You’ve probably noticed the sponsored links creeping into your ChatGPT responses. It feels cheap, right? Like watching a brilliant professor suddenly stop mid-lecture to pitch you a timeshare.
But while OpenAI turns its chatbot into a billboard, Chinese AI giants like DeepSeek, Qwen, and Zhipu are pretending they don’t see the ad money on the table. Instead, they are deliberately bleeding billions to build something far heavier.
OpenAI is turning your AI assistant into an attention merchant; Chinese AI firms are turning theirs into a utility meter.
Let’s look at the numbers. Sam Altman used to hate ads. In 2024, he called them a ‘last resort.’ By 2026, ChatGPT Ads is raking in an annualized $1 billion. They are even testing ‘sponsored agents’—where clicking an ad drops you into a brand-curated AI experience. Why the pivot? Because 92% of ChatGPT’s users don’t pay a dime.
When your free user base hits a billion, you’re no longer a tech company. You’re a media empire with a massive, unpaid audience.
Meanwhile, Chinese AI firms are taking a totally different path. DeepSeek, Kimi, and Zhipu aren’t selling your attention. They are selling tokens. They are selling API calls. They are treating AI like water and electricity—metered, billed, and locked into enterprise infrastructure. DeepSeek’s API revenue is up 10x year-over-year with an 82.9% gross margin. Zhipu’s API revenue exploded by over 2,700%.
But here is the twist: The Chinese players aren’t avoiding ads out of some moral high ground. They are avoiding ads because they have no choice.
Chinese AI startups aren’t avoiding ads out of principle. They’re avoiding them because their corporate parents already own the entire attention economy.
Think about it. If you’re ByteDance, you already own TikTok and Douyin—the most sophisticated ad engines on earth. You don’t need your AI chatbot, Doubao, to sell ads. You need it to sell cloud compute. The same goes for Alibaba and Baidu. The ad ecosystems are already spoken for. The AI firms are forced into the utility ghetto.
And the cost of this structural choice is brutal. Alibaba’s Qwen division posted a single-quarter EBITA loss of $1.3 billion. ByteDance’s Doubao is losing tens of millions a day on free users. They are subsidizing massive C-end losses with the hope of B-end lock-in.
They are bleeding billions to give away AI for free, hoping to lock enterprises into their infrastructure before the money runs out.
Neither model has reached the profit turning point. OpenAI compromises the neutrality of its AI for fast ad revenue, risking the trust that made ChatGPT great. Chinese firms burn cash to subsidize free users, praying that API margins will eventually cover the hole.
For investors, builders, and users, this divergence dictates the future. If AI becomes ad-supported, your data and attention are the product, and the AI’s answers will always be compromised. If AI becomes a metered utility, it becomes infrastructure—reliable, but locked in and expensive.
The AI revolution isn’t about intelligence anymore. It’s a bare-knuckled fight over who owns the tollbooth.
OpenAI is selling your eyes. China is selling the plumbing. Both are burning the furniture to keep the house warm.
We are watching a trillion-dollar bubble float on a foundation of free users and unmet margins. Eventually, someone has to pay the bill.
FAQ
Q: Isn't OpenAI's $1 billion ad revenue proof that the ad model works?
A: It proves they can sell ads, but it masks the fact that 92% of users don't pay them. It's a band-aid on a massive cash bleed, not a sustainable foundation for a trillion-dollar infrastructure play.
Q: What does this divergence mean for everyday AI users?
A: If you're not paying for the AI, you're either the product being sold to advertisers (OpenAI), or you're a loss-leader being used to capture B2B infrastructure contracts (Chinese firms). Free AI is a temporary illusion.
Q: Is the Chinese utility model actually better than the ad model?
A: It's structurally forced. Their parent companies own the ad markets, so they have no choice but to sell plumbing. It's slower and burns more cash, but once a business is locked into their API, that cash flow is far more deterministic than an ad click.