Anthropic Releases AI Agents for Banks and Prepares for IPO This Year
Anthropic has unveiled 10 new AI agents tailored to automate routine tasks in financial services, including compliance and presentation preparation. This is part of a strategy to expand the business ahead of an anticipated initial public offering.
Two days that turned the corporate AI market upside down: first, a $1.5 billion joint venture with private equity giants, then the launch of financial agents, and the first joint appearance of Dario Amodei and Jamie Dimon. What seems like a technological upgrade is actually the most aggressive capture of financial infrastructure since the Bloomberg Terminal. Everyone is silent about the main point: Anthropic is not selling software to banks; it is building an operating system without which a bank will not be able to function in three years.
The Essence: What Is Really Happening
This is not a product launch. It is an operation to capture an exclusive distribution channel through closed portfolio companies to pump up metrics before an IPO. The scheme works like this: Blackstone, Hellman & Friedman, and Goldman Sachs each invest about $300 million in the joint venture. In return, Anthropic gets guaranteed deployment of Claude in hundreds of portfolio companies of these funds—from insurance brokers to regional banks. This is not market competition; it is forced distribution through control over capital.
Why is this important right now? Because revenue from API and chat subscriptions has stopped growing at the previous pace. The enterprise contract market has become the main battlefield: these provide high-margin multi-year commitments. And Anthropic, with 300,000 business clients and over 1,000 companies paying more than $1 million annually, is building a moat around its product that competitors cannot breach: agents are embedded in critical compliance and underwriting infrastructure, where switching costs for banks are prohibitive.
Timeline and Context
The groundwork was laid in July 2025, when Anthropic launched Claude for Financial Services and then expanded the platform. But the key turning point came in January 2026: the launch of Claude Cowork sent shockwaves through Wall Street, driving down stocks of companies whose business could be destroyed by agents.
By February 2026, Anthropic had raised a $30 billion round at a $380 billion valuation, and by April, annual revenue reached $30 billion. By the end of April, the company began discussing a new round with a valuation above $900 billion.
Meanwhile, an IPO advisor was appointed—Wilson Sonsini, the same firm that led Google to its listing in 2004. The window dressing before the IPO peaked on May 5, 2026: the launch of 10 AI agents and the first joint appearance of Amodei and Dimon.
Who Wins and Who Loses
Winners:
- Anthropic: The trillion-dollar valuation is approaching not on hype but on enterprise contracts. Over 80% of revenue is B2B, and each new agent makes the product increasingly indispensable for banks.
- JPMorganChase, Goldman Sachs, and Citi: They get not just software but the status of co-authors of the standard. Their credit memo and compliance check methodologies are embedded in Claude Opus 4.7 and distributed through the joint venture to hundreds of middle-market companies. JPMorgan is effectively exporting its operational standards.
- Blackstone, Hellman & Friedman, and Apollo: Their portfolio companies are the first to get agents that compress AML investigations from days to minutes, reducing regulatory risks and operational costs by 30-40%.
Losers:
- OpenAI: Valued at $852 billion in March 2026, it now risks going public second, with significantly less presence in finance. Anthropic's joint venture with PE giants creates a distribution channel that OpenAI simply does not have.
- Legacy software vendors: SAS, Oracle, niche risk management providers sit on products with 20-year-old codebases. Claude Opus 4.7 with a 64.4% score on the Finance Agent Benchmark leaves them no chance—all risk management, financial statement audit, and KYC checks move to a single AI provider.
- Mid-level finance employees: Vice presidents of compliance, credit analysts, and auditors will find that their work—data collection and checklist verification—is 80-90% automated within 18-24 months. The agent not only works faster but also with a reproducible decision trail, killing the argument that "a human is needed for accountability."
What the Media Is Not Saying
The first fact missed by the market: Moody's embedded its credit rating platform as a native application in Claude. This means Anthropic gets direct access to credit data on 600 million companies. Combined with Dun & Bradstreet, Experian, and S&P Capital IQ, Claude becomes not just an assistant but the primary interface for any credit decision. Moody's effectively handed over its distribution to an AI provider, losing control over how its data is used in the final decision.
The second hidden fact: the joint venture is a mechanism to bypass regulation. If FIS or JPMorgan directly transferred client data to Claude, it would trigger a barrage of regulatory questions about confidentiality. But through the joint venture structure, managed as an independent company, the data formally remains "within FIS's controlled infrastructure" while being processed by Claude. This is a gray area: regulatorily "own" infrastructure, but in reality, full access for Anthropic to sensitive banking data.
The third point almost no one noticed: the launch occurred one day before the publication of the Consumer Financial Protection Bureau (CFPB) report on discrimination in AI underwriting. Anthropic deliberately timed the announcement to hijack the agenda. Instead of discussing bias in credit decisions, the market is discussing the Amodei-Dimon partnership.
Forecast: Next 30 Days and 90 Days
30 days (by June 6, 2026):
JPMorganChase will launch a pilot credit underwriting agent processing $500 million in monthly applications in test mode. There will be no big announcement—information will appear in an internal memo for top management and leak through former employees. FIS will begin migrating at least 15 banks to the Financial Crimes AI Agent. OpenAI will try to force a similar partnership with Morgan Stanley and Bank of America. Anthropic's board will approve the final parameters of the round with a $900 billion valuation.
90 days (by August 6, 2026):
Anthropic will file an S-1 with the SEC with a target IPO date in October 2026. The offering size will be at least $60 billion. The key disclosure item will be the Long-Term Benefit Trust—a structure giving external directors veto power over decisions that contradict the company's safety mission. This will spark a heated debate at the SEC: can a company with such a governance structure be public? The answer will determine not only Anthropic's fate but that of the entire AI industry—for the second time in history (after Google), the market will see a developer of technologies claiming critical infrastructure status disclose its economics to public investors.
— Editorial Team
