The US vs China AI War - and Everyone Else Caught in the Middle.
With Jez Rowe and Cam Lynch
In July 2026, China's Kimi K3 matched top US models at a fraction of the price, and US enterprises started switching to cheaper open source models. Jez and Cam unpack what that does to the idea that America leads AI, and whether a business can trust Chinese AI. Their answer: don't trust or distrust the model, look at where it runs.
What this episode answers
Is it safe to use Chinese open source AI models in my business?
Yes, if the model runs on your own hardware. Cam Lynch, founder of Vibe CFO, puts it simply: it's not about trusting the AI, it's about trusting the server it sits on. A Chinese open source model running on your own computer or server sends nothing out, it just processes tokens locally. The risk is using a Chinese-hosted service, where confidential data passes through servers you don't control and is very likely being saved.
What is the difference between running an AI model locally and using a hosted AI service?
Running a model locally means your data never leaves your machine, while a hosted service sends your data to someone else's server. Open source models can be downloaded and run on your own computer or servers, where the model is just a program processing tokens. A hosted service like a chat app or API processes your prompts in the provider's data centre. For confidential client or financial data, where the model runs matters more than who built it.
Why are companies switching from OpenAI and Anthropic to open source AI models?
Companies are switching because frontier AI gets very expensive at scale. Token costs grow with every user, so giving thousands of staff access to Anthropic or OpenAI can run into millions of dollars. In July 2026, US enterprises were pulling back and moving to open source models, including Chinese ones, that they can download and run on their own servers without a per-token bill. People choose things that work and are cheap.
What is Kimi K3 and why does it matter?
Kimi K3 is a Chinese AI model that, in July 2026, matched Anthropic's Opus 4.8 at roughly 60 to 70% lower cost. It ranked roughly the fourth best model available, competing with Claude Fable 5. It matters because it ends the story that China is well behind in AI. At around 1.6 trillion parameters it's too large to run on a normal computer, so most users reach it through Chinese-hosted servers.
Is the US still ahead of China in AI?
Not clearly, and the gap has closed fast. By July 2026, Chinese models like Kimi K3 were matching leading US models at a much lower price, after three months of enormous progress. That matters beyond tech. Much of recent US economic growth has come from AI companies, so cheap, capable Chinese models have the potential to shake the valuations that growth rests on.
Episode notes
Kimi K3 just landed out of China and gave the US-China AI race a fresh jolt. Two superpowers, endless compute, and a frontier owned by exactly two countries. So Jez and Cam step back from the horse race and ask the bigger question: what about everyone else? Every business and every country outside those two is now running on AI it doesn't own and can't control, and this year we all got a reminder of what that means when access can be switched off from the other side of the world.
Five minutes. One thing that matters. We ignore the rest, so you can too.
Vibe CFO is Australian founded. An AI CFO for small and medium business, out of the box, governed and secure.
Vibe in 5.
Transcript
Read the full transcript
Ah, yet another big week in AI. This is The AI Exhale, hosted by Vibe CFO with Jez and Cam. Five minutes, one thing that matters, or maybe two. We ignore the rest so you can too.
Jez: So for today's episode, we're going to cover a little bit around the global tech war. I think it's exciting, for one. You know, we try and cut through the newsworthy items that come out each week. We pick one. We break it down, we make it simple, easy to digest. Let's talk about Kimi K3, Cam. There's something tiny out of China that's, well, maybe not so tiny now, and what America might be doing in reaction, as it seems to be quite a reactive time for what have been the dominant players, Anthropic and OpenAI, out of the US.
Cam: Oh, the reason this matters, right, is because the story for the past 3 or 4 years is that China's well behind, and the models aren't that good, and so they're not relevant.
The storm had already been coming, but the reason Kimi K3 is big is because it matches Opus 4.8. And it's like 60 or 70% cheaper. It's like the 4th best model. It's right up there taking on Claude Fable 5. So, you know, this whole thing that America leads the frontier game has changed. No way is that story true anymore. So this just massively reshapes the way things happen. Because you can get Kimi K3, right, and this is what is happening in America right now. All these enterprise companies have pulled up on this whole token maxing because they're like, this is costing me a fortune.
If you're an enterprise using Anthropic or OpenAI and using their system and giving thousands of users access to AI, we're talking millions and millions of dollars. And they have pulled up stumps, and now they're switching to Chinese models and using them open source, because they're open source, so they can download these models and put them on their own servers.
This is big. The reason it's big, Jez, is because the American economy is propped up on AI right now, right? SpaceX and OpenAI and Anthropic and Google and all the things they're doing.
The economy is being held up. All the growth has been coming from AI. So if you pop this bubble, right? And I feel like these Chinese models have got the potential to pop the market here. It's a big moment. There's lots of wacko stuff we're hearing in the market. Like, Trump is gonna take 5%, like the actual American government's gonna own 5% of OpenAI. That was the current rumour. The other rumour is that the American government's gonna ban Chinese open source models from being used in America.
Things are happening right now, and the movement in the last 3 months of Chinese models is just enormous. And guess what? It's sort of like, what's the name of the Amazon guy, the founder?
Jez: Jeff Bezos.
Cam: Jeff Bezos. I mean, he basically said this 20 years ago. He said people like things that work and are cheap. And that is like, right? People are drawn to cheap that works. And that is what the Chinese models are doing, which is what makes it so enticing. So look out, it's big.
Jez: Well, I mean, we saw Spain have $51 mil taken off their World Cup win paycheck by the US. I mean, that's the IRS. That's what they do. Just to that point, but you know, Cam, the whole Chinese car came in, right? And we could buy that, and we were worried it might not work. I mean, speaking in general here, I think there's probably something similar in people.
And Kimi K3 has exploded in just the last 1 to 2 weeks. But in terms of, can I trust it? As you say, I'm an enterprise, I'm a director, I've got a job, I can't do something stupid here as head of cyber. Or I'm a small to medium business owner and I definitely can't screw up adopting something that's going to, you know, put me in the sh*tter.
Cam: It's not about trusting or distrusting the actual AI. It's about trusting where the AI sits, like what server it sits on. So Kimi K3 is such an enormous beast, right? It's, I think, 1.6 trillion parameters in size. You can't just go and put that on your computer and use it, right? It's hosted on Chinese servers. And that's the problem. The problem is that people put confidential data through the server, and the Chinese are almost certainly saving all your data.
And that's the conflict of interest. Whereas the lower models are smaller and you can fit them on your own servers. And so if you're running a Chinese open source model on your local computer, don't let that fear you. That's a completely different question. Having a Chinese model on your local computer, there's nothing leaving it. It's just a program that's running tokens inside it.
That's not the danger. And that's what the enterprises are looking at doing. They put Chinese open source models on their servers and they run them for free. And there's actually no risk on that front, in the sense that the data's not going to the Chinese Communist Party the same way. Does that make sense?
Jez: Yep, it does to me. I think for some listening, you know, we probably should do an episode in the future that breaks down, as we move further down this path, how to run AI in the business.
And that's your five-ish minutes back. Vibe CFO, Australian founded. An AI CFO for small and medium businesses, out of the box, governed and secure. vibecfo.ai. Now you know.
Now you know.
The AI Exhale is made by VibeCFO. Australian founded, an AI CFO for small and medium business, out of the box.
