To The Moon

OpenAI, the company that sparked a global AI frenzy, has been overtaken by competitor Anthropic. Now, the firm is fighting hard for a comeback. Growth in its flagship consumer product, ChatGPT, has slowed significantly. The departure of Sam Altman's designated successor, Figi Simo, has prompted a reshuffling of responsibilities among senior executives. The sales team is aggressively pursuing high-value enterprise clients by offering bulk discounts and various incentives, incurring high costs.

In recent months, some of OpenAI's major investors have privately expressed concerns that the startup is burning cash too quickly without matching revenue growth, according to individuals involved in those discussions. Other investors are diversifying their risk by also funding Anthropic. Meanwhile, buoyed by the success of its coding tool Claude Code, Anthropic has recently surpassed OpenAI in both revenue growth and corporate valuation, which is now approaching $1 trillion. Insiders say Anthropic is accelerating its fall IPO plans, initiating discussions with potential investors and emphasizing its lead over ChatGPT developer OpenAI.

Altman posted on X earlier this month: "We haven't delivered our best work in the last 12 months, and the fault is mostly mine. But the next 12 months could be the best period of development we've ever seen. The team is working on major projects, and everyone will see surprises soon." OpenAI's struggle to maintain its industry leadership stems from management's misjudgment of the AI market's direction. The company's current predicament reflects the intense competition for global AI dominance.

Altman initially bet the company's growth on ChatGPT, expecting that as AI became more integrated into daily life, more users would pay for subscriptions to the chatbot. However, the overnight success of Claude Code clearly demonstrated that the bigger opportunity lies in selling tools to high-end software developers and large enterprises that employ many programmers. While OpenAI pursued flashy projects—from video generation models and consumer devices to custom chips—a smaller, more strategically focused competitor seized a market gap, launched a blockbuster coding tool, and pulled ahead.

To catch its primary rival, OpenAI has released a series of new models focused on coding and professional workplace scenarios, appointing President Greg Brockman to lead a product revamp. The company also partnered with Amazon to sell AI tools to that cloud giant's customers and hired former Slack CEO Dennis Dresser as its first Chief Revenue Officer. Brockman said at a media lunch in July: "I believe our entire operating system is now running efficiently." This week, Altman is in Washington, meeting with Trump administration officials and members of Congress to preview new models, as the U.S. government debates how to further regulate the AI industry.

The executive team is still determining the optimal pricing for the Codex enterprise version. A price cut could expand market share but would erode profit margins closely watched by the capital markets, potentially pressuring a future IPO. Insiders indicate OpenAI may postpone its public listing until next year. The company had previously hoped to go public before Anthropic. A week after Anthropic filed its IPO application, OpenAI also submitted listing documents but did not commit to a timeline, stating it was "too early" for an IPO, citing that "many strategic goals are easier to pursue as a private company."

Codex faces serious challenges. In fact, OpenAI once held a first-mover advantage in the enterprise service space but is now scrambling to catch up. In fall 2024, OpenAI launched a series of reasoning models capable of step-by-step problem-solving, a skill well-suited for code writing. However, researchers trained the models to solve high-school-level competitive programming problems, not the fragmented, open-ended engineering tasks of real-world software development. Months later, Anthropic took the opposite approach, releasing the reasoning model Sonnet 3.7. In a February 2025 blog post, the company said its R&D focus was on "real-world business scenario tasks," aligning with how enterprises actually use AI.

Early last year, OpenAI opened a beta version of the new Codex to employees, anticipating the product would quickly take the market by storm, with experienced engineers tracking AI progress eager to use it. But user activity fell short of expectations, prompting repeated internal warnings from executives. Initially, Codex's issues were viewed as short-term volatility. ChatGPT's weekly active users continued to rise, and the company's valuation climbed. In August last year, OpenAI hired former Instacart CEO Simo to prepare for a highly anticipated IPO. However, Codex's cold reception proved to be a warning signal for a series of crises.

After OpenAI officially launched the product in May, its market performance was disappointing, with software engineers flocking to Claude Code. Many developers reported Codex was slow and cumbersome to use, forcing OpenAI to redesign the product, mimicking Anthropic. The company also established a special team called "coding ninja" to ensure new models were designed according to customer needs. Meanwhile, executives were distracted by other urgent matters: fending off poaching attempts by Meta CEO Mark Zuckerberg and repairing the deteriorating relationship with its largest investor, Microsoft. Valuable computing resources were poured into ultimately failed projects, including the video generation app Sora.

By the time the company refocused resources on Codex, the market opportunity was lost. In December, OpenAI released GPT-5.2, a new model focused on coding and professional work scenarios. Some employees suggested delaying the release to polish the product, but management rejected the proposal. Insiders say the company heard Anthropic was in talks with chip startup Cerebras, which develops specialized chips capable of processing code tasks quickly. OpenAI quickly signed a cooperation agreement to try to block its competitor. But this move still couldn't stem the decline.

During the December Christmas holiday, Altman vacationed on St. Barts in the Caribbean, while programmers in San Francisco stayed up all night testing Anthropic's new model, Opus 4.5, marveling at its power. Subsequently, Anthropic's revenue grew rapidly for several months, exceeding management's expectations. At the same time, Google's consumer chatbot surged in popularity, and ChatGPT's user growth suddenly slowed. OpenAI had set a target of reaching 1 billion weekly active users by the end of last year—a goal initially missed, only recently achieved. Soon after, Anthropic surpassed OpenAI in both revenue growth and valuation.

As its leadership position in AI came under threat, OpenAI found itself fighting on two fronts. OpenAI executives pitched a partnership to Blackstone Group senior management to create a new company selling AI tools to Blackstone's portfolio companies, but Blackstone declined, opting instead to work with Anthropic on a project. OpenAI later reached a separate deal with other private equity firms. By March of this year, some OpenAI employees grew frustrated, peppering a senior executive with questions about the company's future direction. One internal query read: "Anthropic has far fewer employees and a lower market cap than us, yet they set the rules on both technology and industry narrative, while we only react. Why does this keep happening?" Another question was more pointed: "If Anthropic's revenue surpasses ours, what impact will that have on our IPO plans?"

Simo recently resigned due to deteriorating health. She commented that the company had previously chased various side projects, losing strategic focus and falling behind competitors in both technology development and product deployment. "Anthropic's revenue is a direct measure of how well they are executing their mission. They are penetrating many core areas of the economy, and this should serve as a wake-up call for us." OpenAI is now trying to capitalize on the momentum from its recent products to turn the tide in the AI race. The company recently launched a super app integrating Codex, ChatGPT, and a web browser. It claims the new product, along with the standalone Codex app, has already surpassed 10 million total users. This month, OpenAI released a new model, GPT 5.6 Sol, which quickly gained traction with developers, prompting Anthropic to expand public access to its own flagship model, Fable, to compete.

Shifting industry sentiment is also benefiting OpenAI. Some enterprises are beginning to push back against Anthropic, accusing it of trying to block access to more cost-effective Chinese AI models in the U.S. market. Dan Shipper, CEO of Every, a media and software company focused on AI, said: "OpenAI has now truly optimized for the daily experience of developers and knowledge workers. I'm very optimistic that OpenAI can reclaim industry dominance from Anthropic."