, the figure is most useful as a directional signal of API growth rather than a precise per-product breakdown, because Anthropic has not yet publicly disclosed the API-versus-subscription split.

Why is the 80% adjusted gross margin number misleading?

The 80% figure is an adjusted gross margin that excludes three significant cost lines: stock-based compensation, revenue-sharing payments to distribution partners (most importantly Amazon, which carries Anthropic on Bedrock), and the cost of training models. The FT story explicitly flags that the 80% margin number comes before the cost of training models. The practical read is that the figure is closer to a description of inference gross margin than to a description of company-wide operating margin, and inference gross margin is the line that matters for API pricing sustainability, not for headline profit.

When did Anthropic file its S-1 with the SEC?

Anthropic confidentially filed an S-1 registration with the SEC on June 1, 2026, the same filing type OpenAI is widely expected to file. The June filing is the trigger that unblocks a public listing once the SEC review wraps up. As of September 14, 2026, Anthropic had not released a public prospectus, sharing them instead with only a small group of investors first. The pricing range and share count are not yet set.

What should Claude API procurement teams do now?

Three action items: (1) re-run your cost forecast under the assumption that Fable-family pricing holds flat through the next two quarters, because the 80% adjusted inference gross margin removes the standard pre-IPO price-hike pressure; (2) keep an eye on the long-context tier as the first place a margin-driven adjustment would show up, since long-context requests carry the highest contribution margin; (3) treat any future public S-1 disclosure as free negotiating leverage on your enterprise contract, because customer concentration, geographic mix, and material contract terms (including Bedrock commitments) will appear in the prospectus.

Anthropic IPO 2026: $2T Filing Math for API Buyers

Q: How does the Anthropic IPO timeline affect Claude API pricing?

The near-term impact is small. Anthropic has no public timeline to change Claude API pricing as a function of the IPO, and the 80% adjusted inference gross margin figure actually removes one of the typical pre-IPO arguments for a price increase, namely the frontier-model-provider-needs-to-fundraise rationale. The most realistic near-term scenario is API pricing staying flat through the listing window, with any post-IPO adjustment happening only if SEC review drags into mid-2027 and the company needs to update its growth narrative. The Claude API pricing tiers as of September 2026 (Fable 5.1 at $10/$50 with cache reads at $0.25, Sonnet 5 and Opus 5 separated by capability tier, Mythos 5 held for the long-context tier) are the operating numbers, and the S-1 changes none of them today.

The S-1 Without the Spin

Anthropic has told its IPO investor group it expects to turn a profit for the second straight quarter as it moves toward a Nasdaq listing targeting a valuation of $2 trillion or more, the FT reported September 14, 2026, as carried by The Decoder. The headline numbers are loud: a 14-fold year-on-year jump in quarterly revenue to $11.5 billion, an annualized run rate of $65 billion at the end of July, and adjusted gross margins above 80 percent. The footnotes are where API buyers need to look, because every one of those numbers has a different meaning once you start cutting them down to what an Anthropic API line item actually costs.

Three things matter for developers and procurement teams right now: the S-1 was filed confidentially with the SEC back in June, so the revenue and margin figures come from a private briefing to investors rather than a public filing; the 80% margin number is an adjusted figure that excludes stock-based compensation; and Anthropic did not release the prospectus publicly last week as expected, sharing documents only with a small investor group first. That sequence matters because the same math that drives an Anthropic API price change is the math in this filing.

What the Revenue Line Actually Covers

The $11.5 billion quarterly figure and the $65 billion annualized run rate are easy to misread. Anthropic's revenue mix is dominated by two channels: the Claude API and the Claude subscription product on the consumer side. The 14-fold year-over-year jump puts the company at a run rate that is comparable to OpenAI's published numbers earlier this year, but with a narrower customer base because the API alone does not generate that volume.

For an API buyer, the most useful framing is this: Anthropic has reached the scale where the inference compute line item is no longer the dominant cost in the income statement. SemiAnalysis analyst Joey Brookhart, as cited in the same FT report, says investors expect $120 billion in annualized revenue by the end of 2026 and nearly triple that by the close of 2027. If those projections hold, the cost-of-revenue line — which is roughly what drives API price changes — has to keep shrinking per token at a rate faster than the revenue line grows.

That is the same trajectory we have already seen in the 2026 Claude API pricing history. Fable 5.1 cut cache reads from $1.00 to $0.25 per million (-75%), Fable 5 Mythos 5 raised output pricing on long-context but held input flat, and the Sonnet 5 / Opus 5 tier separation locked in a clearer per-feature price ladder. None of those moves would have made sense under a "frontier model provider losing money on every token" cost model, and the S-1 confirms that Anthropic is no longer pricing under that model.

The 80% Margin and What It Leaves Out

The 80% adjusted gross margin number is the one that will get quoted most often, and it is also the one most likely to mislead API teams. Anthropic flagged two specific exclusions that the FT highlighted:

Stock-based compensation — not in the adjusted figure. For a company of Anthropic's recent hiring scale, SBC is a non-trivial line, and treating it as a non-cash item is the standard approach in the industry, but it does mean the operating margin is lower than 80%.
Revenue-sharing payments to partners — most importantly Amazon, which carries Anthropic on Bedrock, and to a smaller extent Google Cloud. These payments sit below the gross-margin line in the adjusted framing but above it in cash-flow reality, especially for the Bedrock-served portion of Claude traffic.
Training costs — the FT story explicitly flags that the 80% margin number "comes before the cost of training models." Training cost amortization is a long-tail expense that hits the income statement over years, and an Anthropic-scale frontier training run is a multi-billion-dollar line item that is NOT in the gross margin figure.

The practical read is that Anthropic has positive unit economics on inference — meaning every Claude API call sold at current pricing covers the cost of serving it — and the headline 80% is closer to a description of inference gross margin than to a description of company-wide operating margin. That distinction matters because it tells you what kind of price pressure to expect in the next 12 months. If unit inference economics are favorable and the company needs to grow into a $120B run-rate target, the price pressure is on growth, not on profit extraction, which means API prices are more likely to stay flat or come down than to spike upward.

The S-1 Timeline and Why It Drags Into 2027

Anthropic filed an S-1 confidentially with the SEC on June 1, 2026, the same filing type OpenAI is widely expected to file, and that filing is what unblocks a public listing once the SEC review wraps up. The June filing is the first concrete date in the sequence; the September 14 investor briefing is a follow-up that previews the operating numbers, not a new filing event. The pricing range and share count are not set, and whether the IPO actually happens depends on market conditions.

For an API buyer, the realistic timeline to plan around is: SEC review of the S-1 wraps up sometime in late 2026 or early 2027, the first public prospectus appears within a few weeks of that, and pricing is set inside 30 days of the roadshow. OpenAI is working in parallel and could file confidential paperwork "within days" of the May 21, 2026 WSJ report, with a target listing starting in September 2026 — though OpenAI's CEO Sam Altman has publicly said OpenAI will not go public in 2026. The Anthropic-OpenAI sequencing matters because the first major IPO in a category sets the comparables that the second one is priced against. If Anthropic lists at a $2T valuation, OpenAI's float — when it happens — will reference that number directly.

The near-term API impact is small. Anthropic has no public timeline to change Claude API pricing as a function of the IPO, and the 80% adjusted margin figure actually removes one of the typical pre-IPO arguments for a price increase (the "frontier model provider needs to fundraise" rationale). The most realistic near-term scenario is API pricing staying flat through the listing window, with any post-IPO adjustment happening only if SEC review drags into mid-2027 and the company needs to update its growth narrative.

API Economics and the Quiet Shift

Put together, the S-1 and the FT report describe a company that has crossed an internal threshold that API buyers will feel through behavior rather than through headline price changes:

Cache economics are now first-class. The Fable 5.1 cut (cache reads from $1.00 to $0.25 per million) only makes sense if the inference gross margin can absorb it. The S-1 confirms the inference line can. Expect more cache-pricing moves in the next year, not fewer.
Long-context pricing is the lever. Anthropic's 2026 history shows output pricing held flat on standard context windows while long-context and extended-thinking tiers carry the marginal cost. The S-1's $11.5B quarterly revenue figure is built on that mix, and the company has an incentive to keep that mix because long-context requests have higher contribution margin.
Provider-on-provider pressure is going to grow. The $2T valuation target puts Anthropic at a market-cap-to-revenue multiple of roughly 4x on the annualized run rate. If the IPO prices in that range, every other frontier provider has to defend its own multiple, and the way to defend a multiple is either to grow revenue faster or to expand margin. Both paths tend to benefit API buyers, either through cheaper tokens or through more aggressive pricing tiers.
Compliance pressure on Anthropic increases. Once the S-1 is public, Anthropic has to disclose customer concentration, geographic mix, and material contract terms. Cloud commitments to Amazon (Bedrock) and Google Cloud will show up in the prospectus, which means API buyers negotiating enterprise contracts can reference those disclosures.

What to Watch Next

Three signals in the next 60 days will tell you whether the Anthropic IPO is on the timeline implied by the FT briefing or whether it slips:

A public version of the S-1. The June confidential filing is the trigger. Once the SEC review wraps up, Anthropic has 15 days to file the public version. A public S-1 will carry the revenue mix, customer concentration, and any data on training capex that the FT briefing did not disclose.
An OpenAI filing. OpenAI was the other half of the "two major AI IPOs in 2026" story. Altman's "not this year" comment to Fortune was a softening, but a confidential S-1 filing from OpenAI would force Anthropic to either match the timing or signal a slip.
API list-price moves on long-context tiers. The cheapest signal that Anthropic is recalibrating its revenue model is a price change on the long-context tier, because that is where the contribution margin is highest and where a public-comparable-driven repricing would show up first.

The S-1 and Developer Decisions

None of this changes a developer's near-term Claude API integration. The Claude API pricing tiers as of September 2026 — Fable 5.1 at $10/$50 with cache reads at $0.25, Sonnet 5 and Opus 5 separated by capability tier, Mythos 5 held for the long-context tier — are the operating numbers, and the S-1 changes none of them today. What changes is the answer to a procurement question that comes up every quarter: can we expect Anthropic to stay in this market at current pricing, or do we need to plan for a price hike that funds their growth? The FT briefing and the June S-1 say the former. The $11.5B quarterly revenue and the 80% adjusted inference gross margin imply that Anthropic's growth story no longer requires API price hikes to fund it, which means the Claude API is more likely to be a stable line item in a 2027 budget than a moving target.

For a developer or procurement team, the action items are small but real: re-run your cost forecast under the assumption that Fable-family pricing holds flat through the next two quarters, keep an eye on the long-context tier as the first place a margin-driven adjustment would show up, and treat any public S-1 disclosure as a free piece of negotiating leverage on your enterprise contract. The IPO is Anthropic's event. The pricing math is the part that lands on your invoice.

Sources

The Decoder, "Anthropic eyes Nasdaq listing as a second profitable quarter aims to win over investors ahead of a mega-IPO" — Maximilian Schreiner, September 14, 2026, carrying the FT report.
The Decoder, "Claude maker Anthropic files for IPO with the SEC" — Matthias Bastian, June 1, 2026, carrying Anthropic's announcement of the S-1 filing.
The Decoder, "OpenAI could file confidential IPO paperwork within days" — Maximilian Schreiner, May 21, 2026, carrying the WSJ report on OpenAI's IPO timeline.
Cross-reference (apirank): "Claude API Abuse Report 2026: Developer Impact" (September 14, 2026) — covers the parallel security-side story for the same week.
Cross-reference (apirank): "Anthropic $517B Compute Deal: Claude API Pricing" (September 8, 2026) — covers the prior compute-cost context that frames Anthropic's training capex.