A marketing budget that went from immaterial to nine figures
Anthropic will not put advertising inside Claude when it lists on the stock market, but the company is buying plenty of advertising elsewhere. Its S-1 registration document, filed with U.S. regulators ahead of the listing, records advertising spending of $33 million in 2024 and $147 million last year. That exceeds the $125.7 million Duolingo devoted to sales and marketing combined in 2025, even though that app built its reputation on promotion.[S1][S7]
The outlay has continued climbing. In the first six months of 2026, Anthropic spent $273 million more on marketing and advertising than in the equivalent period a year before. Even so, advertising remains a minor line against total sales, marketing and partnership costs of $3.17 billion in that half-year, most of which went to cloud providers that resell Claude and whose fees made up $2.7 billion of the increase.[S1]
The contrast with 2023 is stark. Back then the filing described Anthropic's advertising outlay as not material, Claude had only just gone on commercial sale in the first quarter, and the company booked $41 million in revenue for the entire year. Since early 2025 it has released a steady stream of products, among them the Claude Code coding agent that February, a series of models culminating in Opus 4.5 in November, and Claude Cowork for office work in January 2026.[S1]
Product launches drive the spending
Anthropic ties the 2025 increase to greater media and public relations outlays in support of product launches, and gives the same explanation, minus public relations, for this year's figure. Sales and marketing costs rose in every quarter the document covers, with the sharpest gains in the first half of 2026. The filing lists Opus 4.5 among the factors behind quarterly revenue growth accelerating from 48% in the final three months of 2025 to 132% in the first quarter of 2026.[S1]
First-half revenue this year came to $16.3 billion, up from $1.2 billion during the equivalent period in 2025. Billing works on a usage basis: clients pay for the volume of Claude they consume and for the specific model doing the work, and firms that build Claude into their own products are charged for the traffic those products create. Of the $16.3 billion total, usage charges accounted for $13.1 billion while subscriptions made up the rest, and the filing projects usage will remain the bulk of revenue for the foreseeable future.[S1][S4][S6][S8]
That structure means growth depends on customers leaning on Claude more heavily. Every additional coding task run through a metered account raises the bill, and subscribers who exhaust their limits can move to a costlier plan or buy extra capacity. The filing states that promoting the brand demands substantial expenditure that is expected to grow as the market becomes more competitive, and warns that Anthropic could fall behind rivals with more established or widely recognized brands.[S1]
Brand risk and a public fight with Washington
Competitive anxiety runs through the document. Anthropic cautions that it could be lumped together with other industry players and assessed in the same way regardless of differences in its approach, and flags the risk that public perception of its products as contributors to job displacement could produce strongly negative coverage. Nate Elliott, principal analyst for AI in marketing and commerce at eMarketer, observed that the filing treats brand as one of the company's most valuable assets, noting how often the word appears in the risks section.[S1]
Throughout 2026 the company has marketed itself as an AI business of a different stripe, and that stance has carried a price. In February it stated publicly that its technology ought not to power mass domestic surveillance or fully autonomous weapons. President Trump subsequently ordered federal agencies to halt their use of Claude, and the Department of War labeled Anthropic a supply-chain risk to national security. Anthropic filed suit; a California federal court sided with it largely in August, yet a Washington appeals court later let the designation stand 2-1, prompting the company to seek review by the full court.[S1]
Roughly two months after that clash began, Anthropic held back its strongest model, offering Mythos in April only as a restricted preview to a handful of organizations in cybersecurity, pointing to its heightened capacity for uncovering cyber vulnerabilities. Excluding advertising from Claude fits the same strategy: the filing lists monetization approaches that set it apart from some rivals among its distinguishing traits, though it does not identify them. OpenAI started running ads in ChatGPT in February, and Anthropic devoted part of its Super Bowl spot to pledging Claude would stay ad-free, contending that serving businesses is more lucrative since enterprise clients yield higher monetization than consumer AI.[S1][S7]
