At 9:24 Monday morning Eastern, while traders were still positioning for the most crowded week of the fourth quarter, a 14-megabyte document landed on the SEC’s EDGAR servers and instantly rewrote the AI capital-markets calendar: Anthropic, PBC filed its S-1, registration number pending, CIK 0002130357, listing applied for on the Nasdaq Global Select Market under a three-letter ticker that says everything about the company’s ambitions — “ANT.” Bloomberg called it what it is: the first frontier AI lab to list. The AP went further: “potentially the biggest IPO ever” — a crown that, as our SpaceX GPU-debt coverage documented in June, currently belongs to SpaceX’s $85.7 billion debut at $135 a share.
The market voted immediately. The S&P 500 closed Monday at 7,676.77, up 0.70%, and the Nasdaq Composite gained 0.82% to 26,169.38 — Reuters’ closing wrap attributed the lift directly to the filing buoying AI shares ahead of Wednesday’s Fed decision. By Tuesday morning in Asia, Reuters was reporting stocks surging “as Anthropic IPO lifts sentiment ahead of earnings, central banks.” And on Polymarket, per Yahoo Finance, traders are now pricing a genuinely strange proposition: that Anthropic could be worth more than Bitcoin by year-end — which at BTC’s current ~$79,900 means a market cap north of $1.6 trillion.
This is the eighth chapter of the AI-capital supercycle series on this site (after SpaceX’s $40 billion GPU debt, Meta’s first negative free cash flow, Tesla’s 43% profit fall and the SpaceX-Tesla merger committees). Unlike those chapters, this one comes with a primary source you can read yourself: the S-1 discloses numbers no journalist had ever seen — $16.3 billion of revenue in six months, a headline net loss of $80.7 billion that turns out to be an accounting mirage, net revenue retention of approximately 900%, and a governance structure in which a single share hands the seven co-founders 50.1% of the vote. Here is everything the filing actually says, what it deliberately leaves blank, and — as always — what it means for people who earn independently.
What Was Filed Monday: Nasdaq “ANT,” a 20-Bank Syndicate, and the Blanks
The mechanics first. Anthropic registered (for fee-calculation purposes) $1 billion of Class A common stock, paying an $87,000 filing fee — the standard placeholder before pricing. The underwriting syndicate runs more than 20 banks deep, led by Goldman Sachs & Co., Morgan Stanley and J.P. Morgan, followed by Citigroup, Barclays, Wells Fargo, BofA, Deutsche Bank, RBC, UBS and a long tail through Jefferies, Mizuho, BTIG and Cantor. Reuters noted the deliberate spread: “Anthropic rewards firms with IPO roles as it eyes future legal risks” — the same coalition-building that Revolut’s IPO road required, at a hundredth the scale.
The offering is partly a primary raise (“general corporate purposes, including… investments in compute and infrastructure”) and partly a liquidity event: selling stockholders — employees with vesting RSUs — will sell Class A shares alongside the company, with IPO proceeds earmarked to cover the RSU tax withholding (the “RSU Settlement”). The company receives nothing from the employee sales.
Then there is what’s missing, which WIRED built an entire piece around (“Why Anthropic’s IPO Prospectus Is Missing So Many Key Figures”). The offering price range: blank. Q3 2026 revenue — for the three months ended September 30, when the S-1 says revenue went from $197 million in Q4 2024 “to $___”: blank. Run-Rate Revenue as of October 31, 2026 — a metric Anthropic itself defines (28 days of consumption fees × 13, plus monthly subscriptions × 12): blank. Total contractual obligations, “substantially all of which is related to compute and infrastructure”: blank. All of it gets filled in by amendment before pricing, but the pattern tells you this S-1 was rushed to file at the exact moment the market could absorb it — Monday of Fed week, two days before Microsoft and Alphabet report.
The Growth Story the S-1 Actually Tells
The revenue line is the most violent growth curve ever printed in an S-1. Anthropic’s own summary: $41 million (2023) → $386 million (2024, +841%) → $4.585 billion (2025, +1,088%) → $16.3 billion in the first six months of 2026 alone (+1,300% year-over-year), against $1.163 billion in H1 2025. The quarterly cadence is even steeper: $376 million in Q1 2025 to $11.6 billion in Q2 2026 — 30x in five quarters. And the company warns, in the same breath: “We do not expect our revenue will continue to grow at comparable rates, and we cannot assure you that it will continue to grow at all.”
| Period | Revenue | YoY Growth | Consumption / Subscription Split |
|---|---|---|---|
| FY 2023 | $41M | — | Substantially all consumption |
| FY 2024 | $386M | +841% | $288M (75%) / $98M (25%) |
| FY 2025 | $4.585B | +1,088% | $3.8B (83%) / $789M (17%) |
| H1 2025 | $1.163B | — | — |
| H1 2026 | $16.3B | +1,300% | $13.1B (80%) / $3.2B (20%) |
| Q2 2026 (quarterly) | $11.6B | +1,368% | Compute costs grew only +300% |
The customer engine underneath is equally absurd. As of September 30, 2026: more than 11,900 customers spending $100,000+ on a trailing-twelve-month basis (up from ~950 a year earlier — a 12.5x in twelve months), more than 1,900 above $1 million (Yahoo Finance: “nearly doubled since April”), roughly 250 above $10 million, and approximately 40 million customers below $100,000 — the prosumer long tail that includes most readers of this site. Over 50% of the Fortune 500 and over 30% of the Global 2000 now spend six figures with Claude. Net Revenue Retention — the expansion-churn composite — stood at approximately 900%. There is no SaaS company in history with a 900% NRR at $30 billion of annualized scale.
| Customer Tier (TTM revenue) | Count (Sep 30, 2026) | Trajectory |
|---|---|---|
| $10M+ | ~250 | The whale tier — no single customer >10% of revenue; top five = ~25% |
| $1M+ | 1,900+ | Nearly doubled since April 2026 (Yahoo Finance) |
| $100K+ | 11,900+ | ~400 (Mar 2025) → 6,000+ (Jun 2026) → 11,900+ (Sep 2026) |
| Under $100K | ~40 million | Prosumers, solo developers, Claude subscription base |
Two operating metrics explain why retention is 900%. First, the ecosystem hook TechCrunch zeroed in on: over 50% of $100K+ customers now use both the Claude API platform and a first-party app (Claude Code or the newer Claude Cowork, which extends the coding-agent architecture to general knowledge work) — up from ~30% at the end of last year. Second, the spend multiplier: in Q3, the highest-spending 5% of Claude Enterprise customers who used Claude Code spent about 20x as much per month as the average enterprise user who didn’t. Anthropic is explicit that this is strategy, not accident — it pitches Claude as “not a static product that serves a single use case,” runs an Applied AI organization to embed Claude inside customer stacks, and has launched a forward-deployed-engineer venture with Blackstone and other PE firms called Ode with Anthropic. The filing even says the company intends to spend less effort than a typical business “entrenching our current products” and more on “new classes of products whose aim is to dwarf the economic value of our existing ones.”
And it does all of this with roughly 5,300 employees (up from ~940 at the end of 2024). TTM revenue per employee went from $1.4 million to over $6.6 million in a single year — revenue grew ~1,200% while headcount grew ~170%. The S-1’s own engineers, it notes, now merge 45x more code per day than the company-wide average in 2024. That is the “talent density” argument, in numbers, for the first time in a public filing.
The $80.7 Billion Loss Is an Accounting Mirage — Here’s the Bridge
NBC’s headline — “Anthropic reveals $80 billion loss, huge spending spree ahead of IPO” — is technically true and substantively misleading, and understanding why is the single most useful finance lesson in this document. The net loss for H1 2026 was indeed $80.748 billion, or $(39.88) per share. But walk the bridge:
| H1 2026 Income Statement (in millions) | Amount | What it is |
|---|---|---|
| Revenue | $16,300 | Up 1,300% YoY |
| Total operating expenses | $(16,979) | Compute, personnel, sales, G&A (incl. $2.2B stock comp) |
| Operating loss | $(693) | The actual business: ~4% operating margin — nearly breakeven |
| Interest income | $1,111 | On the $131.7B cash pile |
| Change in fair value of convertible notes, derivatives, SAFEs | $(72,388) | Non-cash. The notes got more valuable because the company did |
| Other expense, net | $(8,757) | Strategic-investment marks, December 2025 acquisition derivatives, FX |
| Net loss | $(80,748) | The headline |
| Adjusted Operating Income (non-GAAP) | +$1,618 | 10% adjusted operating margin — positive |
| Adjusted Net Income (non-GAAP) | +$2,671 | vs. $(1,741)M adjusted loss in H1 2025 |
The $72.4 billion charge is a mark-to-market on convertible promissory notes — the IOUs Amazon, Google and others accepted instead of equity. As Anthropic’s enterprise valuation soared (culminating in the Series H), the notes became worth more to their holders, and GAAP makes Anthropic book that increase in value as an expense. The S-1 says so plainly: the changes were “primarily driven by an increase in the fair value of our outstanding convertible promissory notes resulting from an increase in our enterprise valuation.” Translation: the company lost $80 billion on paper because it became worth too much. Those notes convert to Class N stock at the IPO and the phantom bleeding stops.
The real trajectory is the adjusted line: from a $(522)M adjusted operating loss in 2023 to $(4.854)B in 2025 (a −106% margin — every revenue dollar cost two) to +$1.618B positive in H1 2026 at a 10% margin. Fortune’s summary — “$16.3 billion in revenue, an $80.7 billion net loss, and goal of an ‘age of miracles'” — captures the whiplash. The honest one-sentence version: Anthropic’s underlying business just crossed into profitability at the exact moment its GAAP loss became the largest half-year loss in corporate history.
The Compute Machine: $5.6 Billion a Quarter and $35 Billion in Big-Tech IOUs
What does frontier AI cost to run? The S-1 finally answers: compute and infrastructure expense rose every single quarter, from $944 million (Q1 2025) to $5.6 billion (Q2 2026) — an annualized run-rate above $22 billion, against $16.3 billion of half-year revenue. Anthropic procures compute 12 to 18 months in advance, so these costs are “generally fixed in the near term” regardless of demand. That is the same stranded-compute exposure Musk warned about for the neoclouds — with one difference: Anthropic’s efficiency curve is outrunning it. Q2 revenue grew 1,368% year-over-year against 300% compute-cost growth, and each model generation cuts prices: Claude Fable 5.1 (September 2026) reduced typical workload costs ~25% versus Fable 5 and up to 45% for complex coding and agentic tasks; Opus 5.5 cut cost per task ~40% versus Opus 5. Remaining performance obligations — contracted future revenue — stand at $10.8 billion, ~75% due within twelve months.
The funding stack behind it is a web of mutual dependency with the hyperscalers that has never been disclosed this precisely:
| Counterparty | Instrument | Size / Terms | Status |
|---|---|---|---|
| Amazon / AWS | 2023 + 2024 Note Purchase Agreements | ~$6.2B aggregate principal (2023 NPA: $1.3B Nov ’24, $1.3B May ’25, $1.4B Nov ’25), 8% interest, 10-year | Converted to preferred; conversion capped so Amazon never exceeds 12% of capitalization without waiver |
| Series H equity (May 2026) | $5.0B | Funded | |
| Amazon 2026 NPA | $15.0B remaining, drawn in $5.0B increments tied to compute-capacity milestones; notes at 8%, 10-year | Available until 30 months post-IPO | |
| Series H equity (May 2026) | $10.0B | Funded | |
| Google 2026 NPA (March 2026) | $20.0B remaining; notes at 9%, 6-year | Available in tranches | |
| All investors | Series H-1 preferred, May–June 2026 | 77,637,878 shares at $589.0095 = $45.73B raised | Last private mark |
Total capital raised since inception: $141.1 billion. Cash and equivalents at June 30: $112.4 billion, plus $19.3 billion of short-term investments — a $131.7 billion treasury that threw off $1.1 billion of interest income in six months. And still the risk factors concede the dependency: Anthropic “has in the past sought quantities of compute from CSPs in excess of what those partners were willing to provide,” and Amazon, Google and Microsoft “may prioritize their own requirements or rivals they’ve invested in.” Nearly half the revenue (48% in H1 2026, up from 11% in 2023) flows through those same clouds’ marketplaces, which bill customers and remit net of fees — and 58% of accounts receivable sits in those channels. The landlords are also the distributors, the creditors, and (via OpenAI at Microsoft, Gemini at Google) the competition.
What the Risk Factors Confess
Every S-1 is also a confession booth, and this one is unusually candid. Five disclosures deserve attention:
1. The month-long training pause. Reuters flagged it from the filing: Anthropic “recently chose to delay pre-training a new frontier base model for more than a month to allow time to further enhance our security measures.” A company whose product cycle is its revenue growth paused the machine for security work — and told the SEC it may do so again (“we may delay or pause a stage when we are not confident in our safety measures”).
2. Safety as legal architecture. The word “safety” appears exactly 250 times in the document (we counted; Forbes noticed too, alongside the CEO’s personal-security costs — the board approved enhanced protection for both Dario and Daniela Amodei in 2025). The company is a Delaware public benefit corporation whose charter purpose is “the responsible development and maintenance of advanced AI for the long-term benefit of humanity,” and the filing warns that pursuing that purpose “may, in any given period, reduce our revenue, increase our expenses… and reduce our profitability.” No IPO has ever asked public shareholders to underwrite an existential-risk thesis this explicitly — AP’s headline paired the record-size filing with “CEO warns of risk to humanity from AI.”
3. Concentration everywhere. Top five customers ≈ 25% of nine-month revenue (though no single customer exceeds 10%); CSP marketplaces = 48% of revenue; compute supply controlled by three companies that compete with it; data centers “primarily in the United States,” dependent on long-lead-time power deals, with the filing acknowledging “public controversy and regulatory scrutiny regarding environmental impact, land and water use, and the allocation of limited power resources.”
4. The debt cross-default web. Anthropic carries convertible notes plus an amended-and-restated credit agreement, and warns that a default under “certain commercial agreements” (read: the hyperscaler compute deals) could trigger cross-defaults accelerating “substantially all” of its indebtedness. The compute it locks in today, it adds, “may become less efficient or economically attractive relative to future alternatives” — the stranded-asset risk in the company’s own words.
5. The blanks. Q3 revenue, October run-rate revenue, the offering range, the total compute obligations — all deferred to an amendment. WIRED’s skepticism is fair: the most anticipated IPO since SpaceX is asking you to commit before the last four numbers exist.
The Governance Fine Print: One Share, 50.1%, and a Trust You’ve Never Heard Of
CNBC’s read — “Anthropic’s IPO puts power in the hands of founders, giving everyday shareholders little voice” — undersells how deliberately the machine is engineered. There are five classes of stock:
| Class | Who holds it | Voting power | Economics |
|---|---|---|---|
| A | You — the IPO buyer | 1 vote per share | Full |
| N | Strategic partners (notes convert here at IPO; Amazon capped at 12%) | None | Full |
| T | The Long-Term Benefit Trust (1 share, after a 1-for-1,000 reverse split on Sep 17, 2026) | Elects the Class T directors — a majority of the board | None |
| F | The Founder LLC — Dario Amodei, Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish, Christopher Olah | A fixed 50.1% of all voting power, from a single share | None |
| E | Employees (excluding co-founders) | Limited; cannot elect directors | Full |
Stack the layers: the Long-Term Benefit Trust — five financially disinterested trustees with backgrounds in AI safety, national security, public policy and social enterprise — elects and can remove the majority of the board through that single Class T share. The founders separately hold 50.1% of general voting power through one Class F share with zero economic rights, and amending the trust against the trustees’ wishes requires, until the third anniversary of the IPO, the unanimous Continuing Founders plus employee-class majorities. A public Class A holder voting against everything controls roughly nothing. The one softening gesture: the co-founders “have each pledged to dedicate 80% of their personal equity in Anthropic to philanthropic causes” — though the filing is scrupulously honest that these pledges are “voluntary, revocable expressions of philanthropic intent” and “do not constitute legally binding obligations.”
Is it anti-investor? It’s actually consistent: you are buying the economics of the fastest-growing revenue base in software history, explicitly without control, from a company whose charter subordinates profit to its AI-safety purpose. Know which of those three facts you’re underwriting. The lock-up mirrors the unusual structure: a 180-day market standoff with early-release tranches — 7% at day 62, 7% at day 76, another 12.5% after the FY26 earnings release — all waivable by Goldman and Morgan Stanley.
Sizing ANT: Aramco, SpaceX and the $1.8 Trillion Napkin
“Potentially the biggest IPO ever” (AP) needs context, because this timeline has moved the goalposts twice:
| IPO | Raised | Valuation at listing | Year |
|---|---|---|---|
| Alibaba (NYSE: BABA) | $25.0B | ~$231B | 2014 |
| Saudi Aramco (Tadawul) | $25.6B ($29.4B w/ greenshoe) | ~$1.7T | 2019 |
| SpaceX (Nasdaq: SPCX) | $85.7B at $135/share — current record | Peaked at $225.64 within days; Musk became the first trillionaire (Forbes) | Jun 2026 |
| Anthropic (Nasdaq: ANT) | TBD — price range still blank | Series H mark (May–Jun 2026): $589.0095/share | Filing: Oct 26, 2026 |
The napkin math on valuation: ~2.02 billion weighted-average common shares (the H1 EPS denominator) plus 1,040,440,652 preferred shares outstanding at September 30 — call it ~3.06 billion as-converted — times the Series H price of $589.0095 lands around $1.8 trillion, before a single public trade and before any IPO pop. That is why the Polymarket crowd is pricing “Anthropic worth more than Bitcoin by year-end”: BTC’s market cap at ~$79,900 is roughly $1.6 trillion, and bitcoin has been sliding while ANT hasn’t even priced. If Anthropic raises anything like SpaceX’s $85.7 billion, it takes the record; if it raises even $40 billion, it becomes the second-largest IPO in history while selling roughly 2% of the company. The Financial Times, characteristically, begs to differ with the whole genre: “Why you shouldn’t bother reading the Anthropic IPO filing.” The NYT split the difference — “Promise and Peril.”
The trickle-down is already visible beyond the ticker. ABC7 in San Francisco reports the IPO is “expected to reshape San Francisco’s housing market,” quoting employees past and present: “Most of us can’t afford houses” — a suddenly-liquid RSU base colliding with Bay Area real estate is the CNBC “massive trickle-down effect” story in one sentence. And the wealth effects don’t stop at Anthropic’s cap table: Amazon’s 12%-capped stake, Google’s $10 billion Series H position, and the 20+ underwriters’ fee pools all reprice on the day ANT trades.
The Week It Lands: Fed Wednesday, Hyperscaler Earnings, PCE Thursday
The filing didn’t pick Monday by accident. This is the calendar we mapped in Sunday’s week-ahead preview: the FOMC meets today and tomorrow, deciding Wednesday at 92% hold odds (the market’s verdict since October hike odds collapsed from 54% to 15%); Microsoft and Alphabet report Wednesday night; Amazon and core PCE land Thursday. Two of those three companies are Anthropic’s landlords, creditors and distributors — their capex lines on Wednesday and Thursday are Anthropic’s compute budget. Gold steadied above $4,000 Monday (futures settled ~$4,038) after the two-month low, and Bloomberg’s morning note had it “steadying with Fed decision in focus.” An S-1 of this magnitude, filed into the single most crowded macro week of Q4, is either supreme confidence or perfect camouflage. Probably both.
The 9-Move Freelancer & Digital Nomad Playbook
Same as every week: what this actually means for people who earn independently, hold their own portfolios, and can’t outsource the thinking to an institution.
Move 1: Respect the blanks — no day-one market orders on ANT
The price range, Q3 revenue, October run-rate and total compute obligations are all still “$___” in Monday’s filing. They arrive in an S-1/A amendment before pricing — read it (it takes ten minutes; Move 9 shows you how). Then remember the lock-up ladder: 7% of insider shares unlock at day 62, another 7% at day 76, 12.5% after the FY26 earnings release, with Goldman and Morgan Stanley able to waive everything early. If you want exposure, size it in tranches against that supply calendar, not against the headline. The rules-over-moods framework from our all-time-highs guide applies verbatim: decide your limit levels before the ticker exists in your brokerage app.
Move 2: Learn the loss-bridge trick on a document that matters
Eighty billion dollar “loss,” $1.6 billion adjusted profit — the difference is a non-cash fair-value mark on convertible notes that converts away at the IPO. Every future mega-listing (OpenAI is next; xAI won’t be far) will ship with some version of this distortion: SAFE marks, warrant derivatives, earnout liabilities. The skill transfers directly to reading any 10-K in your portfolio: find the cash operating line first, then ask what the non-cash noise is marking to. When a company’s losses grow because its valuation grows, that’s a footnote, not a red flag. When losses grow while valuation stagnates — that’s Meta’s negative-free-cash-flow quarter all over again.
Move 3: Reprice your AI cost stack every quarter — the deflation is contractual now
Fable 5.1 cut typical workload costs ~25% (up to 45% for coding/agentic) versus its predecessor; Opus 5.5 cut cost-per-task ~40%. This isn’t promotional — it’s in an SEC filing, which means it’s a margin strategy: Anthropic lowers cost-per-task to grow usage, and usage compounds at 900% retention. For you: any fixed monthly AI spend you set more than a quarter ago is overpriced. Re-benchmark your tools, and if you sell fixed-price deliverables built on AI (content systems, code audits, data pipelines), your COGS is falling while your prices don’t have to. That margin gap is the cleanest tailwind a solo operator gets this year.
Move 4: Sell the 20x, not the hour
The most actionable number in 14 megabytes: Claude Code power users in the top 5% of enterprise spend consume 20x the monthly dollars of average enterprise users. Deployment, not access, is the product. The “Ode with Anthropic” model — forward-deployed engineers embedding Claude into company stacks, backed by Blackstone money — is a franchise blueprint for one-person consultancies: pick a vertical, productize a deployment package (intake → agent workflows → training → handoff), charge per outcome. Our AI-proofing playbook made the defensive case in the spring; the S-1 just published the offensive one.
Move 5: Audit your CSP and AI-lab exposure — concentration cuts both ways
48% of Anthropic’s revenue rides through AWS/GCP/Azure marketplaces; 58% of its receivables sit in those channels; its risk factors admit the clouds “may prioritize their own requirements or rivals they’ve invested in.” If your freelance pipeline is AI-lab-adjacent — marketplace listings, agency overflow from Anthropic/OpenAI ecosystem work, neocloud clients — you are downstream of the same chokepoint. Check receivables aging (net-30 stretching to net-60 is the tell), keep six months of runway, and remember Musk’s “guppy-level death” forecast for 200+ neoclouds: consolidation at the infrastructure layer is an accounts-payable event at every vendor above it.
Move 6: Follow the enterprise budget migration
~250 customers now spend $10M+/year with Anthropic; the top five average ~$1.6B of annual commitment each. Money like that doesn’t appear — it’s reallocated from SaaS seats, outsourcing contracts and junior-headcount budgets. If your clients are mid-market companies, the pitch that wins in 2027 is “I make the AI spend you already committed to actually work” — integration, workflow redesign, team training. If your clients are the SaaS vendors and agencies being displaced, diversify now, before the renewal cycle prices it in.
Move 7: Sequence around the macro cluster — nothing big until Friday
Fed Wednesday (92% hold, but the statement language on 2027 is the real event), Microsoft + Alphabet Wednesday night, Amazon + core PCE Thursday, and the long end of the Treasury curve knocking on 5% — which is the number that actually prices your debts. Self-employed and mortgage-shopping? The lock-versus-float logic in our self-employed mortgage guide hasn’t changed. Don’t make irreversible portfolio or pricing decisions inside a 72-hour window this dense.
Move 8: Realize you already own the picks and shovels
You can’t buy ANT yet — but if you hold QQQ or any broad index, you already own Microsoft and Alphabet (Wednesday’s earnings), Amazon (Thursday’s), Nvidia (the silicon under the $22B compute run-rate), and the SpaceX that closed $40 billion of GPU debt at 8%+ last week. The Anthropic S-1 is, among other things, a forward guidance document for all of them: $15B more from Amazon and $20B more from Google remain committed to its compute buildout. Before you chase the IPO, audit what your existing funds already express about it — most portfolios are more Anthropic-exposed today than their owners think.
Move 9: Use the 14-megabyte free textbook
The S-1 is public, free, and the single best document ever written about how a frontier AI company actually makes money: EDGAR → search CIK 2130357 → anthropicpbc-sx1.htm. Two hours in the Business section (pp. 178+) and the Contractual Obligations table will teach you more about the AI economy than a year of newsletters. Read it with WIRED’s critique open alongside — noticing what a company chooses not to fill in is half of financial literacy. Practice now, because the OpenAI filing is coming, and the people who can read an S-1 cold are the same people who get the next 900%-NRR story at the right price.
FAQ
1. When does Anthropic’s IPO actually happen, and what’s the ticker?
No date yet — Monday’s S-1 is the initial filing, with the price range still blank. Expect an amended filing (S-1/A) after this week’s Fed decision and hyperscaler earnings fill in the macro picture, then a roadshow and pricing on the usual 4–8 week timeline. Listing: Nasdaq Global Select Market, symbol “ANT.” If the standard clock holds, first trade lands somewhere in Q4 2026 or very early Q1 2027.
2. Did Anthropic really lose $80.7 billion?
On GAAP, in H1 2026, yes — $(80.748) billion, or $(39.88) per share. But $72.4 billion of that is a non-cash mark on convertible notes that got more valuable because Anthropic’s valuation rose, and another $8.8 billion is other non-operating marks. The operating loss was $693 million on $16.3 billion of revenue (~4% margin), and adjusted operating income was positive $1.618 billion (10% margin) with adjusted net income of $2.671 billion. The notes convert to stock at the IPO; the phantom loss ends there.
3. How much revenue does Anthropic make?
$41M in 2023, $386M in 2024, $4.585B in 2025, and $16.3B in H1 2026 alone — including $11.6B in Q2 2026. Q3 2026 revenue is one of the blanks to be filled by amendment. The mix is ~80% consumption-based (API tokens) and ~20% subscriptions (Claude plans), with 48% of revenue billed through AWS/GCP/Azure marketplaces.
4. What will Anthropic be worth?
Unknown until pricing, but the last private mark (Series H-1, May–June 2026) was $589.0095 per share — $45.7 billion raised in that round alone, including $5B from Amazon and $10B from Google. Rough as-converted math (~3.06 billion shares × $589) implies ~$1.8 trillion. Polymarket traders are betting it exceeds Bitcoin’s ~$1.6 trillion market cap by year-end. For scale: that would make ANT a top-five company on Earth before its first earnings call.
5. Will public shareholders have any control?
Effectively none. Class A shares (what you buy) carry one vote each; a single Class F share held by the seven co-founders’ LLC carries a fixed 50.1% of all voting power; and a single Class T share held by the five-trustee Long-Term Benefit Trust elects the majority of the board. Anthropic is a Delaware public benefit corporation whose charter purpose is the responsible development of AI “for the long-term benefit of humanity” — the filing explicitly warns that purpose may reduce revenue and profitability. You are buying economics, not governance.
6. How is this different from SpaceX’s June IPO?
Scale and structure. SpaceX raised a record $85.7B at a $135 price with straightforward (if founder-heavy) governance and $5.2B of quarterly Starlink revenue underneath it. Anthropic’s raise is still unpriced, but its governance is a five-class-stock contraption with a trust at the top, its revenue is growing faster (30x in five quarters), its losses are larger in absolute terms (and more accounting-driven), and its compute supply is controlled by three companies that also compete with it. SpaceX IPO’d into the AI-debt boom; Anthropic is IPO-ing as the boom’s stress test — Reuters’ Instant View literally titled it “putting the AI investment boom to the test.”
7. Should freelancers and digital nomads care beyond the stock?
Yes — three ways. Your tools get cheaper on a contractual cadence (each model generation cuts task costs 25–45%). Your competitive bar rises (the 20x Claude Code stat means deployment skill, not tool access, is what gets paid). And your clients’ budgets are moving (250 enterprises spending $10M+/year each is reallocated from the SaaS and services spend that funds freelance invoices — position on the right side of that migration).
Bottom Line
Anthropic’s S-1 is the first time the AI economy’s engine room has been audited in public, and the numbers justify both the euphoria and the skepticism in the same paragraph: $16.3 billion of half-year revenue growing 1,300%, 900% net retention, a $131.7 billion treasury and a first-ever adjusted operating profit — wrapped around an $80.7 billion accounting loss, $22 billion of annualized compute spend committed 18 months in advance, three landlord-creditor-competitors holding the supply line, and a governance structure that gives public money 100% of the exposure and approximately 0% of the control. The market’s opening vote — Nasdaq up, Asia up, Polymarket pricing ANT above Bitcoin — says investors will take that trade. Wednesday’s Fed decision and Wednesday-night Microsoft/Alphabet earnings say whether the macro lets them. Watch the blanks get filled; that’s where the real price discovery lives.
Sources
- SEC EDGAR — Anthropic, PBC Form S-1, filed October 26, 2026 (CIK 0002130357, accession 0001628280-26-067168): revenue, loss bridge, adjusted measures, customer tiers, NRR, compute expenses, CSP concentration, Series H terms, Amazon/Google NPAs, share classes, LTBT, lock-up, risk factors, underwriters, “safety” count (250)
- SEC EDGAR — Anthropic EX-FILING FEES exhibit: $1B placeholder registration, $87,000 fee, Class A par $0.00001
- TechCrunch — “Anthropic’s IPO filing reveals its strategy for keeping enterprise customers hooked,” Oct 26, 2026: 50%+ dual-product adoption, 20x Claude Code spend stat, Ode with Anthropic/Blackstone, joint go-to-market, “reinvent our business” quotes
- Reuters — “Instant View: Anthropic IPO filing to put AI investment boom to the test”; “Anthropic’s soaring sales, steep losses put AI frenzy to IPO test”; “Anthropic discloses month-long training pause, AI security breaches in IPO filing”; “Anthropic rewards firms with IPO roles as it eyes future legal risks”; “Equities end higher as Anthropic’s IPO filing buoys AI stocks ahead of Fed”; “Asian stocks surge as Anthropic IPO lifts sentiment ahead of earnings, central banks,” Oct 26–27, 2026
- AP News — “Anthropic files for potentially the biggest IPO ever as CEO warns of risk to humanity from AI,” Oct 26, 2026
- Bloomberg — “Anthropic Files for IPO to Become First Frontier AI Lab to List”; “Anthropic’s IPO Sets Up Defining Moment for Wall Street and AI Boom”; “Gold Steadies Above $4,000 With Fed Rate Decision in Focus,” Oct 26, 2026
- Fortune — “Anthropic’s IPO prospectus shows $16.3 billion in revenue, an $80.7 billion net loss, and goal of an ‘age of miracles’,” Oct 26, 2026
- NBC News — “Anthropic reveals $80 billion loss, huge spending spree ahead of IPO,” Oct 26, 2026
- CNBC — “Anthropic’s IPO puts power in the hands of founders, giving everyday shareholders little voice”; “From Wall Street to Silicon Valley, Anthropic IPO poised to have massive ‘trickle-down effect’,” Oct 26, 2026
- WIRED — “Why Anthropic’s IPO Prospectus Is Missing So Many Key Figures,” Oct 26, 2026
- Forbes — “10 Surprises Revealed In Anthropic’s IPO Filing: CEO’s Personal Security Costs, ‘Safety’ Mentioned 250 Times And More,” Oct 26, 2026
- Yahoo Finance — “Anthropic Could Be Worth More Than Bitcoin by Year-End, Polymarket Traders Say”; “Anthropic IPO Filing Shows Customers Paying Over $1M A Year Has Nearly Doubled Since April,” Oct 26, 2026
- ABC7 Bay Area — “Anthropic’s IPO expected to reshape San Francisco’s housing market: ‘Most of us can’t afford houses’,” Oct 27, 2026
- Financial Times — “Why you shouldn’t bother reading the Anthropic IPO filing,” Oct 26, 2026; The New York Times — “The Key Numbers From Anthropic’s IPO Filing” / “Anthropic’s I.P.O. Prospectus Spells Out a Future of Promise and Peril,” Oct 26, 2026; WSJ — “The Biggest Takeaways From Anthropic’s IPO Prospectus,” Oct 26, 2026
- Yahoo Finance chart API — S&P 500, Nasdaq Composite, Dow closes Oct 23–26, 2026; BTC-USD and GC=F levels Oct 27, 2026 (author’s own pulls)
- finelypick.com prior coverage — SpaceX $40B GPU debt (post #284, incl. June 12 SPCX IPO record table), Meta Class C / negative FCF (#274), Tesla Q3 (#272), SpaceX-Tesla merger (#270), week-ahead macro calendar (#278), Warsh Fed / Polymarket odds (#260), Revolut IPO road (#235), bitcoin break below $81,000 (#239)