Anthropic's $81.3B IPO Gives Agencies a Quarterly Vendor Check

The AI company's costs and losses are now public, which gives agencies on Claude an early read on pricing and policy changes.
Anthropic's $81.3B IPO Gives Agencies a Quarterly Vendor Check
watch video
Coral Cripps
By , B2B Reporter
Reviewed by Katherine Maclang, Lead News Editor

Anthropic priced its IPO at $83 a share on Wednesday night, above its $70 to $78 range.

The offering raised about $81.3 billion, the largest IPO total on record before overallotment shares.

It tops the $75 billion SpaceX raised in its June base deal, which later totaled $85.7 billion after underwriters exercised an overallotment option.

Orders reached nearly seven times the shares on offer.

The maker of Claude began trading on Nasdaq as ANT on Thursday.

Shares opened at $111.50 and closed at $97.17, 17% above the offer price.

Agencies using Claude for research, creative work, or campaign tasks can now check their vendor's finances every quarter through public filings.

Anthropic's S-1 filing details its losses, infrastructure spending, and future commitments, giving agencies a better picture of the financial pressures that could affect AI pricing.

Brands now read the same vendor numbers as their agencies, so AI risk becomes a shared agenda item in client reviews.

Here, we look at what the filing reveals, plus three checks agencies should run before tying client work to an AI vendor.

Anthropic's Finances Are Now Public

Anthropic reported third-quarter revenue of $16.96 billion, up 47% from the second quarter.

Its amended filing also reported a net loss of $23.29 billion.

About $21.9 billion of this loss came from non-cash changes in the value of convertible notes and similar instruments.

Anthropic's operating loss for the quarter was $567 million.

Its annualized revenue run rate reached roughly $81 billion by the end of October, up $10 billion from the end of September.

At the IPO price, Anthropic's fully diluted valuation was about $1.5 trillion, while the company also carries $6.2 billion in debt.

Thursday's close lifted Anthropic's fully diluted value to about $1.76 trillion, as reported by CNN.

During the first trading session, the stock ranged from $92.72 to $115.10.

The $81.3 billion figure covers the base offering, while underwriters also have an option to buy up to 147 million additional shares through December 18.

A bar chart showing Anthropic's compute and infrastructure costs as a share of revenue, by quarter

Anthropic disclosed about $900 billion in purchase and lease commitments, including roughly $70 billion due within 12 months.

These disclosures help teams judge the risk of a price change before it hits a client's marketing budget.

An agency running Claude across many accounts can read each filing once and brief every client, which spreads vendor monitoring across its roster.

Compute Costs Could Affect AI Pricing

Anthropic spent $9.86 billion on compute and infrastructure in the third quarter, equal to 58% of revenue, up from 48% in the second quarter.

The company said that $5 billion of this expense came from short-term compute, which costs more than capacity secured through multiyear agreements.

Consumption-based services accounted for 79% of revenue during the first nine months of 2026, while subscriptions made up the remaining 21%.

Agencies using Claude through its API pay according to usage, tying their bills to the same infrastructure costs Anthropic reports each quarter.

The filing, however, does not announce a price increase.

Anthropic says its own tests show Claude Opus 5.5 costs about 40% less to run than Opus 5 on typical workloads.

Agencies that price client work by task can monitor compute costs in each quarterly report, giving them an additional data point when reviewing their own AI budgets.

Brands hiring an agency for AI-heavy work should ask whether usage costs sit inside the retainer because this decides who absorbs a price change.

Customer Mix and Policy Rules Affect Agencies

Anthropic's five largest customers accounted for about 25% of revenue during the first nine months of 2026, while no single customer contributed more than 10%.

Cloud marketplaces operated by Amazon, Google, and Microsoft handled 48% of first-half revenue, alongside 58% of net receivables at the end of June.

Anthropic estimates that 5% to 7% of revenue came from customers later found to have breached its usage policy or commercial terms.

The company expects revenue from these customers to decline and warns that the crackdown may reduce its overall revenue.

Agencies buying Claude through a cloud marketplace use a sales channel responsible for nearly half of Anthropic's first-half revenue.

Brands, on the other hand, can ask their agency to keep their work on a separate Claude account, so enforcement against another client stays off their projects.

Shareholders Will Watch Claude's Ad-Free Pledge

Anthropic's February blog post says "Claude will remain ad-free" and promises that sponsored links will not appear next to conversations.

Its Super Bowl campaign in February closed every spot with the line "Ads are coming to AI. But not to Claude."

The founders' letter in the prospectus repeats this commitment for public investors.

However, one analysis of the S-1 found no binding commitment to keep Claude ad-free in the filing itself.

OpenAI has expanded ChatGPT ads since it began testing them with U.S. Free and Go users on Feb. 9.

A Comscore study found sponsored ads in 47% of tracked ChatGPT video game conversations in June, up from 3% in March.

The figures cover U.S. desktop activity from an opt-in panel, offering an early indication of how often ads appeared in those conversations.

Brands seeking visibility in Claude answers have to earn it organically, which makes a content and SEO agency the right hire for this channel.

Agencies can run these three checks before a client workflow depends on an AI tool:

Ask About Pricing Changes

Agencies should request written notice periods for usage-price changes and ask whether committed-spend agreements lock rates for the contract term.
Anthropic's S-1 says that some customers agree to minimum spend commitments in exchange for a committed price list.

Check Workflows Against Usage Rules

Agencies should list the client tasks running on each model and check them against the vendor's usage policy.
The New Stack reports that Claude Opus 5.5 automatically reroutes flagged cybersecurity and biology requests to Opus 4.8 or Opus 5.

Test a Backup Model for Each Task

Agencies should approve a second model for critical tasks and test its prompts and outputs in advance. This gives teams a fallback if pricing or terms change during a project.
Models Anthropic released from April to September drove more than 80% of its Q3 revenue.

Brands running an agency review can ask to see these checks in the pitch to learn how a shop protects client work when its martech stack changes.

Our Take: Can Agencies Trust Claude?

The biggest vendor risk in Anthropic's filing lies in its usage rules and access controls.

We'd trust Claude with client work, as long as every critical task has a tested backup model ready.

A backup adds testing hours and a second vendor bill, so agencies should price this redundancy into their AI budget.

Anthropic just raised $81.3 billion, so the odds of it running out of money mid-retainer look remote.

Model access has already slipped once this year.

Claude Fable 5 and Claude Mythos 5 launched on June 9, then Anthropic suspended both three days later to comply with U.S. export controls.

Each quarterly filing now gives agencies a set date to retest their backups and update client contracts.

Our earlier report on Claude's watermark policy covers another Anthropic rule that changes how agencies deliver client content.

Brands using AI tools need agency partners who understand the companies behind the technology.

Take a look at these top AI companies and agencies in our directory.

👍👎💗🤯
Latest AI News
Receive our NewsletterJoin over 70,000 B2B decision-makers growing their brands