Anthropic forces OpenAI to rethink its “too many bets, too little direction” strategy

OpenAI is pulling back from its broad product strategy and zeroing in on coding tools and business customers, as the company’s top leadership admits the current approach has left it scrambling to keep up with rival Anthropic.

Fidji Simo, OpenAI’s CEO of applications, told employees at a company-wide meeting last week that senior leaders, including CEO Sam Altman and chief research officer Mark Chen, were actively reviewing which products to cut or scale back. Staff have been told to expect a formal update in the coming weeks.

“We cannot miss this moment because we are distracted by side quests,” Simo said, according to remarks reviewed by The Wall Street Journal. “We really have to nail productivity in general and particularly productivity on the business front.”

Over the past year, OpenAI rolled out a long list of new products, including the video generator Sora, a web browser called Atlas, a hardware device, and e-commerce features for ChatGPT. Altman had compared the model to “betting on a series of startups” inside the company.

But the pace of releases created confusion internally, with current and former employees saying it was often hard to understand where the company was headed. Computing resources shifted between teams at the last minute, and the structure grew unwieldy. The Sora team, for instance, sat inside the research division despite launching one of the company’s most visible products.

Sora’s standalone app launched last September alongside a TikTok-style social feature. It briefly hit the top of Apple’s App Store but lost steam in the months after. OpenAI is now looking to fold its video tools into the main ChatGPT app instead.

Anthropic’s rise forces a rethink

Simo told staff that Anthropic’s momentum should be seen as a “wake-up call.” Anthropic has grown into the dominant AI provider for businesses, thanks to the success of its Claude Code and Cowork products, which let users hand off complex tasks to AI agents. The products have drawn a large following among software engineers and set off a global stock-market selloff last month. Unlike OpenAI, Anthropic has stayed away from image and video generation, keeping its focus on enterprise and coding customers.

OpenAI has made some progress catching up. It released a new version of its Codex app last month, along with a model called GPT 5.4 tailored for professional use. Simo said on X that Codex now has more than two million weekly active users, up nearly four times since the start of the year. The company is also placing engineers with consulting firms and business partners to push AI adoption across industries.

OpenAI is also benefiting from the Pentagon’s decision to label Anthropic a supply-chain risk after the company refused to allow unrestricted military use of its AI, including for domestic surveillance and autonomous weapons. Anthropic is suing the U.S. government over the designation.

The Information reports that the company has struck a new agreement with Amazon Web Services to bring its AI tools to federal employees, covering both sensitive and general government work.

OpenAI also chasing private equity and public markets

On the financial front, both companies are in talks with private equity firms ahead of potential public listings, possibly as soon as later this year.

OpenAI is in advanced discussions with TPG, Advent International, Bain Capital, and Brookfield Asset Management to form a joint venture valued at roughly $10 billion before investment.

The PE firms would put in about $4 billion in exchange for equity and board seats, with TPG as the lead investor. Anthropic is in separate talks with Blackstone, Permira, and Hellman & Friedman for a similar arrangement, with investors taking roughly $1 billion in common equity.

OpenAI is facing a fresh lawsuit from Encyclopedia Britannica and its Merriam-Webster subsidiary, which accuses it of scraping nearly 100,000 protected articles and dictionary entries to train its ChatGPT models without permission. The publishers also claim ChatGPT reproduces their content closely enough to cut into their web traffic.

“We are very much acting as if it’s a code red,” Simo told staff at the all-hands meeting.