"We Didn't Do Well This Past Year!" How OpenAI Lost the AI Crown and Its Plan to Reclaim Dominance

Once the tech giant that single-handedly sparked the global artificial intelligence (AI) craze, OpenAI has fallen behind its archrival Anthropic. According to The Wall Street Journal on July 31, the maker of ChatGPT is launching an all-out counteroffensive to reclaim the AI throne.

OpenAI’s flagship consumer product, ChatGPT, was once synonymous with large language models, but user growth has slowed significantly. The departure of Fidji Simo, whom CEO Sam Altman viewed as his successor, triggered a major executive reshuffle. Meanwhile, the sales team, competing for lucrative enterprise clients, resorted to heavy discounts and incentives—a strategy that plunged OpenAI into a costly price war.

Insiders reveal that several core OpenAI investors have recently expressed greater concern over the company’s alarming burn rate than its revenue growth during private discussions. Some investors have even hedged their bets by redirecting funds to Anthropic.

At the same time, Anthropic’s revenue growth has surpassed OpenAI’s following the massive success of its code development tool, “Claude Code.” Its valuation now approaches $1 trillion, leaving OpenAI behind in both financial and technological terms. Sources indicate Anthropic is accelerating plans for an initial public offering (IPO) expected this fall and has already begun meeting with potential investors, repeatedly emphasizing its competitive edge over ChatGPT.

"The past 12 months haven’t been our best—that’s mainly my responsibility—but the next 12 months will be the best in our history," OpenAI CEO Sam Altman posted last month on X. "The team is delivering incredible results, and you’ll be very happy with the new products we’re preparing to launch."

Strategic Misjudgment: Underestimating Software Developers

The Wall Street Journal reports that OpenAI’s struggle to regain dominance stems from early senior leadership misjudgments about the AI market’s trajectory. Altman initially bet on ChatGPT for growth, believing more everyday users would pay for subscriptions as AI became embedded in daily life. However, the overnight success of Claude Code proved the real goldmine lies not in general consumers, but in selling tools to elite software engineers and the deep-pocketed enterprises that hire them.

While OpenAI poured resources into flashy side projects like video generators, consumer hardware, and chip development, the smaller but sharply focused competitor Anthropic seized the opportunity. It developed a breakout coding tool and instantly captured market leadership.

To catch up, OpenAI launched a series of new models focused on coding and professional tasks and assigned President Greg Brockman to overhaul the product line. The company also struck a deal with Amazon to sell its AI tools to the cloud giant’s customers and hired former Slack CEO Denise Dresser as its first Chief Revenue Officer (CRO).

"I think we’re now entering a state where it feels like the machine is truly running efficiently," OpenAI President Greg Brockman said at a media lunch in July. Meanwhile, Altman personally traveled to Washington, D.C., to meet with Trump administration officials and lawmakers to discuss further industry regulation, while OpenAI unveiled a new model.

Company leaders are still working to find the most effective pricing strategy for its enterprise product "Codex," including considering price cuts to gain market share—though this could erode the closely watched gross margin ahead of an IPO.

According to sources familiar with the plan, after failing in its ambition to go public before Anthropic, OpenAI may now delay its IPO until next year. When OpenAI filed its own listing application one week after Anthropic submitted theirs, it remained tight-lipped about timing, stating only that "it might take some time" because "there are things we want to do that are easier as a private company."

Codex’s Failure

The Wall Street Journal argues that OpenAI once held a commanding lead in the enterprise market it now desperately wants to reclaim.

In autumn 2024, the company launched a series of so-called "reasoning models" capable of step-by-step thinking before answering—ideal for writing code. However, researchers trained these models by having them solve high school programming competition problems to achieve perfect scores, rather than tackling the complex, open-ended practical work involved in real-world software development.

Months later, Anthropic took the opposite approach with its reasoning model Sonnet 3.7. In a February 2025 blog post, Anthropic explicitly stated its R&D focus was on "real-world tasks" that reflect how enterprises actually use AI.

Last year, OpenAI allowed employees early access to the new Codex, expecting instant popularity, especially among tech-savvy engineers who follow technological advances. But actual usage fell short of expectations, frustrating executives who had repeatedly raised internal warnings. After OpenAI officially launched the tool in May this year, performance failed to meet external expectations, and software engineers flocked to Claude Code.

Many developers found Codex cumbersome and slow, prompting OpenAI to redesign its products to align more closely with Anthropic’s style. The company even formed a new team called "Coding Ninja" dedicated to ensuring new models meet actual customer needs. At the same time, executives were distracted by other urgent crises, including countering Meta CEO Mark Zuckerberg’s aggressive talent poaching campaign and repairing its deteriorating relationship with largest investor Microsoft.

By the time OpenAI refocused attention on Codex, it was already far behind.

OpenAI executives launched GPT-5.2, a new model focused on coding and professional tasks, overriding employee requests to delay release for further refinement. When the company learned Anthropic was negotiating a partnership with chip startup Cerebras, OpenAI rushed to sign its own agreement first, attempting to block the competitor—but these efforts failed to stop the bleeding.

While Altman vacationed in St. Barts in the Caribbean last December, engineers in San Francisco spent hours immersed in a "Claude coding frenzy," marveling at the capabilities of Anthropic’s new model Opus 4.5. Soon after, Anthropic experienced explosive growth that even surprised its own leadership. Meanwhile, after Google boosted its own consumer chatbot, ChatGPT’s growth suddenly stalled. Anthropic’s growth rate and valuation surged past OpenAI’s, leaving OpenAI caught in a two-front battle and severely challenged in its lead in the AI race.

When OpenAI executives proposed partnering with private equity giant Blackstone to form a new company selling AI tools to Blackstone’s portfolio firms, Blackstone declined and instead chose to collaborate with Anthropic. OpenAI had no choice but to pursue alternative partnerships with other private equity firms.

By March this year, frustration among OpenAI employees peaked. During internal meetings, sharp questions were raised: "It feels like Anthropic, with far fewer employees and a much smaller market cap, keeps setting the standard technically and culturally, while we’re just reacting. Why does this keep happening?" "If Anthropic’s revenue surpasses ours, what impact will that have on our IPO plans?"

Fidji Simo, brought in to assist with IPO preparations, stated that OpenAI had lost direction by chasing "side quests" and now lagged behind in both research and product development. She emphasized that Anthropic’s revenue "reflects how well they’re achieving their mission," adding, "They’re accelerating such an important part of the economy—this should be a real wake-up call for us." Simo later stepped down from her full-time role at OpenAI due to health issues.

OpenAI’s Last Stand

Now, OpenAI is making an all-out effort to reverse its decline, including developing a new product that integrates Codex, ChatGPT, and web browsing.