Today, OpenAI launched its new Atlas web browser in a surprise livestream. The show started with CEO Sam Altman himself, speaking directly to the audience.
“We think AI represents a rare, once-a-decade opportunity to rethink what a browser can be,” Altman said. “In the same way that, for the previous way people used the internet, the URL bar and the search box were a great analogue, what we’re starting to see is that the chat experience and the web browser can be a quick analogue.”
It was an inspiring note, in the classic Steve Jobs mode. But even more important than Altman’s browser was the detritus he was sweeping aside to make room. It wasn’t just casting present-day browsers as old, but part of a whole package of goods that are about to be replaced by AI — as Altman put it, part of “the previous way people used the internet.” And most of those soon-to-be obsolete services trace back to a single company: Google.
OpenAI’s browser project has been an open secret in Silicon Valley since at least this summer — and it was clear from the beginning that it would be a potential threat to Google, current owner of the world’s most popular browser. But Tuesday’s product and presentation details made it clear exactly how much the web giant has to lose in the AI era — and how little the Google’s success with Gemini seems to have helped.
The immediate threat is simple enough: ChatGPT draws 800 million users a week, and if those users switch to Atlas, they’re most likely switching away from Chrome. Losing those users doesn’t have an immediate dollar cost for Google (it’s a free product, after all) but it limits Google’s ability to target ads to those users or nudge them to Google Search — a particular sore point because, just last month, Google was barred by the US Department of Justice from making any search exclusivity deals.
Then, there’s how OpenAI deals with search itself. AI has already strained the search model of the web, surfacing processed information instead of content that can be advertised against. But on OpenAI’s livestream, Atlas head of engineering Ben Goodger (himself a central figure in developing both Firefox and Chrome) described the new kind of chat-oriented search as a paradigm shift.
“This new model of search is really powerful,” Goodger said. “It’s a multi-turn experience. You can have this back-and-forth with your search results instead of just being sent off to a web page.”
Techcrunch event
San Francisco
|
October 27-29, 2025
Of course, Google has done a lot to integrate AI into the normal search experience — but the company has mostly approached it the same way as product listings or reviews: by adding a box to the results page. But OpenAI’s kind of engaged back-and-forth is beyond anything you can get on Chrome, and given its profoundly different approach, it’s not something that can be easily copied. If OpenAI’s search interface proves popular, it could be a serious threat to Google’s dominance.
Then there’s the advertising question. OpenAI doesn’t serve advertising at the moment, but it has been careful not to rule it out. The company has also been listing a lot of adtech jobs lately, fueling speculation that an ad pivot might be on the way. With Atlas, ChatGPT can now collect context directly from a user’s browser window — providing a lot of extremely valuable data for ad targeting. It’s an unprecedented level of direct browser access: literally looking at the words on your screen as you type them. And after decades of privacy scares, it’s not the kind of sensitive information that users are likely to give to Google or Meta.
It’s still early days for Atlas and a lot will depend on the product itself — and whether users really want what OpenAI is offering here. But the company has plotted a surprisingly commercial path here, one focused on user and revenue growth rather than hazy ambitions around AGI. As infrastructure wonks ponder the $300 billion question of whether OpenAI’s revenues can ever live up to its enormous data center buildout, products like Atlas may be the first place to look for an answer.
