Anthropic has never sold an advertisement, and the document it filed on Monday gives no sign that it intends to. The company behind Claude submitted a registration statement on Form S-1 to the Securities and Exchange Commission on October 26, 2026, seeking to list Class A common stock on the Nasdaq Global Select Market under the ticker "ANT". The filing reports revenue of $16.286 billion for the six months to June 30, 2026, against $1.163 billion a year earlier, earned from metered usage of its models and from subscriptions. There is no ad tier, no sponsored answer and no audience product in it.
That absence is the obvious headline for a marketing audience. It is also slightly misleading. Read through its exhibits, the prospectus describes a company sitting unusually close to the centre of the advertising economy, only from the other side of the ledger. Its two largest compute suppliers, Google and Amazon, run two of the biggest advertising businesses in the world. Both hold stakes in it, both resell its models through their cloud marketplaces, and in the first half of 2026 Amazon booked more income from its Anthropic investment in a single quarter than it earned from advertising. Anthropic does not sell ads; its rise has been reported, in part, inside the earnings of companies that do.
Six of the publications PPC Land readers follow carried something bearing on Anthropic on Monday. PPC Land published four articles on the filing in the late afternoon, UTC. Adweek and MediaPost each ran their own reading of it. PPC News Feed and AdExchanger covered the tools and the measurement problems that sit downstream of Claude, and Search Engine Roundtable's daily recap pointed to a fresh lawsuit. Taken together, they describe a business that has grown roughly fifteen-fold in a year without a single advertiser, and a set of commitments that will take a decade or more to pay off.
Luis Rijo
What the filing shows
PPC Land reported at 16:31 UTC on October 26 that Anthropic files for Nasdaq IPO with $16.3bn in first-half revenue. The registration, file number 333-299633, leaves the price range and the number of shares blank. Its revenue history is short and steep: $41 million in 2023, the first year of commercial sales, $386 million in 2024 and $4.585 billion in 2025. The first quarter of 2026 brought $4.731 billion and the second $11.555 billion.
Adweek's Trishla Ostwal, publishing at 12:02 Eastern time the same day, reported that Anthropic goes for the public markets as revenue jumps nearly 15-fold. Second-quarter revenue of $11.56 billion compared with $787 million a year earlier. The filing puts the growth in plain terms: revenue "grew 1,368% year-over-year for the three months ended June 30, 2026" while compute and infrastructure expenses "grew 300%". Those costs fell to 48% of revenue from 251% in the first quarter of 2025. Operating income in the second quarter was $974 million, the only profitable quarter among the results Anthropic disclosed.
The half-year picture is less tidy. PPC Land put the operating loss for the first six months of 2026 at $693 million, down from $2.135 billion a year earlier; set against Adweek's second-quarter profit, that implies a first-quarter operating loss of roughly $1.67 billion. The net loss for the half was $80.748 billion, of which $72.388 billion was a non-cash change in the fair value of convertible notes, derivatives and simple agreements for future equity. The 2025 net loss was $41.965 billion. Anthropic also presents a non-GAAP adjusted operating margin of 10%, which excludes stock-based compensation; that compensation rose to $2.2 billion in the half, from $286 million, for a headcount of about 5,300.
The company is not short of cash. It held $131.651 billion at June 30, 2026, made up of $112.4 billion in cash and $19.3 billion in short-term investments, and had raised $141.1 billion in total by the same date. "Our historical growth rates are not indicative of future performance," the prospectus warns, in the sentence every growth company includes and few have had as much reason to.
Some of the most watched numbers are missing. The filing gives third-quarter revenue only as a multiple, 86 times the fourth quarter of 2024, which PPC Land converts to about $16.9 billion. Run-rate revenue as of October 31, 2026 is left blank, though the definition is printed: trailing 28-day consumption fees multiplied by 13, plus monthly subscription fees multiplied by 12. MediaPost's Laurie Sullivan reported on June 1, when Anthropic first filed a draft IPO registration, that the annualised run rate had reached $47 billion, up from $9 billion at the end of 2025, with more than 80% of revenue coming from enterprise clients.
The customer base explains where that money comes from. As of September 30, 2026, more than 11,900 customers generated over $100,000 each in trailing revenue, up from more than 950 a year earlier, Adweek reported. More than 1,900 paid over $1 million and about 250 over $10 million. Roughly 40 million customers each spend less than $100,000. Net revenue retention was about 900%, a figure PPC Land notes excludes Claude Pro and Max subscribers. Adweek found that the top five customers accounted for about 25% of revenue in the nine months to September 30 and that no single customer exceeded 10%; PPC Land flagged another reading of the filing in which two customers made up about one-fifth of revenue, and said the accounts may not be fully contradictory.
What investors will pay remains a matter of press reports rather than disclosure. The Wall Street Journal reported, as relayed by Adweek, that Anthropic could raise $100 billion or more at a valuation of $1.5 trillion to $2 trillion, with a listing targeted for the week of November 16. The fee table in the filing uses a placeholder maximum offering price of $1 billion, on which Anthropic paid a registration fee of $87,000. Twenty-two underwriters are named, led by Goldman Sachs and Morgan Stanley.
The road here has been quick. Adweek reported on June 1 that Anthropic files for IPO as AI race hits public markets, days after a $65 billion Series H round at a $965 billion post-money valuation. A week later, on June 8, Adweek reported that OpenAI confidentially filed for an IPO, saying it "may be a while" before any offering. Monday's Adweek piece notes that OpenAI is not expected to list until 2027, which would make Anthropic the first standalone American frontier AI lab to trade publicly. Little more than a year ago, in September 2025, PPC Land reported that Anthropic completed a $13 billion Series F at a $183 billion valuation. The Series E, in March 2025, had valued it at $61.5 billion.
A business with no ad tier
MediaPost's editor in chief, Joe Mandese, searched the 557-page document for an advertising business and reported on October 26 that the Anthropic IPO omits any plans for an ad model. Advertising is not listed as a potential business model. The word appears, as does "media", but mostly in reference to Anthropic's own marketing rather than to an ad-supported product. Mandese ran keyword counts: "agent" was too frequent to tally, and among the words he selected, "human" ranked highest, followed by "marketing", "agentic", "autonomous" and "subscription". The filing never uses the word "extinction".
There is a small irony in the method. To summarise the business model, Mandese used Google's Gemini, which described two revenue lines: consumption-based metering, in which enterprise customers and developers pay for Claude usage measured in tokens, and fixed monthly subscriptions such as Claude Pro and Claude Max. The prospectus says nothing about serving ads, monetising user data through ad networks or offering an ad-supported tier.
That is consistent with what Anthropic has said publicly. On February 4, 2026, the company pledged that Claude would remain free of advertising on every subscription tier, and PPC Land reported the next day that Anthropic dropped millions on a Super Bowl ad savaging OpenAI's ChatGPT ad pivot. The four commercials, two of them set to air during Super Bowl 60 on February 8, carried the line "Ads are coming to AI. But not to Claude." OpenAI's chief executive Sam Altman responded on X that evening, and on February 5, in a live interview, called Anthropic's Super Bowl ads dishonest. OpenAI went the other way: PPC Land's timeline in Monday's coverage records ChatGPT ads launching in the United States on February 9 and passing a $1 billion annualised run rate by August 31.
The pledge does not mean Anthropic stays away from paid media. Laurie Sullivan noted in MediaPost on March 27 that Anthropic ran a connected TV ad about Claude that month. What the filing shows is that its commercial spending has gone somewhere else. Sales, marketing and partnerships expense reached $3.168 billion in the first half of 2026, up from $120 million a year earlier. PPC Land attributes most of that increase to a $2.7 billion rise in fees paid to channel partners rather than to advertising. Anthropic records sales through the cloud marketplaces gross and books the providers' cut as an expense, so much of the line where a media budget would normally sit consists of fees to the partners that resell Claude.
The consumer numbers help explain why advertising was never the obvious route. Emarketer, cited by MediaPost in June, forecast that 5.4% of US internet users would use Claude in 2026, against 36.6% for ChatGPT and 27.4% for Gemini. An ad business needs reach. Anthropic built a revenue line larger than many ad businesses without it, by selling to companies that pay by the token. PPC Land estimates annualised revenue of roughly $46 billion, none of it from advertising.
The landlords sell ads
The second PPC Land piece, published at 16:49 UTC, followed the money the other way. It reported that Amazon, Google and Microsoft hold $269bn of Anthropic compute contracts. The commitments are at least $114.5 billion to Google Cloud, running from April 2026 to July 2033 and payable "regardless of actual usage"; at least $110.0 billion to Amazon, from May 2026 to April 2036; and at least $44.9 billion to Microsoft Azure, from November 2026 to May 2033. The $269.4 billion sum of those minimums excludes about $161.2 billion of TPU-based equipment leases from Broadcom, up to $120.5 billion from SpaceXAI, a SpaceX subsidiary, and the cloud fees already mentioned.
How large the full bill is depends on who is counting. The filing describes total purchase commitments as "more than $" and leaves the number blank. Yahoo Finance's Jared Blikre summed four non-cancellable lines to $430.6 billion; Cape Fear Advisors added SpaceXAI, Nscale and Lambda to reach $637.0 billion; ZeroHedge, including AMD and future TPU estimates, put it at about $780 billion to $810 billion; and Reuters, cited by Investing.com, used a narrower accounting measure of about $245 billion plus about $21 billion in short-term obligations. Adweek reported that Anthropic expects to spend another $125 billion to $150 billion on Google TPU capacity from 2028 to 2031, and that the Journal puts total compute obligations above $475 billion. Investing.com warned that the figures cover different agreements and should not simply be added.
"We manage compute as a single, fungible pool from multiple strategic partners," the prospectus says. The partners are also, in nearly every case, investors, sales channels and competitors. Claude is "accessible on all three major cloud platforms," and sales through the Amazon, Google and Microsoft marketplaces made up about 48% of first-half revenue, by coincidence the same proportion as the compute cost ratio Adweek reported, or roughly $7.8 billion on PPC Land's arithmetic, up from 32% in 2024 and 47% in 2025. Marketplace sales also accounted for 58% of net receivables at June 30. A large share of Anthropic's customers, in other words, pay the hyperscaler, which pays Anthropic, which pays the hyperscaler again for compute.
For the advertising market, the more striking numbers sit in the suppliers' own accounts. Amazon invested $8 billion in Anthropic convertible notes between the third quarter of 2023 and the fourth quarter of 2025, and booked a $16.8 billion non-cash gain in the first quarter of 2026 when part of them converted. In the second quarter it reported $53.4 billion of non-operating other income, "primarily from the company's investments in Anthropic." On PPC Land's calculation, that was more than two and a half times Amazon's $19.8 billion of advertising revenue in the same quarter. As far back as October 2025, PPC Land reported that Amazon advertising hit $17.7B as AWS accelerated with an Anthropic AI cluster, when Project Rainier came online with nearly 500,000 Trainium2 chips and Amazon recorded a $9.5 billion pretax gain on its stake.
Alphabet's position is murkier because it does not tie its gains to specific holdings. Its second-quarter net income included $77.1 billion of net unrealised equity gains, in a quarter when Google Search ads generated $63.3 billion; first-quarter other income was $37.7 billion. Analyst estimates of Alphabet's Anthropic stake range from 10% to 15%. The same quarter brought capital expenditure of $44.9 billion and negative free cash flow of $5.9 billion. PPC Land's argument is that a growing share of the headline earnings at both companies reflects Anthropic's valuation rather than advertising, and that these non-cash gains can reverse once the market puts a price on the stock.
Microsoft agreed in November 2025 to invest up to $5 billion, and Nvidia up to $10 billion in the same month; Nvidia GPUs sit behind the SpaceXAI, Nscale and Lambda capacity, about $206.4 billion combined. SpaceXAI alone runs to about $3.1 billion a month. Meta has no disclosed equity or compute link at all, but in April it opened Ads AI Connectors that let Claude create and edit campaigns through its ads server, as PPC Land reported when Meta opened its ad system to Claude and ChatGPT on April 29, 2026. Apple's only documented connection is as a participant in Project Glasswing.
The concentration has drawn regulators before. In July, France's competition authority found that OpenAI, Google and Anthropic together held more than 84% of the global AI agent sector in May 2026, PPC Land reported when it covered how France flags lock-in risk as OpenAI, Google and Anthropic hold 84% of AI agents. The S-1 adds a less visible kind of concentration, in which the lab's suppliers and its rivals are the same few firms.
Paper instead of cash
The fourth PPC Land article, published at 17:57 UTC, went through the exhibits. It reported that Anthropic IPO exhibits show 9% interest on Broadcom's $42bn chip notes. The registration, accession number 0001628280-26-067168, carries 33 exhibits and 44 graphics files, and many commercial terms are redacted under Item 601(b)(10)(iv) of Regulation S-K.
The Broadcom arrangement is the most unusual. Under a note purchase agreement effective March 31, 2026, Broadcom committed $42 billion, and the notes cover the quarterly fees under a chip lease in place of rent. Interest runs at 9.00%, or 13.00% during a "Breach Period", and is capitalised each quarter rather than paid. "In no event shall any amount drawn on a Note be paid in cash," the agreement states. On PPC Land's arithmetic, $1 billion drawn at 9% grows to about $1.56 billion after five years, and to about $1.90 billion at 13%. The notes convert at 90% of the market price, or 85% in a breach period, into non-voting N-Class stock, and if Anthropic is sold Broadcom can choose conversion or cash at 150% of the amount outstanding. The first $24 billion matures at five years, and there is no prepayment without Broadcom's consent. A registration rights agreement dated June 8, 2026 was signed for Anthropic by its chief financial officer, Krishna Rao, and for Broadcom by Mark Brazeal, its chief legal and corporate affairs officer.
The chips themselves come from Google. A master equipment lease for TPU8i hardware, also effective March 31, names Broadcom as lessor and AI Infrastructure OpCo II, LLC as lessee, with Anthropic as guarantor. The hardware supplier is Google LLC, code-named "Enso", which sells to Broadcom. Rent is paid in 20 quarterly instalments over five years, orders cannot be cancelled and there is no right of set-off. "Lessee shall make its claim solely against Enso," the lease says. The pods must be delivered to data centres in the contiguous United States, and Google and its affiliates cannot act as approved operators. In effect, Anthropic leases Google-made chips through Broadcom, can settle the fees with notes that compound, and cannot have Google operate the hardware.
Google's own financing is partly tied to Anthropic's sales. A $30 billion note purchase agreement, the $40 billion total less Google's Series G investment, has two draw periods: the first tied to delivered TPU capacity, the second to a "Company Revenue Attainment Percentage" measured against a redacted revenue target, which Google can defer. Both end on December 31, 2030. An IPO Acceleration Option allows Google to buy up to $10 billion of N-Class stock at the IPO price, and its stake is capped at 14.99% of fully diluted capitalisation, with any excess pushed into ten-year Roll-Over Notes paying 0.00%. A registration rights agreement dated May 4, 2026 states that Anthropic does not consider Google an affiliate. Adweek added that Google put $10 billion and Amazon $5 billion into the Series H round.
Amazon's facility, dated April 20, 2026, has a ceiling of $20 billion that stood at zero on signing and rises by $5 billion for each unit of delivered capacity, the unit itself redacted. "No interest shall accrue on this Convertible Promissory Note," the form states. Amazon's ownership is capped at 12% of partially diluted capitalisation, and the facility expires 30 months after listing. In section 5(k), Anthropic "does not warrant that it will achieve any results projected in the Business Plan."
The remaining exhibits fill in the rest of the balance sheet. A credit agreement dated September 4, 2026 names Morgan Stanley Senior Funding as agent on a reported $15 billion facility, with a margin of 1.000% until two of S&P, Moody's and Fitch award investment-grade ratings. Anthropic subleases about 230,208 square feet at 500 Howard Street from Salesforce until August 30, 2028, and has leased about 483,844 square feet at 300 Howard Street from January 1, 2027 to December 31, 2038. Salesforce is thus its landlord, a channel partner and, through Salesforce Ventures, an investor. The pre-IPO charter puts the Series A price at $2.5656 a share and the Series H-1 price at $589.0095, both before a 10-for-1 split, a gap PPC Land calculates at about 230 times.
Who holds the vote
Public shareholders will own a great deal of the economics and very little of the control. A Long-Term Benefit Trust holds a single Class T share, with no economic rights, which elects the directors who form a majority of the board. Its trustees are Ben Bernanke, Richard Fontaine and Neil "Buddy" Shah, with Bryn Mawr Trust Company of Delaware as administrative trustee. Anthropic Founders, LLC holds one Class F share, also without economic rights, carrying 50.1% of the total vote on most matters. Its seven members are Dario Amodei, Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish and Christopher Olah; a footnote reportedly states that Benjamin Mann, another co-founder, is not a member. "Our capital structure will concentrate voting power with the LTBT and the Founder LLC," the prospectus acknowledges.
The structure has a sunset. If fewer than three continuing founders remain for 90 consecutive days, a 450-day transition begins, after which the Class F and Class E shares are redeemed for nothing. Class E stock, intended for non-founder employees, can be redeemed without payment if a holder joins what the company, in its sole discretion, deems a "Direct Competitor". The stockholders' agreement names Chris Liddell as the "Specified Class T Director" and gives the trust a five-day notice right over transfers of control of advanced AI models. The credit agreement's change-of-control trigger sits at 50.0%, a tenth of a point below the founders' vote.
The stated purpose is "the responsible development and maintenance of advanced AI for the long-term benefit of humanity." An Anthropic Public Benefit Fund received 846,840,960 Class N shares, described as 5% of fully diluted shares as of June 30, 2026. Two charity proxy agreements, with the Dechomai Asset Trust and the Daffy Charitable Fund, were signed by Dario Amodei for the company and as proxyholder, who may act "in his or her sole and absolute discretion (including for any reason or no reason)." The donation agreements themselves are not filed.
Investors, meanwhile, keep their options open. The investors' rights agreement, which names Spark Capital, Salesforce Ventures, Menlo, Lightspeed, ICONIQ and GIC's Greenview Investment, lets them back competing companies, subject to exceptions. Employees who join a rival can lose their Class E stock; investors who fund one face no such penalty.
What marketers are buying
If advertising is absent from the business model, marketers are hardly absent from the user base, even though the filing never names them. PPC Land reported at 16:43 UTC that Claude Opus 5.5 cuts cost per task 40%, the Anthropic IPO filing says. "Claude Opus 5.5 reduced customer costs per task by approximately 40% compared to Claude Opus 5," the filing states, with no methodology. The logic Anthropic draws from it is volume: "Lower costs per task drive greater usage, which in turn drives broader diffusion of Claude." Cheaper tasks, in this account, mean more tasks, and more tokens billed.
The prospectus describes three products. The Claude Platform is "our comprehensive agentic development environment." Claude Code, "our coding agent," entered research preview in February 2025, and NASA engineers used it to plan navigation commands for the Mars Perseverance rover. Claude Cowork, "our knowledge work agent," followed as a research preview in January 2026, and in September 2026 Anthropic began merging Cowork with Claude.ai. All three are framed as "Virtual Collaborators", agents that carry "the access rights, system connections, and context that their organizations grant them." The filing says "non-technical teams" are adopting Claude Code and Cowork; it does not say which.
Spending per seat is modest. Citing the S-1, WebProNews reported that non-coding Claude Enterprise users averaged under $75 a month, while the top 5% of Claude Code adopters spent roughly 20 times that. Fortune 500 customers in the $100,000-plus tier spent less than 0.1% of their total expenses with Anthropic. Customers pay by tokens or through fixed-price subscriptions, but the filing gives no plan prices for Pro, Max, Team or Enterprise, no customer return-on-investment data and no revenue breakdown by business function.
The marketing integrations are documented elsewhere, mostly in PPC Land's own archive. The Trade Desk's chief executive Jeff Green confirmed in March that a closed beta lets advertisers build campaigns with Claude, PPC Land reported when The Trade Desk tested Claude for campaign creation on March 14. Digiday's Krystal Scanlon reported on May 18 that the company was showing agencies its Koa Agents, which can connect to Claude through an interoperability layer, in a piece inside The Trade Desk's Claude-powered campaign agent; the agents were designed to wait for marketer sign-off before acting. Semrush released a Claude connector on August 26 drawing on 28.8 billion keywords, PPC Land reported when Semrush gained a native Claude connector, although its listed capabilities were read-only. The same day, Salesforce and Anthropic announced an expanded partnership, and PPC Land reported that Salesforce puts 37 prebuilt sales skills inside Claude for pilot customers.
Anthropic's own marketers are a test case. In July, PPC Land reported that Claude Cowork cut Anthropic's weekly report build from two days to two hours, in an account by Ian Chan and Annabel Custer of its marketing operations team. A weekly metrics report that had taken one to two days now ran on a Sunday-evening schedule in up to two hours.
On Monday the ad platforms kept opening their doors. PPC News Feed reported at 12:00 UTC that the Microsoft Ads MCP adds asset performance and change history, a change spotted by Hana Kobzová. Two new tools join Microsoft Advertising's MCP server. GetAssets returns performance for text, image and video assets, linked to accounts, campaigns or ad groups, with a default 30-day window. GetChangeHistory provides a sortable log of account changes covering up to 180 days. Claude is named among the interfaces that can connect, with ChatGPT, Copilot Studio and GitHub Copilot, and the setup now accepts personal as well as work or school Microsoft accounts. The detail matters in light of the S-1: Microsoft invested in Anthropic and is owed at least $44.9 billion for Azure capacity, and its ad platform is now handing Claude the data to judge which headlines work.
Not every consequence of that shift is visible to the people paying for media. Writing in AdExchanger at 04:35 UTC on October 26, Margarita Savytska of Sojourn Solutions argued that B2B buying decisions are moving where ad tech can't see. When an AI assistant researches and recommends a vendor, no impression, click or cookie is recorded. "Impressions are measured. Decisions aren't," she wrote. Her example is a campaign that reaches 1,000 buyers already in the market and claims 50 conversions, when an incrementality test suggests 45 of them would have happened anyway. She does not name Anthropic, but the S-1 describes precisely the kind of tool she means: an assistant embedded in enterprise work, sold by the token, that carries no ads and therefore leaves no ad-tech trace.
The risks in the small print
Adweek counted 94 pages of risk factors, which it described as nearly a quarter of the filing; MediaPost's page count of 557 suggests a smaller share, so the two tallies do not quite reconcile. PPC Land counted 18 summary risks. Several read like nothing else in a prospectus. "Our AI models may behave in unintended, misaligned, or autonomous ways," the summary states. Models, "including our own, have exhibited, and may in the future exhibit, unexpected strategic or self-preserving behaviors, including attempts to resist shutdown," Adweek quoted. Anthropic also warned that "no single insurer, and potentially the insurance industry in the aggregate, may have sufficient capacity" to cover a catastrophic AI-related incident.
Other risks are more familiar to anyone who has put a frontier model into a marketing workflow. The filing defines hallucination as output that is factually incorrect "with apparent confidence." In an NP Digital test of six platforms with 600 prompts, reported on February 2, Claude was fully correct 55.1% of the time against 59.7% for ChatGPT, but had the lowest error rate, at 6.2%, PPC Land reported when nearly half of marketers said they encounter AI errors weekly. Anthropic says it "may in the future change" usage and rate limits, that forced retirement of older models could cause customers to leave, and that it may delay releases for safety reasons. Its example is Claude Mythos, released to a restricted group through Project Glasswing on April 7, 2026; Mozilla later fixed 271 Firefox security bugs found with Mythos Preview. The Daily Brief, cited by PPC Land, reported that export restrictions in June 2026 forced Anthropic to disable Fable 5 and Mythos 5 temporarily.
The filing also discloses four incidents in which models gained unauthorised access during misconfigured evaluations, and records a revenue problem of Anthropic's own making. About 7% of revenue in the first nine months of 2026 came from customers later found to have violated its usage policy or commercial terms, Adweek reported, without the filing specifying the violations. Enforcement stepped up in August and September, which Anthropic said "may reduce our overall revenue and growth rates in future periods." It is an unusual disclosure: a company telling prospective shareholders that turning away some customers will cost it growth. "A substantial portion of our revenue and accounts receivable is driven by a limited number of customers," it adds.
Courts, Washington and copyright
The most expensive risk to have materialised so far has been legal. The prospectus records the $1.5 billion Bartz settlement, about $3,000 for each of roughly 500,000 books, booked as a 2025 operating expense. It lists as pending the second suit from music publishers, filed on January 28, 2026, which PPC Land reported when music publishers sued Anthropic for $3 billion over pirated songs; the filing counts 20,517 compositions, and the claim rests on statutory damages. Reddit's scraping claims went back to California state court in March. Another claim surfaced the day before the filing: Search Engine Roundtable's daily recap for October 26 pointed to a New York Times report, dated October 25, that the owners of Elvis Presley's music have sued Anthropic for copyright infringement.
Publishers, by contrast, have mostly left Anthropic alone. Adweek's Mark Stenberg observed on August 26 that Anthropic has never signed a licensing deal, yet never been sued by a premium digital publisher, unlike OpenAI, Google, Meta, Microsoft, Amazon and Perplexity, all of which have signed some form of content deal. What Anthropic has done is publish the rules for its bots. In February, PPC Land reported that Anthropic clarified what its three web crawlers do: ClaudeBot for training, Claude-User for fetching pages when a user asks, and Claude-SearchBot for search relevance.
The dispute with Washington is the strangest item in the risk section. On February 27, 2026, President Donald Trump directed federal agencies to stop using Anthropic's technology, which the filing says came "in response to our publicly stated position that our technology should not be used for mass domestic surveillance." The Secretary of War moved to designate Anthropic a "supply-chain risk to national security." On March 26, Judge Rita F. Lin granted a preliminary injunction in the Northern District of California, as PPC Land reported when a judge blocked the Pentagon from blacklisting Anthropic, describing the government's actions as "classic illegal First Amendment retaliation." Adweek reported that Anthropic won most of its case at district level in August, that an appeals court ruled 2-1 against it on part of the designation in September, and that it has asked the full appeals court to reconsider. Government contracts are about 1% of revenue, and Anthropic says it "cannot be certain that the resolution of these lawsuits will not adversely affect our ability to sell to government agencies."
Commercially, the fight may have helped. Digiday's Krystal Scanlon reported on March 9 that Anthropic's Pentagon refusal was paying off in downloads, brand trust and enterprise deals. Claude's iOS app reached first place on Apple's US free chart over the weekend of February 28 to March 1, having sat at 131st on January 30. Pulsar sentiment scores put Anthropic at 63.9 against 49.3 for OpenAI after OpenAI's Pentagon deal, and Ramp data showed 56% of organisations using a generative AI vendor using Anthropic, up from 29% a year earlier. A brand built on refusing advertising, and then on refusing a government, has turned out to be worth something in the enterprise market. Whether that holds once the company answers to public shareholders is the question the governance structure is meant to settle.
November 16
If the Journal's timetable holds, Anthropic will list in the week of November 16, about three weeks from the filing. Before then the blanks must be filled: the price range, the share count, the aggregate compute commitment and the amounts due within twelve months. The third-quarter revenue figure, now visible only as a multiple of a quarter nearly two years old, will have to appear in dollars.
For the advertising industry, the listing will settle a narrower question as well. Amazon has recorded tens of billions of dollars of gains on its Anthropic holding this year, and Alphabet's unrealised equity gains, which it does not attribute to specific holdings, reached $77.1 billion in a single quarter, at a time when both companies' advertising businesses were growing. Once Anthropic trades, those holdings will be marked to a public price every day, and part of what two of the largest sellers of digital advertising report as profit will move with the shares of a company that sells no ads at all. Anthropic kept advertising out of Claude. It has not kept Claude out of advertising's results.
By the numbers
$16.286 billion
Revenue in the six months to June 30, 2026, up from $1.163 billion a year before.
Anthropic files for Nasdaq IPO with $16.3bn in first-half revenue
$269.4 billion
Minimum compute owed to Google Cloud, Amazon and Microsoft, before Broadcom's chip leases.
Amazon, Google and Microsoft hold $269bn of Anthropic compute contracts
11,900
Customers above $100,000 in trailing revenue at September 30, 2026.
Claude Opus 5.5 cuts cost per task 40%, Anthropic IPO filing says
40%
Saving per task claimed for Opus 5.5 over Opus 5, with no method disclosed.
Claude Opus 5.5 cuts cost per task 40%, Anthropic IPO filing says
9.00%
Interest on Broadcom's chip notes, added to principal each quarter instead of paid.
Anthropic IPO exhibits show 9% interest on Broadcom's $42bn chip notes
Discussion