OpenAI Files IPO Draft, Listing Still Uncertain

OpenAI Files IPO, Altman Says Listing Not Urgent, AI Self-Improvement May Be Bigger Variable

On June 8, 2025, OpenAI officially submitted an S-1 draft to the U.S. Securities and Exchange Commission (SEC), initiating the IPO process. Meanwhile, competitor Anthropic had secretly filed its S-1 draft on June 1. Both companies are valued near $1 trillion, but OpenAI CEO Sam Altman stated internally that if recursive self-improvement (RSI) accelerates, delaying the IPO could be more advantageous.

Background: From Nonprofit to IPO

OpenAI was founded in 2015 as a nonprofit with the mission to "ensure that artificial general intelligence benefits all of humanity." In 2019, due to high training costs for large models, OpenAI established a subsidiary structure with a profit cap and began accepting investments, with Microsoft injecting $1 billion initially. After the launch of ChatGPT in late 2022, the company's commercial value surged. In late 2023, the board briefly ousted Altman, who was reinstated under pressure from investors and employees. From 2024 to 2025, OpenAI restructured: the for-profit arm became a public benefit corporation (PBC), the nonprofit parent was renamed the OpenAI Foundation (holding about 26%), Microsoft holds about 27%, and employees and investors collectively hold about 47%. This S-1 filing is a natural result of the restructuring.

Key Details: IPO and AI Model Progress

IPO Process: After submitting the S-1 draft, OpenAI made it public early due to potential leaks. The company stated that being private is more conducive to strategic advancement, but after weighing options, it still filed the IPO; whether to list remains under consideration. Anthropic's S-1 draft is under SEC review, with share count and offering price yet to be determined.
Model Iteration Accelerates: OpenAI Chief Scientist Jakub Pachocki confirmed that a new model, codenamed 5.6, will be released this month, with performance "significantly surpassing" its predecessor. Previously, GPT-5.4 was released on March 5, and GPT-5.5 on April 23, with intervals of about 6-7 weeks, and no slowdown in generational capability improvement. Community leaks show that GPT-5.6's internal codename is iris-alpha, with subsequent versions including ember-alpha, beacon-alpha, kepler, and kindle, with kindle-alpha being the current release candidate.
Product Redesign: On June 10, ChatGPT's model selector was revamped, changing from model names (e.g., Thinking-Light, Thinking-Standard) to intelligence tiers (Instant, Medium, High, Extra High, Pro Standard, Pro Extended), aiming to simplify user choices.
Price War Signal: According to the Wall Street Journal, OpenAI is planning to significantly lower API pricing to compete with Anthropic. Currently, GPT-5.5 API pricing is $5 per million input tokens and $30 per million output tokens, while Anthropic's Fable 5 and Mythos 5 are priced at $10 input and $50 output.

Reactions and Data

Altman's Internal Statement: In OpenAI's internal Slack, Altman suggested that if AI recursive self-improvement accelerates fast enough, delaying the IPO could be more beneficial, as technology and the world may change in unexpected ways, and private companies have more flexibility.
Anthropic Data: Internal reports show that AI task completion time doubles every four months, and engineers' quarterly code output has surged to eight times previous levels.
Valuation Comparison: As of end of May, Anthropic surpassed OpenAI with a valuation of $965 billion versus OpenAI's $852 billion, becoming the world's most valuable AI startup. Both companies are expected to have market caps exceeding $1 trillion after listing.
Financial Data: OpenAI's CFO confirmed annualized revenue has exceeded $20 billion, but losses are severe, with projected losses for 2026 between $14 billion (non-GAAP) and $25 billion (GAAP).

Impact and Outlook

OpenAI and Anthropic are both racing toward IPOs, with combined valuations of about $3.6 trillion, equivalent to France's annual GDP. Investment banks believe the first to list will define the valuation framework for the AI sector. However, Altman's RSI comments suggest that technological breakthroughs could change business rules: if AI achieves self-improvement, the lead will expand exponentially, potentially lowering the priority of an IPO. Currently, three flagship models (GPT-5.6, Claude Fable 5, Gemini 3.5 Pro) will compete head-on in June, focusing on reasoning, coding, agents, and front-end generation. The real race may be at the RSI level, not the IPO timeline.