Anthropic announced on June 2 that it had confidentially submitted a draft S-1 registration statement to the U.S. Securities and Exchange Commission (SEC) that same day, becoming another super unicorn targeting the U.S. stock market after SpaceX. Confidential submission is a standard practice in the IPO process: companies can communicate with the SEC, undergo review and revisions, and then publicly file the prospectus once underwriters are ready and market timing is favorable. If the listing ultimately fails, it also avoids disclosing extensive internal information. By regulation, the prospectus must be made public at least 15 calendar days before the roadshow begins. Anthropic just completed a $65 billion funding round last week, reaching a post-money valuation of $965 billion — for the first time surpassing OpenAI to become the world’s most valuable AI startup. In February of this year, its valuation was only around $380 billion, roughly half of OpenAI’s, meaning it has nearly tripled within a few months. Departing from its usual low-key approach, the company’s high-profile announcement of the confidential submission is seen by the industry as a direct response to the competitive landscape: reports had previously indicated that OpenAI is also preparing to confidentially file its prospectus soon, meaning the two giants’ IPO timelines are highly likely to collide. Both are expected to raise tens of billions of dollars, so being first to market means gaining an advantage in a market where SpaceX’s IPO has already absorbed massive liquidity.
On the fundamentals side, Anthropic’s annualized revenue has surged from $14 billion to $47 billion in the first five months of this year, and it is expected to turn profitable for the first time in Q2, driven primarily by strong demand for Claude Code and the Mythos large model. Sam Rubin, Senior Vice President of Threat Intelligence at Palo Alto Networks, disclosed that during approximately three weeks of internal testing, Mythos identified over 20 critical vulnerabilities — about five times more efficient than traditional tools. However, during the same period, it consumed roughly $1 million worth of token credits (no actual payment was required during the testing phase). To control costs, Palo Alto Networks established a tiered calling mechanism where Mythos formulates attack plans and less expensive models like Opus 4.7 execute them. Rubin said he has spoken with hundreds of enterprise security executives, all of whom plan to increase security budgets due to the growing threat of AI-driven attacks. “Mythos has elevated the importance of security to a new level — board members are asking about it, CFOs are asking, and CEOs are asking.” Anthropic stated that Mythos will be made available to all customers in the coming weeks after the internal testing transition period ends.