OpenAI Lost the AI Crown to Anthropic — Here's How It's Fighting to Win It Back

OpenAI, the company that ignited the modern AI boom, has spent the past year watching a smaller, more focused rival pull ahead. Anthropic's valuation is now nearing $1 trillion — surpassing OpenAI's — fueled almost entirely by the runaway success of Claude Code, its AI coding tool. Anthropic's revenue growth rate has also overtaken OpenAI's, and the company is accelerating plans for a fall IPO, pitching investors on its lead over the ChatGPT maker. OpenAI, meanwhile, may be forced to delay its own public offering until next year.

The root cause, according to people close to the situation, was a fundamental misread of the market. CEO Sam Altman bet heavily on ChatGPT as a mass-consumer product, while the real money turned out to be in developer and enterprise tools — specifically, AI that could write and manage code for software engineers and the companies that employ them. While OpenAI chased flashy side projects — a video generator called Sora, consumer hardware, custom chips — Anthropic quietly built Claude Code into a must-have product. Software engineers found Codex, OpenAI's competing tool, too slow and clunky, and defected en masse. OpenAI's internal Codex launch underperformed so badly that it was flagged repeatedly by executives before the public rollout, which also disappointed.

Distractions compounded the problem. Executives burned cycles fending off a talent raid from Meta's Mark Zuckerberg and repairing a strained relationship with Microsoft, its largest investor. Over the December holiday break, while Altman vacationed in St. Barts, coders in San Francisco were spending marathon sessions marveling at Anthropic's Opus 4.5 model. When OpenAI pitched Blackstone on a partnership to sell AI tools to its portfolio companies, the private-equity giant chose Anthropic instead. Internal employee frustration boiled over by March, with staff openly questioning why a company smaller by headcount and market cap was consistently setting the technical and cultural agenda. Former CEO-in-waiting Fidji Simo warned internally that OpenAI had lost its way chasing side quests and fallen behind on both research and product — then departed after a health condition worsened.

OpenAI is now mounting a comeback. It released GPT 5.6 Sol in July, which became an immediate hit with developers and forced Anthropic to broaden access to its own Fable model in response. A new "super app" merges Codex, ChatGPT, and a web browser, and the company claims 10 million users across those combined products. President Greg Brockman has been handed responsibility for overhauling the product lineup. OpenAI also hired former Slack CEO Denise Dresser as its first chief revenue officer, struck a distribution deal with Amazon, and is experimenting with pricing cuts on Codex to win enterprise market share — though it risks compressing margins right before an IPO. Some observers are bullish: at least one AI-focused investor says OpenAI has recently nailed the day-to-day experience for developers and is positioned to reclaim its lead.