OpenAI’s Free Users Were Its Biggest Problem. This Year They Started Paying.

The case against OpenAI has always been simple. About 900 million people use ChatGPT every week. Around 95% of them pay nothing. Every question they ask costs money to answer. This is why the company keeps losing money, as the paying customers alone cannot foot the bill for the remaining 95% of users.

That was true until this year. The company has now started showing ads to help close the gap between the free tier and the paying subscribers.

What Changed

OpenAI said in January it would start testing ads in the United States. The ads are shown to free users and to Go, its cheapest plan at $8 per month. Subscribers on Plus, Pro, Business and Enterprise do not see them. The ads sit below ChatGPT’s answers and are labelled as sponsored. On March 26, an OpenAI spokesperson said the test had passed $100 million in annualized revenue. That took less than six weeks, with more than 600 advertisers already signed up. The Go plan helps as well. It was launched internationally first and has only just arrived in the US. The plan that shows ads only just went live in the region it is worth the most.  

Most of the Audience Still Sees No Ads

An annualized recurring revenue of a hundred million dollars is small for a company this size. It must be noted, however, that only a small part of the total user base generates this. About 85% of free and Go users in the United States can be shown ads. Fewer than 20% see one on a given day. Under-18s are excluded. OpenAI has said ads will not appear near political, health, or mental health topics. So the money is coming from a small portion of just one country. A report from Axios suggests OpenAI could reach as high as $25 billion a year in advertising by 2028. Whether that number is reached or not remains uncertain, as these are only projections. But given how much of the market OpenAI has yet to target, a meaningful increase in advertising revenue seems probable.  

What Investors Have Paid So Far

OpenAI raised $122 billion in March at a valuation of $852 billion, the largest private raise on record. SoftBank, Microsoft, Nvidia, and Amazon all contributed. The company was valued near $300 billion in early 2025 and around $500 billion that October. The price has nearly tripled in a year, and the buyers are among the most informed in the industry.

Where the Chinese Models Actually Hurt

In mid-July, Chinese firm Moonshot AI released a model called Kimi K3. Moonshot says it still trails OpenAI’s GPT-5.6 Sol and Anthropic’s Claude Fable 5 overall. It beat GPT-5.5 and Claude Opus 4.8 on several benchmarks, including coding. The cost is what makes it dangerous. Bank of America put K3 at roughly half the cost of GPT-5.6 Sol. Moonshot also plans to publish the model itself on July 27, meaning any company can download it and run it on its own computers without paying Moonshot anything. 

That threatens the money OpenAI makes from selling access to its models. However, a cheaper model does not hand Moonshot 900 million weekly users. It does not build the habit, the brand, or the advertising space that comes with them. Google faced this years ago as other search engines started competing. But its advantage was never the technology alone. It was that everyone already used it. If frontier models become cheap and freely available, more of OpenAI’s value sits in the part nobody can copy.

There is a political layer too. Washington is debating whether American firms should use Chinese open models at all. That debate is live and unresolved, and nobody should assume how it ends. The consensus is that the government will create regulatory hurdles for enterprises to adopt the Chinese models rather than outright banning them. This will help protect the margins of the US companies, similar to what the government does to protect the margins of US EV makers.

What Happens When OpenAI Lists

OpenAI filed confidentially with the SEC on June 8. Reports say it is leaning towards a 2027 IPO rather than listing below the current $1 trillion mark. Whenever it happens, it will matter beyond OpenAI. There is no listed AI company of this scale. Once OpenAI trades, every other AI business gets measured against it. This is why the price matters so much to Sam Altman. He is not only selling shares, but he is also setting the number the rest of the industry gets compared to. 

The IPO will also settle multiple arguments. It could possibly reveal what OpenAI spends on paying customers, what it spends on everyone else, and whether advertising is growing fast enough to close the gap. At the same time, Wall Street will start asking tough questions on future revenue and growth, something OpenAI is already struggling with. The company missed its internal revenue targets earlier in the year, sparking concerns about whether it will be able to keep up with its spending commitments. Until then, the free tier is what needs to be watched. It has been the reason the numbers look weak. It is now the reason they might not anymore.