OpenAI is pulling the plug on its consumer experiments to refocus on developer tools and enterprise sales, a strategic retreat driven by Anthropic's rapid gains in exactly the market OpenAI now wants to own. Fidji Simo, OpenAI's CEO of applications, laid out the new direction at an all-hands meeting reported by the Wall Street Journal on Monday.
"We cannot miss this moment because we are distracted by side quests," Simo told staff. "We really have to nail productivity in general and particularly productivity on the business front."
The Anthropic problem
Simo was blunt about why. She called Anthropic's success a "wake-up call" and told employees the company is operating in "code red" mode, a phrase that's becoming a recurring motif at OpenAI. Simo seemed aware of the pattern, adding that she doesn't think "declaring codes for everything makes a ton of sense," which is a strange thing to say right after declaring one.
The urgency is real, though. Anthropic's Series G in February valued the company at $380 billion, fueled by enterprise revenue that has apparently quadrupled since January. Claude Code alone now generates an estimated $2.5 billion in annualized revenue and, according to Anthropic, accounts for over half of all enterprise spending on its products. According to market data, Claude Code's daily VS Code installs surged from around 17.7 million to 29 million since the start of 2026. Those numbers come from third-party estimates and should be taken with some caution, but the direction is clear.
Current and former OpenAI employees told the Journal that last year's approach of betting on multiple products at once made it hard to understand the company's strategy. That's a polite way of saying the product roadmap was a mess.
Sora tells the story
The clearest example of what's being left behind: Sora, the AI video generator that OpenAI launched as a standalone app in October 2025. It hit the top of the App Store and reached a million downloads faster than ChatGPT. Then it cratered. Appfigures data shows downloads dropped 32% in December and another 45% in January, falling to 1.2 million installs. Consumer spending peaked at $540,000 in December and slid to $367,000 by January. Total lifetime revenue: $1.4 million. For a company valued at $500 billion, that's a rounding error.
OpenAI is reportedly folding Sora into ChatGPT rather than maintaining it as a separate product. The Atlas browser, an eCommerce layer for ChatGPT, a hardware device: all of these were part of Sam Altman's strategy of running what he described as "a series of startups" inside OpenAI. That experiment is over.
Why now, specifically
The timing maps neatly onto OpenAI's IPO preparations. The company has been in informal talks with Wall Street banks about a potential fourth-quarter 2026 listing, according to Fortune. It recently hired a new chief accounting officer and a business finance officer to run investor relations. Current valuation sits around $500 billion, with a fundraising push that could take it to $830 billion. The company doesn't expect to turn a profit until 2030.
Investors looking at an IPO prospectus want predictable revenue growth from customers who sign annual contracts, not App Store download charts that spike and collapse. Enterprise software and API access provide that. Consumer products like Sora, with its $1.4 million in total spending across 9.6 million downloads, do not.
Anthropic is running the same calculation in reverse. Its enterprise market share reportedly grew from 24% to 40%, it just launched a major partnership with Accenture involving 30,000 trained consultants, and it recently committed $100 million to a new partner network. Anthropic is also eyeing a public listing, potentially by late 2026 or 2027. OpenAI executives have reportedly expressed concern about Anthropic listing first and absorbing the pent-up retail investor demand for AI stocks.
The Pentagon wrinkle
There's an unusual subplot here. The Pentagon recently designated Anthropic a supply chain risk after CEO Dario Amodei refused to let the military use Claude without restrictions on autonomous weapons and mass surveillance. OpenAI swooped in within hours to replace Anthropic in classified military environments. Some businesses have reportedly grown hesitant to work with Anthropic as a result, which OpenAI clearly sees as an opening. Anthropic is suing, and over a million people signed up for Claude daily in the days after the designation, so the PR damage cuts both ways.
Simo's pivot is less about new ideas and more about acknowledging which bets didn't pay off. OpenAI already launched Frontier, an enterprise platform for building and managing AI agents, in February. The infrastructure for the enterprise push exists. The question is whether shedding the consumer experiments fast enough will let OpenAI close the gap before both companies face public market scrutiny.
Employees were told they'd learn in the coming weeks which projects are being deprioritized. For a company that launched a video app, a browser, a hardware device, and a shopping feature in a single year, that's a long list.
