Sam Altman Admits GPT-5 Mistakes and Outlines OpenAI's Trillion-Dollar Plans

Sam Altman, CEO of OpenAI, recently met with reporters from TechCrunch and The Verge over dinner in San Francisco, where he spoke candidly about the recent launch of the GPT-5 model and the company's future plans. This informal meeting, which included dishes such as lamb skewers and fish specialties costing hundreds of dollars, offered a rare glimpse behind the curtain of one of the most influential companies in artificial intelligence. Altman admitted that the launch of GPT-5 was not ideal, while also highlighting massive user growth and an ambitious vision that extends beyond models like ChatGPT.

The Mistake of Removing GPT-4o and Lessons from the Rollout

During the dinner, Altman directly admitted that OpenAI made a mistake when it removed the popular GPT-4o model without warning following the launch of GPT-5. "I legitimately think we totally screwed that up," Altman told reporters. Users on platforms such as Reddit and X (formerly Twitter) expressed dissatisfaction with the new model's tone, which was too direct and cold. OpenAI therefore quickly restored GPT-4o for paying subscribers and introduced a model selector so users could switch between versions.

Altman explained that GPT-5 is now focusing on "warmer" responses that are not excessively flattering, in order to avoid reinforcing negative behavior among users. Nick Turley, vice president of ChatGPT, added that many people appreciate it when AI asks how they are doing, but the model must remain balanced. OpenAI worked with mental health experts to create a rubric for evaluating GPT-5's responses, ensuring that the AI pushes back against unhealthy tendencies. Altman estimated that fewer than 1% of ChatGPT users have an unhealthy relationship with the chatbot, which could nevertheless amount to tens of millions of people, and the company is holding internal discussions about it.

Despite the controversy, there was enormous growth: API traffic doubled within 48 hours of the launch, ChatGPT reached new user records—more than 700 million per week—and became the fifth-largest website in the world. OpenAI is now "out of GPUs" due to overwhelming demand, and tools such as Cursor have switched to GPT-5 as their default model.

Computing Capacity Constraints and Trillion-Dollar Investments

Altman revealed that OpenAI has even better models ready, but cannot offer them due to a lack of computing capacity. "Hardware is now the main constraint, not algorithms," he said. The company plans to spend "trillions of dollars" (equivalent to approximately CZK 23 trillion) on data centers in the near future to support billions of daily users and become a global utility-scale infrastructure provider. Altman compared the current AI boom to the dot-com bubble of the 1990s, when company valuations were "insane," but the technology ultimately justified the investments. According to him, AI is similarly transformative and warrants massive spending.

Plans for Social Networks and Much More

Altman hinted at expansion beyond ChatGPT. Incoming CEO of Applications Fidji Simo will oversee multiple consumer applications, including a potential AI browser that would compete with Chrome. When Perplexity's bid to acquire Chrome was discussed in connection with the antitrust trial against Google, Altman said: "If Chrome is really going to be sold, we should take a look at it." He did not provide a specific figure, but indicated interest if the court forced Google to sell.

Altman also expressed interest in developing an AI-powered social network. "There is nothing inspiring today about how AI is being used on social media," he said, wondering whether it would be possible to create "a much better social experience with AI." OpenAI also plans to back brain-computer interface startup Merge Labs, which would compete with Elon Musk's Neuralink. "We haven't closed the deal yet, but I would like us to," Altman said, describing it as an investment in a separate company where users could interact with ChatGPT through their thoughts.

On the hardware side, Altman spoke about working with Jony Ive on a new AI device. "It will be so beautiful that if you put a case on it, I will personally find you," he joked. According to him, it will be "worth the wait" and represents a "new computing paradigm" that does not emerge often. Altman praised Apple's products and hinted at continued collaboration between OpenAI and Apple.

A Look at OpenAI's Future

This dinner showed that OpenAI is now focusing on ambitions beyond models such as GPT-5. With growth on the scale of Alphabet (Google's parent company), the company plans to disrupt areas such as search, hardware, and social networks. Altman's openness about mistakes and massive investments offers a rare glimpse into the thinking of one of the most powerful people in the tech industry. Despite the controversies, OpenAI is growing astronomically, with trillion-dollar ambitions that could change the way we interact with technology.