The AI industry is forking into two paths. They are non-binary and actors can be in both at the same time, but it was expected that we would see this branching:
- Products aimed at the general consumer. ChatGPT wins there and Gemini is growing lately.
- Tools for companies. Gemini and Copilot from Microsoft stand out more there, but Anthropic is growing a lot thanks to Claude Code.
This division also marks something else: who is going to capture the real economic value of AI.
Why is it important. Companies pay up to $200 per month per employee for the best model. ChatGPT “domestic” users, one tenth. And in business environments, the difference between the best and the second best matters a lot. At least much more than in domestic environments.
Yao Shunyu, who worked at OpenAI and is now at Tencent, sums it up: “If your salary is $200,000 and you have 10 tasks a day, an excellent model does eight or nine. A weaker model does five or six. And when you don’t know what those five or six are, you waste time monitoring it.” according to the newsletter ChinaTalk.
The contrast. “If you compare the Today’s ChatGPT with the one from a year agothere’s really no perceptible difference,” Yao points out. “On the other hand, AI-assisted programming has already changed the entire coding industry. “People no longer write code, they talk to their computer in natural language.”
Most ordinary people still use ChatGPT as a kind of enhanced search engine. In companies, more intelligence directly means more productivity with a clear economic value.
Between the lines. Anthropic has bet everything on this. Claude Code has changed the way developers work. And now just launched Coworkwhich seeks to bring that same idea to office workers, outside of programming. They are not going after occasional users: they want entire teams that depend on AI to work.
OpenAI dominates in adoption figures and brand recognition, but it has a problem: many users who pay little on average. Companies are increasingly looking for tools that truly improve productivity.
The threat. In the business market, whoever has the best model wins. Companies will always pay for number one if its price is comparable to the rest of the proposals. In consumption, something decent is enough and the price sensitivity is greater.
And OpenAI needs a lot of money. Training and operating these models costs a lot. The consumer market has a fairly low profitability ceiling, and that is why they seek to shore it up with advertising revenue and pointing towards those of affiliation.
And now what. This year we will see it clearly:
- OpenAI needs to prove that its enterprise agents are worth the money.
- Anthropic will continue to refine its position in code and productivity.
- Google is a little late but has hit the nail on the head with Gemini 3.
At the end of the year we will know if OpenAI’s generalist strategy works or if the AI business ends up divided between those who dominate the office and those who dominate the couch.
Featured image | Anthropic, OpenAI, Xataka
GIPHY App Key not set. Please check settings