Just as rumors were circulating in the market that OpenAI was planning an astonishing "hundred-billion-dollar funding round," the company's CEO, Altman, stepped forward to launch a new round of evangelism on the "AI narrative."
Early Friday morning Beijing time, Altman's interview on the popular podcast Big Technology was released. The OpenAI CEO discussed recent competition with Google, future product plans, and the economic implications of the "trillion-dollar investment."
Google's moat is also its weakness.
Regarding OpenAI's declaration of a "red alert" following the release of Google's Gemini 3, Altman stated that this situation is not uncommon within the company: it's always better to remain vigilant and act swiftly when potential threats emerge. Typically, such situations don't last long, usually six to eight weeks is sufficient.
He also revealed that DeepSeek had triggered a "red alert" from OpenAI earlier this year .
That said, Google still possesses one of the strongest distribution channels and business models in the tech industry. However, Altman believes that precisely because it is unwilling to easily relinquish this advantage, it becomes Google's weakness.
He said, "Google probably has the best business model in the entire tech industry, and I don't think they'll give it up easily. But with artificial intelligence... " Forcing it into web search... I don't think that will be as effective as completely redesigning the entire system.
Altman further explained that while it would be slightly better to cram an AI into a messaging app that can summarize information and draft replies, that wouldn't be the final form of the product. The same applies to search and office suites. In the future, we are likely to see new products built entirely around AI in major product categories, rather than simply piecing AI together.
The head of OpenAI used messaging applications as an example, saying that he doesn't need an AI that can summarize messages and draft them. Instead, he wants an AI that understands him and the people he's communicating with, to "handle everything it can," provide feedback every few hours, and then pop up a notification when it's needed to intervene. This is completely different from current chat applications.
Regarding new products: New models are coming in the first quarter of next year, and AI hardware is also on the way.
Altman stated that the release date of GPT-6 is currently unknown, but a new model with significant improvements over GPT-5.2 is expected in the first quarter of next year .
He also stated frankly that what consumers want most right now isn't a smarter model, but businesses still want larger models to have higher intelligence. Therefore, the company will improve the model in different ways for different uses, with the goal of making everyone like the new model more.
Altman also discussed OpenAI's main approach to developing "AI hardware".
He said, "I think the way people use computers will gradually change, from something that is somewhat passive and merely reacts, to something that is very intelligent and proactive ... The existing devices are not suitable for that kind of world."
Altman believes that the screens on electronic devices today are designed to confine users to the interaction methods that have existed for decades, while keyboards are designed to slow down the speed of input.
Concerns about an "AI bubble": This is a good thing.
OpenAI, with its annualized revenue just reaching $20 billion, has already placed orders for $1.4 trillion to build data centers . Altman reiterated that the biggest problem now is that computing power is not growing fast enough. Providing the energy, chips, systems, and networks needed for trillion-dollar data centers will take time. More importantly, demand is currently sufficient; if computing power could double, OpenAI's revenue would double .
Altman explained, " Our computing power has roughly tripled in the past year. We plan to triple it again next year , and hopefully continue doing so after that. Revenue growth is even slightly faster than computing power growth, but overall it's in line with the scale of our computing power . So far, we haven't encountered a situation where we can't fully monetize our existing computing power. If our computing power doubled, I think our revenue would have doubled by now."
Regarding the difference between "20 billion in revenue" and "1.4 trillion in expenditure," Altman emphasized that this comparison ignores the implications of exponential growth.
He said, " We believe we can maintain a steep revenue growth curve for a considerable period of time, and everything we're seeing right now supports that judgment , but we can't do it without computing power. Our computing power is extremely strained, and that has a huge impact on our revenue line."
Altman also stated that the current skepticism of capital markets towards AI investment is a good thing, as the market did indeed go crazy earlier this year.
He recalled, "We met with a company, and the company's stock price rose by 15% or 20% the next day. That was very unhealthy. I'm actually glad there's more skepticism and rationality in the market now, because in my view, we were heading straight for a very unstable bubble."
However, Altman also emphasized that there is nothing inherently wrong with building data centers through financing or other financial innovations. Not only will the capabilities of large-scale models be much better in the future, but even if AI capabilities remain at the GPT-5.2 level, there is still a huge amount of economic value that has not yet been realized.
The head of OpenAI also admitted to "miscalculations" regarding the progress of AI applications. He once believed that as long as more capable models were introduced, the world would quickly figure out how to use them. But the reality is that, apart from specific groups such as programmers and scientists, most people's questions about AI are still stuck in the GPT-4 era, where there is a huge excess of potential.
(Article source: CLS)
