OpenAI is scrambling to defend its lead in generative AI after CEO Sam Altman declared an internal “code red” and ordered the company to focus almost entirely on improving ChatGPT. The move comes as Google and Anthropic roll out new models that outperform OpenAI on key benchmarks and are rapidly gaining users and enterprise clients.
Code red inside OpenAI
According to multiple reports, Altman told employees that OpenAI will pause or slow several side projects — including advertising tools, shopping and health agents, and a personal AI assistant codenamed Pulse — to redirect engineers onto ChatGPT’s core experience. The memo sets daily development calls and emphasizes speed, reliability, personalization and broader question‑handling as immediate priorities, signaling the most urgent internal reset since ChatGPT’s launch in 2022.
OpenAI still commands enormous scale, with around 800 million weekly active users on ChatGPT and an annualized revenue run rate above 20 billion dollars this year, yet leadership now sees that early advantage narrowing. Altman has previously said the firm is betting on massive infrastructure build‑out to support future models, but the latest directive suggests that product quality and user experience can no longer lag while that long‑term bet plays out.
Google’s Gemini momentum
Google has emerged as the most immediate threat after launching its latest Gemini 3 model, which independent and internal benchmark tests show beating OpenAI’s GPT‑5 across several performance measures. The model’s gains are helped by Google’s custom AI chips and deep integration of Gemini across Search, Workspace and Android, which together give the company a powerful distribution and data advantage.
Usage patterns are also shifting: while OpenAI leads in raw weekly users, recent data indicate people are spending more time inside Google’s Gemini chat interface, suggesting stronger engagement per user. For OpenAI, that raises the risk that casual and professional users begin defaulting to Google’s tools embedded in services they already use daily, from Gmail to Docs to Android phones.
Anthropic’s enterprise push
Anthropic, long positioned as a safety‑first rival, has quietly built a large base of business customers around its Claude family of models. The company says it has surpassed 300,000 business clients, with the number of large accounts generating more than 100,000 dollars in annual revenue growing more than sevenfold in the past year.
Its newest flagship, Claude Opus 4.5, is cited in recent testing alongside Google’s Gemini 3 as outperforming GPT‑5 in several areas, particularly coding tools and complex reasoning. Anthropic’s focus on reliability, guardrails and human‑in‑the‑loop workflows has made it especially attractive to risk‑sensitive enterprises, from financial services to healthcare, that want powerful models but tight governance.
Costs, profits and pressure
Behind the product race sits a brutal economic reality: OpenAI’s training and inference costs are enormous, and the company still does not have a clear path to sustainable profitability without massive future growth. Internal projections suggest OpenAI would need to reach on the order of hundreds of billions of dollars in annual revenue by around 2030 to fully justify its infrastructure spending and turn a meaningful profit.
That funding strain contrasts with Google, which can finance AI investment from its existing ad and cloud businesses, and Anthropic, which so far spends more conservatively while growing high‑margin enterprise deals. The combination of rising competition, soaring compute costs and an intense fight for AI talent has led analysts to question whether OpenAI can maintain its leadership without reshaping both its business model and product roadmap.
What the shift means for AI
For users, this new phase of competition is likely to bring faster updates, more personalized assistants and better integration of AI into everyday tools, as all three companies race to prove their systems are more useful in real‑world work and life. For regulators and customers, the rise of strong alternatives in Google and Anthropic means the generative AI market is no longer effectively synonymous with OpenAI, which could both ease concentration fears and intensify scrutiny on safety and governance across the sector.
If OpenAI’s “code red” succeeds, the company could regain product momentum and convert its huge user base into a more durable ecosystem before rivals pull too far ahead. If it stumbles, the narrative of the AI race may shift decisively from a single dominant pioneer to a three‑way contest where model quality, economics and trust matter as much as scale.
