OpenAI Lost 6% of Users To Google's Gemini 3 In A Week; No Profit Until 2030

It has been clear for over a generation that things move fast in tech - but they may be accelerating. A month ago many headlines were asking whether OpenAI was too big to fail. Now there are questions about whether it can survive.

Reports that Google was openly and successfully challenging OpenAI appeared just over a week ago. Data now reveal that OpenAI lost as much as 6% of its users to Google's Gemini 3, especially surprising since commentary earlier in the year suggested Google itself was roadkill in the face of AI. It also seems that Anthropic's Claude is emerging as the corporate AI of choice, challenging OpenAI's dominance as many consumers will likely convert to whichever brand their employers' use. And as if that weren't enough, OpenAI itself acknowledged that it would probably not be profitable until 2030, a lifetime in tech terms. As worrisome, the company's financial model assumes it will have captured 44% of the world's non-Chinese users by then, a figure which seems...aspirational...JL 

Joe Hindy reports in Mashable and Pyush Shukla reports in Economic Times:

OpenAI lost 6% of its traffic since Google's Gemini 3 launched. ChatGPT went from 203 million average daily visits to 191 million, meaning OpenAI lost 12 million people per day over the last week. Some of that may be due to Thanksgiving, but the problems started months ago when GPT-5 was released. Criticism was so intense that OpenAI brought back GPT-4o. In addition, OpenAI may not make money until 2030. New financial estimates suggest the company could face a $207 billion shortfall. OpenAI's model assumes 3 billion users by 2030— 44% of the world’s adult population outside China—and a paid conversion rate rising from 5% to 10%.The question many investors are asking is: Can OpenAI’s revenue ever catch up to its infrastructure bill?

Recently, Google introduced Gemini 3, which the company called its "most intelligent" AI model yet. So far, the AI leaderboards agree, with Gemini 3 surging ahead of rivals like ChatGPT at LMArena. Now, OpenAI head honcho Sam Altman has reportedly declared "code red" at the company in response to increased pressure from competitors. 

The announcement came in the form of a memo that was leaked to The Wall Street Journal. The contents of the memo are largely unknown, but the message was direct — OpenAI has some serious competition in the likes of Gemini 3 and Anthropic’s Claude, and the company needs to do something about it.

 

According to a post on X from former Googler Deedy Das, OpenAI has lost nearly 6 percent of its traffic since Gemini 3 launched. The data is sourced from SimilarWeb, which says that ChatGPT went from 203 million average daily visits to 191 million. If that data is accurate, it means OpenAI lost about 12 million people per day over the last week. Some of that may be due to Thanksgiving, but it's a big shift. The problems started months ago with some slips from the reigning champion of AI chatbots. When GPT-5 was released earlier this year, it was immediately criticized for its less friendly tone compared to earlier models. Criticism was so intense that OpenAI ultimately decided to bring back GPT-4o. The data shows that GPT-5 is more capable than its predecessors, but its lack of personality irked hardcore users who used the bot every day. Plus, OpenAI has been walking a tightrope to balance functionality and safety, which it has not done particularly well. 

 

Despite the criticisms, it seemed OpenAI was destined to keep its top spot among the AI chatbot competitors. However, Google released Gemini 3 in late November to nearly universal praise, with Google’s offerings topping nearly all of the benchmark leaderboards. Due to the sheer size and reach of Google, the company has a significant advantage in accessing training data, the lifeblood of AI models.

 

The competition is also ramping up on the enterprise side of things. Anthropic boasts more than 300,000 business customers as of September 2025, which includes some Fortune 500 companies. 

 OpenAI may not make money until 2030 — and the gap is massive. New financial estimates suggest the company could face a $207 billion shortfall even as ChatGPT continues to dominate the AI market. The growth is explosive, but so are the costs behind training and running large language models. And now, the question many investors are asking is simple: Can OpenAI’s revenue ever catch up to its infrastructure bill? 

Despite ChatGPT’s popularity, OpenAI is still operating without profitability. Forecasts show revenue could grow to more than $213 billion by 2030, driven by more paid users and expanding enterprise adoption. But those numbers still fall short of what OpenAI must spend to keep its AI running at scale. The biggest cost is not product development — it’s compute power. 

HSBC’s semiconductor research team, led by Nicolas Cote-Colisson, says OpenAI’s long-term commitments to cloud and AI infrastructure will reshape the company’s finances. The company has signed a $250 billion deal with Microsoft and a $38 billion deal with Amazon, part of a plan to secure 36 gigawatts of compute capacity by 2030.

HSBC estimates these commitments translate into $792 billion in cumulative data-center rental costs between late 2025 and 2030. If that capacity runs at full use, the annual rental bill alone could reach $620 billion. Total compute needs may hit $1.4 trillion by 2033, aligning with Sam Altman’s public comments about long-term requirements. 

OpenAI’s revenue is expected to soar. HSBC forecasts over $213 billion in annual revenue by 2030, driven by subscription growth, enterprise AI services, and a capture of roughly 2% of the global digital ad market. The model assumes 3 billion users by 2030—equal to 44% of the world’s adult population outside China—and a paid conversion rate rising from 5% to 10%.

These are all likely reasons why OpenAI is in code red; losing that many visitors is certainly noticeable. Of course, OpenAI is hardly staying still. The company is working on tons of initiatives, including inking infrastructure deals with the likes of Nvidia, Oracle, and other companies. 

In November, Sam Altman wrote a long post on X addressing fears of an AI bubble and concerns that OpenAI is overextending itself.