Anthropic
Anthropic, PBC · AI research and enterprise software
Anthropic, PBC is an AI safety and research company that builds the Claude family of AI models. It sells model access through a usage-based API, enterprise and consumer subscriptions (Claude Pro, Max and Ultra), and through cloud service providers' marketplaces, which billed 44% of revenue in the first nine months of 2026. The company is incorporated as a public benefit corporation with a Long-Term Benefit Trust that elects a majority of its board.
The offer
- Price range
- TBA
- Implied valuation
- TBA
- Expected listing
- TBA — no listing date has been set in the filing
- Ticker / exchange
- ANT · Nasdaq
The price range and share count are deliberately left blank in the latest amendment. This dashboard will populate them the day the filing is updated, and shows TBA until then rather than guessing.
TTM revenue
Calculated$35.3B
FY2025 $4.59B + 9M 2026 $33.24B − 9M 2025 $2.55B
Revenue growth, 9M 2026
Reported+1,206%
Versus the nine months to 30 September 2025
Operating margin, 9M 2026
Calculated-3.8%
Operating loss $1.26B on revenue of $33.24B
Net loss, 9M 2026
Reported-$104.0B
Includes a $94.3B non-cash change in fair value of convertible notes
The one-minute version
What you are actually buying
How the business makes money, what the filing proves, and what it leaves out.
AI investment summary
AI interpretationAnthropic is growing faster than almost any enterprise software company on record: revenue rose from $41m in 2023 to $4.59bn in 2025, then to $33.2bn in just the first nine months of 2026. The business has already crossed into operating profit in one quarter (Q2 2026, +$974m), because compute costs fell from 172% of revenue a year ago to 57% today. The tension is the balance sheet: against $131.9bn of cash and investments sit $895.4bn of contracted compute and infrastructure commitments, and the reported $104.0bn net loss is dominated by a $94.3bn non-cash charge from revaluing convertible notes, not by operations. In plain English: a rapidly scaling, capital-hungry business that has proven it can price above cost, but not yet that it can fund decades of compute from its own cash flow. AI interpretation of reported figures, not investment advice.
Written by GMG Pro AI from the figures in the S-1/A below. It is interpretation, not a company statement and not advice.
Three reasons this could work
Operating leverage is real, and it arrived fast
Compute and infrastructure costs fell from 172% of revenue in the first nine months of 2025 to 57% in the same period of 2026. The company reported its first quarterly operating profit in Q2 2026 (+$974m) and a near-breakeven Q3. If that ratio holds as revenue scales, small changes in compute pricing swing profit dramatically.
9M 2026 compute and infrastructure $18.8B on revenue of $33.2B; Q2 2026 operating income $974m.
Enterprise land-and-expand is compounding
Customers spending at least $100,000 on a trailing-twelve-month basis grew from more than 950 to more than 11,900 in a single year, and they generated $24.5B of the $33.2B nine-month revenue. Remaining performance obligations of $13.8B give visibility on contracted work.
Filing: customers with $100,000+ TTM revenue, and remaining performance obligations.
The balance sheet funds the build-out
Anthropic has raised $141.1B in aggregate and holds $131.9B of cash and investments. That pre-funding is what makes $895.4B of contracted compute commitments financeable without immediate reliance on the public markets — although the gap between the two is the central question for IPO investors.
Filing: capital raised, cash and short-term investments, purchase commitments and leases.
Three things that could go wrong
Commitments dwarf liquidity
$895.4B of contracted spend — substantially all compute and infrastructure — sits against $131.9B of cash and investments. $69.8B falls due within twelve months and $447.5B before the end of 2030. If demand or pricing disappoints, the company must still pay, and may need to raise equity or debt on worse terms.
A short history of losses and cash burn
Operating losses were $551m, $2.98B and $8.06B in 2023, 2024 and 2025, and operating cash flow was negative in all three years. 2026 has flipped both trends, but one good nine-month period is not a track record, and the company warns it may not sustain its growth rate.
Channel and customer concentration
44% of nine-month 2026 revenue reached customers through cloud service provider marketplaces, up from 11% in 2023, and those platforms bill end customers and remit cash net of fees. Losing or renegotiating a hyperscaler relationship would disrupt access and collections at once.
The numbers behind the story
Source documentTTM revenue
FY2025 $4.59B + 9M 2026 $33.24B − 9M 2025 $2.55B
$35.3B
Revenue growth, 9M 2026
Versus the nine months to 30 September 2025
+1,206%
Operating margin, 9M 2026
Operating loss $1.26B on revenue of $33.24B
-3.8%
Net loss, 9M 2026
Includes a $94.3B non-cash change in fair value of convertible notes
-$104.0B
Cash and investments
$95.6B cash plus $36.3B short-term investments at 30 Sep 2026
$131.9B
Contracted commitments
Substantially all compute and infrastructure; includes $69.8B due within 12 months
$895.4B
Remaining performance obligations
About 75% expected to be recognised within 12 months
$13.8B
Customers with $100k+ TTM revenue
Up from more than 950 a year earlier
11,900
Gross margin, 9M 2026
Revenue less cost of revenue, which is mostly compute
85.5%
Weighted-average shares, 9M 2026
Post-offering share count is left blank in the amendment
2.08B
Reported results
The financials, without the spin
Every figure is from the S-1/A. Balance-sheet style ratios are calculated from it and labelled.
Fiscal years 2023–2025 plus the nine months to 30 September 2026. FY2026 is not complete.
Revenue · 9M 2026
$33.2B
+1206% year on year
Operating result
-$1.3B
Margin -3.8%
Net loss
-$104.0B
Includes $94.3B non-cash note charge
Operating cash flow
$6.6B
Cash generative
Revenue · annual & interim
S-1/A- Revenue
- Year-on-year growth
What does this mean for investors?
The reported financial table
Every line from the consolidated statements of operations, including personnel, sales and marketing, and stock-based pay.
Start free trialGrowth
How fast, and for how long
The filing documents one of the steepest revenue ramps in software history. Growth rate is also the input that moves a valuation most, so it is worth seeing clearly.
Quarterly revenue, first to latest
$376M → $17.0B
45.1× in seven quarters
Latest quarter growth
+1127%
Year on year, Q3 2026 vs Q3 2025
Enterprise customers
11,900
Spending $100k+ TTM, up from 950 a year earlier
Quarterly revenue and sequential growth
S-1/A- Revenue
- Growth vs prior quarter
What does this mean for investors?
Revenue against compute spend
- Revenue
- Compute as % of revenue
What does this mean for investors?
Revenue milestones
FY2023
Operating margin -1343.9%
$41M
—
FY2024
Operating margin -772.0%
$386M
+841%
FY2025
Operating margin -175.8%
$4.6B
+1088%
9M 2025
Operating margin -205.3%
$2.5B
—
9M 2026
Operating margin -3.8%
$33.2B
+1206%
The nine-month 2026 column is not a full year. Annualising it would overstate the run-rate because the quarterly sequence is still steeply rising — a simple annualisation implies about $68bn.
Business model
Where the money comes from, and where it goes
Anthropic sells access to the same underlying models in several ways. The filing does not break out product revenue, so anything showing a precise split is an estimate.
Usage-based API access
Split not disclosedPer-token consumption
Customers pay for what they use through the Claude API. The filing does not break out consumption revenue from subscription revenue, so any split you see elsewhere is an estimate.
Consumer subscriptions
Not disclosedFixed monthly fee
Claude Pro, Max and Ultra. The filing excludes these fixed-price subscribers from its net revenue retention metric because their spend does not reflect usage expansion.
Enterprise and team plans
Not disclosedContracted seats plus usage
The fastest-growing customer cohort: 11,900 organisations now spend at least $100,000 a year on a trailing basis, up from more than 950.
Cloud marketplaces
44% of 9M 2026 revenueBilled by the cloud provider, net of fees
Cloud service provider marketplaces handled 11%, 32%, 47% and 44% of revenue in 2023, 2024, 2025 and 9M 2026 respectively. The platform bills the end customer and remits cash net of platform fees.
Compute per $1 of revenue
$0.57
Gross margin
85.5%
Sales & marketing intensity
16.6%
Stock-based pay intensity
9.9%
Customer spending tiers
S-1/A| Tier | Latest | Year earlier |
|---|---|---|
$100,000+ annual spend (TTM) Reported at 30 September 2026 versus 30 September 2025. | 11,900 | 950 |
$1 million+ annual spend (TTM) Defined in the filing but the count is not stated in this amendment. | Not disclosed | Not disclosed |
$10 million+ annual spend (TTM) Defined in the filing but the count is not stated in this amendment. | Not disclosed | Not disclosed |
Net revenue retention The filing defines the metric but leaves the value blank in this amendment. | Not disclosed | Not disclosed |
What does this mean for investors?
Gross margin, where disclosed
S-1/A- Revenue
- Gross margin
What does this mean for investors?
Profitability simulator
HypotheticalSet how fast revenue grows and what share of it compute and other operating costs absorb. The model keeps cost ratios constant, which is the simplest honest assumption: it shows the operating leverage already visible in the filing.
Run your own profitability scenarios
Adjust growth, compute intensity and operating costs to see when the business breaks even.
Start free trialWhy costs behave the way they do
Nearly every dollar of cost is compute. In the first nine months of 2026, compute and infrastructure expense was $18.8B against revenue of $33.2B — 56.6% of the top line. That spend is largely pre-committed years in advance, which means gross margin is decided when contracts are signed, not when tokens are sold. It also means the model can look extremely profitable while revenue grows into fixed capacity, and extremely unprofitable if demand pauses. The full contractual picture is in the risk section.
Revenue for the twelve months to 2026-09-30 was $35.3B.
Valuation
What would it have to be worth?
No IPO price range has been set, so there is no official valuation to compare against. Build your own from the largest listed assumption in any technology valuation: revenue and the multiple the market pays for it.
The interactive valuation calculator
Model revenue growth, price-to-sales multiples, margins and dilution, then compare bear, base and bull outcomes.
Start free trialWhat has to be true for the base case
- Revenue compounds at roughly the scenario rate — far below the 1,206% of the last nine months.
- Compute stays available at contracted prices, so the 57% cost ratio does not drift back up.
- The $895.4B of commitments is funded by operating cash flow and existing liquidity, not by repeated equity issuance.
- The multiple holds: 15× sales is a premium software rating that assumes durable growth.
How to read a pre-IPO multiple
Price-to-sales is the only multiple that works here, because earnings are negative and will stay volatile while convertible notes are revalued. At $35.3B of trailing revenue, each turn of sales multiple is worth about $35B of valuation. That sensitivity is why the growth assumption matters more than any other input.
Why we do not show a current valuation
Competition
Who else is selling the same thing
Private AI labs publish no audited accounts, and the listed giants bury AI revenue inside larger segments. The table below says so rather than inventing numbers.
Competitor benchmarking
Compare Anthropic's disclosed growth and scale with the frontier labs and listed platforms it competes against.
Start free trialOnly Anthropic's row is sourced to an audited regulatory filing. Everything else is either a listed company that does not break out AI revenue, or a private company that publishes nothing. We update this table when a peer files or reports.
Ownership & governance
Who controls the company after the IPO
This is a controlled company. Buying Class A shares buys an economic interest, not a say in how Anthropic is run.
Share classes
| Class | Voting | Economic | Held by |
|---|---|---|---|
| Class A | 1 vote per share | Yes | Public investorsNot convertible. The stock being sold in the offering. |
| Class N | None, except as required by law | Yes | Strategic business partnersGenerally converts to Class A on transfer to an unaffiliated third party. |
| Class T | None on general matters | No | Long-Term Benefit TrustElects the Class T directors, who are a majority of the board. |
| Class E | Limited specific matters only | No | Employees excluding co-foundersIssued immediately before the offering. |
| Class F | 50.1% of total voting power | No | Founder LLC (seven co-founders)Single share; sunsets on defined triggering events. |
Who holds what
Founder LLC (co-founders)
Dario and Daniela Amodei plus five other co-founders are the sole initial members.
50.1% of total voting power
Class F
Long-Term Benefit Trust
Independent trustees with AI safety, national security and public policy expertise.
Elects a majority of the board
Class T
Employees (excluding co-founders)
Created immediately prior to the offering.
Limited matters only
Class E
Strategic business partners
Primarily strategic partners; converts to Class A on transfer.
None
Class N
Public investors
Detailed beneficial ownership percentages are blank in the amendment and will be set at pricing.
1 vote per share
Class A
Dilution: what to watch
Weighted-average shares rose from 1.53B in 2024 to 2.08B in the first nine months of 2026, and a tranche of RSUs settles around the offering. Add future grants, plus any equity raised to fund compute commitments, and per-share value grows more slowly than enterprise value. The valuation calculator applies a dilution input for exactly this reason.
The Long-Term Benefit Trust
A single Class T share, held by a trust of financially disinterested trustees, elects the directors who make up a majority of the board. A single Class F share, held by an LLC owned by the seven co-founders, carries 50.1% of total voting power. Class A shareholders vote on the remainder and cannot outvote either structure.
Why this is unusual
Risks
What could actually go wrong
Every risk below is drawn from the filing, written in plain English, and linked to the section that covers it. Severity is our judgement, not the company's.
Contracted commitments
Total $895.4B- Next 12 months$69.8B
- 2027 to 2030$377.7B
- 2031 to 2045$447.9B
What does this mean for investors?
Risk scorecard
Commitments dwarf liquidity
High severity
A short history of losses and cash burn
High severity
Channel and customer concentration
High severity
Competition and pricing pressure
Medium severity
Reported losses are distorted by convertible note accounting
Medium severity
Public shareholders get limited voting power
High severity
Regulation and litigation
Medium severity
Dilution from equity compensation
Medium severity
Commitments dwarf liquidity
High$895.4B of contracted spend — substantially all compute and infrastructure — sits against $131.9B of cash and investments. $69.8B falls due within twelve months and $447.5B before the end of 2030. If demand or pricing disappoints, the company must still pay, and may need to raise equity or debt on worse terms.
A short history of losses and cash burn
HighOperating losses were $551m, $2.98B and $8.06B in 2023, 2024 and 2025, and operating cash flow was negative in all three years. 2026 has flipped both trends, but one good nine-month period is not a track record, and the company warns it may not sustain its growth rate.
Channel and customer concentration
High44% of nine-month 2026 revenue reached customers through cloud service provider marketplaces, up from 11% in 2023, and those platforms bill end customers and remit cash net of fees. Losing or renegotiating a hyperscaler relationship would disrupt access and collections at once.
Competition and pricing pressure
MediumFrontier labs, hyperscalers and open-weight models all compete for the same enterprise budgets, and model quality advantages have historically been short-lived. Price per token has fallen industry-wide, so revenue depends on usage growing faster than prices decline.
Reported losses are distorted by convertible note accounting
MediumThe $104.0B nine-month net loss includes a $94.3B non-cash change in the fair value of convertible notes, derivative liabilities and SAFEs, driven by rising enterprise valuations. The operating loss was $1.3B. Investors should read the operating line, not the bottom line, but the notes are real liabilities and will need repaying or converting.
Public shareholders get limited voting power
HighA single share of Class F stock held by a founder-owned LLC carries 50.1% of total voting power, and the Long-Term Benefit Trust elects a majority of the board through Class T stock. Class A buyers are effectively minority partners in a controlled company with a public-benefit mandate.
Regulation and litigation
MediumAI regulation is still being written in the US, EU and UK, and the company discloses litigation and a $1.5B Bartz settlement in 2025. Compliance costs, usage restrictions or adverse judgments could change the economics of the core product.
Dilution from equity compensation
MediumStock-based compensation reached $3.3B in the first nine months of 2026, and a tranche of RSUs settles around the offering. Future grants, plus any equity raised to fund compute, dilute per-share economics.
AI research
Ask the filing
Answers are grounded in the figures held on this page and the S-1/A itself. Facts are cited; anything that is judgement is labelled as AI interpretation.
Ask GMG Pro AI
Grounded in the S-1/ALog in to ask the filing assistant. Free accounts get one question a session.
Unlock the full research desk
Advanced financial analysis, valuation models, competitor benchmarking and unlimited AI research across every company we cover.
IPO timeline
Where the process has got to
The journey from a public S-1 to a traded stock, updated as the filings land.
Amendment No. 1 (S-1/A) filed
2 Nov 2026Adds quarterly results through September 2026 and updated risk factors. Still marked 'Subject to Completion'.
FilingProposed IPO price range
TBAThe price range is left blank in the amendment, so no valuation or share price has been set yet.
Final IPO pricing
TBASet the night before listing, after the roadshow.
First day of trading
TBAClass A common stock proposed for the Nasdaq Global Select Market under the symbol ANT.
First results as a public company
TBAFirst quarterly report after listing.
IPO alerts
Get an email when the price range is set, when pricing is confirmed and when trading starts.
Create an account for alertsSources and accuracy
Figures are taken from Anthropic's Form S-1 (26 October 2026) and Amendment No. 1 (2 November 2026) as filed with the SEC. Reported figures are shown as filed; calculated figures are derived from them and labelled; estimates are labelled. The IPO price range, final pricing, post-offering share count and some ownership percentages are blank in the amendment and are shown as TBA or Not disclosed. Multiples used in the valuation tools are your assumptions, not the company's guidance.
Last refreshed 4 November 2026.
PRO