Ep 886: The U.S. vs China AI Cold War Is Starting: What It Means and How It Impacts You (Replay)

A US versus China AI cold war is starting, and most business leaders have no idea they may already be participants. And it's moving much faster than hardly anyone can keep up with. About 2 weeks ago, Reuters reported that China is considering locking down its most advanced AI models, keeping its best from the rest of the world. Then just yesterday, the Trump administration is signaling it may restrict Chinese AI models inside the United States. Yeah, read that again, because both sides are now moving to wall off AI access within days of each other. And here's why 2 governments suddenly care so much about who runs which model. Well, that's because this was never really about who has the most powerful AI or who has the best benchmarks. It's actually about economic growth, geopolitical power, and who sets the standards the rest of the world builds on. That's because whoever leads AI doesn't just win a tech race towards superintelligence, they gain leverage over every other country's future. And that's the fight your business may now be caught in the middle of, whether you signed up for it or not. So let's dig in. Here is the big picture. The AI power is flipping right in front of our very eyes. That's because for 3+ years, the US has been pretty far ahead, at least when it came to the open-source Chinese models. But now China is actually closing the gap with open-weight models. Yeah. Closing the gap between the best that Anthropic and OpenAI have to offer. But now both governments are treating their frontier models as strategic weapons worth potentially restricting. And US businesses that have been chasing models and maybe using these open-source Chinese models because they were very capable now might accidentally inherit some serious geopolitical deployment risk. So on today's show, you'll learn how these Chinese models became cheaper, closer and harder to ignore. You're going to know why a trillion parameter— yes, trillion with a T— a trillion parameter open model no longer means that you can simply run it on your computer. You're going to know why Microsoft testing Chinese models could signal the enterprise AI future. And you're going to know what leaders should deploy, document, and avoid before restrictions may harden. All right, let's get into it. Welcome to Everyday AI. My name's Jordan Wilson, and we do this every day, and it's yours. This is your unedited, unscripted daily livestream podcast and free daily newsletter, helping business leaders like you and me keep sense of all of these developments because yeah, they're happening at warp speed. I tell you what matters, what doesn't. You take that information and you're the smartest person in AI at your company. So it starts here with the podcast, but make sure to go to our website at youreverydayai.com. We're gonna be recapping the highlights from today's show and a whole lot more. Let's get straight into it. What the heck has happened? I mean, my gosh, as someone that's been doing this Everyday AI thing for 3 and a half years, the last like 5 or 6 weeks, at least when it comes to the US versus China tensions and even what each country is pushing out, Aside from just AI acceleration is at an all-time high, the battle between these two nations couldn't be any higher. So, like I said, in the past two weeks, we saw reports first that Beijing is looking at curbing overseas access to its top AI models, and then we got a report just yesterday saying the Trump administration may also ban Chinese AI models, and that's. Big news. That's because we've seen a lot of reports. I've talked to people personally that are moving entire enterprises off, you know, maybe like an OpenAI or Anthropic or Google models, right? And on to other Chinese models. I think especially when we saw GLM 5.2 from ZAI, which was technically a more affordable, you know, version of some of the frontier models, it wasn't quite yet punching at the state-of-the-art kind of AI class, but it was getting close. That's when it kind of started, but it's really snowballed just the past week with models like KIMI K3, which is now just under the Fable 5 and GPT-5.6 SOL class. And then Quen, their new 3.8 model, which we don't have full benchmarks on yet, but It's released, which is weird, right? Normally companies don't release models and not release benchmarks, but it's live and it's seemingly really good. And it also may be entering that top tier. So in that upper echelon now, we might have models from Anthropic, OpenAI, and 2 open-weight Chinese companies. So in theory, these are models, right? Again, you can't really run them on your computer, but these are models that large, large enterprises that have the means to do it, They can run all of these things locally, again, assuming they have a couple servers to throw on. So here's why AI leadership now decides the global superpower status. It's 3 things, right? It's the economy, power, and this multi-access. Okay, so here's what each of those boils down to. Whoever automates thinking work fastest compounds growth over every rival nation, and that helps whoever's actually winning AI have a leg up on the economy. Power, I mean, aside from, you know, these systems, these AI systems, because that's what they are, they're much more than models, right? They're being used in military use cases. And I've been talking about this since the very beginning, right before we had models that should have even been touching the battlefield. I've said the future of AI is definitely— it is going to become the new oil, it is going to become the new gold. And I think that most people that hang out on the bleeding edge understand that to be true. Maybe the rest of the world is— we'll see that in a year or two. But that is the reality. If you control AI, if you control You know, if you are in the lead toward AGI or artificial superintelligence, that means that your country will wield a, or yield a power over every other country, and there may not be much that anyone can do about it. I mean, when you think about things like being able to, you know, put cyber attacks and being able to, in theory, take down entire countries' power grids, their banking systems, right? That's what we're at. It's things that physical weapons, You know, would try to do, right? But this is something that could in theory be launched autonomously at scale. That's the down and the ugly side of AI. But that's ultimately anything as powerful as artificial intelligence and these models that we have now. You have to think it's much more about, you know, it's about much more than helping us all write better emails or helping us triage our days better. There is a bigger and sometimes badder purpose between, um, behind what nations, especially the, uh, nations jostling for power at the top of the global, uh, pyramid, uh, this is what they want it for. And then last but not least, it is this, um, this parity. There is this, uh, parity that's happening right now because right now no nation leads in every axis, right? When it comes to model capability, uh, available compute costs, adoption and deployment, uh, right? And I think that's one of the things that China is really working on. And just FYI, as I'm talking about this, I'm obviously right, um, I'm based in the US. Most of the people and companies I work with are based in the US, so I'm obviously coming at it from that perspective, if you couldn't tell already. So one of the biggest questions is like, why? Why does China— have they been coming with this open source or open weight approach, um, and the US just really hasn't? Well, first You know, some recent models from the US have done fairly okay on the open source scheme. So Thinking Machines Labs' new Inkling model, really good. You know, NVIDIA's open models, fairly good, but no one's been able to touch the Chinese models. Part of that is because of distillation, which we'll get to, but it always gets to why would people always ask, why would China put out these open source models that You know, at least 6 months ago, you know, you could download them on very powerful consumer hardware and run them, and everyone was always confused. And I think that there's a couple of reasons, but one, China wants to be the default on what the rest of the world builds on, because if so, that makes all the other services that you might need to run those models more valuable. And every enterprise, here's the thing, it is a competition. Every enterprise that switches from US models drains the revenue funding essentially Silicon Valley's next training runs, right? So if you take the fuel out of the car, the car can no longer run. Also, huge models make enterprises rent Chinese-compatible hosting tools and support, and then China gains adoption and they weaken US pricing power and they keep the leverage. The other thing that most people don't talk about is controlling the narrative, something that China is obviously very concerned about. And has been concerned about for many decades. But these Chinese models, right, studies have shown that they avoid sensitive topics that Beijing does not want discussed globally. So people are just sometimes copying and pasting whatever an AI model spits out, and they're maybe sending it to colleagues, they're sending it to clients, or in many cases they're just putting it on the internet, right? And then large language models start to regurgitate this and this becomes part of the training data. So this is a way, and you know, This gets, you know, put out in schools, media, government documents, everything. So, and you also have people using these models to distill and create other models, right? As an example, I believe Cursor's KIMI, I believe Cursor's models, their first ones that were based off of KIMI's open source models. So why does that matter? Why does China want to control the narrative? Well, a Stanford study even found that China-origin models answered political questions less directly. So it's not like, you know, these models are going to say something that, you know, overtly slams the US or overtly, you know, puts, you know, Chinese, you know, morals and ethics on a pedestal. That's not what I'm saying. But it's just the nuance. It's the describing things in a slightly different way. And if you think of the game of telephone, right? Each time that happens, each time someone just blindly copies and pastes something on the internet, and then the next round of frontier models get trained on that information, and it just starts to weaken and distill maybe certain talking points that Beijing would rather not be out there. So there's, it's much more than just about controlling the what the rest of the world builds on and maybe sucking the US's power supply dry. It's also about controlling the narrative. All right, so let's talk about some of the more recent models. So I think this all started earlier this summer or late spring with ZAI's GLM-5.2. So that is not nearly on the same tier as the most recent ones from this past week, and that's Moonshot's Kimmy K3. And Alibaba's Qwen 3.8 Max. And they essentially on some things undercut US pricing, especially GLM 5.2. And in many cases they have been good enough, right? We've literally read stories where, you know, I would say more tech-forward or AI-native companies, but large ones essentially took their Claude spend, right? Because Anthropic's models are the most expensive. And there was a period, right, where Anthropic's Mythos 5, Fable 5 came out before OpenAI released their competing model in GPT-5.6. So there was this time and period where Anthropic had a lead in the quote unquote model wars, but it was just ridiculously expensive. And everyone's like, wait, we could use a model like GLM-5-2, which isn't that far behind their Opus class model, and we could get like 95% of the power of like an Opus model for a fraction of the cost. So they were just saying they're trying to undercut US pricing with good enough models, not necessarily by being the number one model in the world. And that's kind of where we stand now. And I said this on a show earlier, I think previously Chinese models were like 6 months behind. Now it's like 1 to 2 months. Part of that is I think their distillation efforts and their architecture under, right? It's not just distillation. Obviously these labs have some of the most talented engineers in the world. So it's a combination, I think, of, you know, number one, their distillation efforts have increased. Number two, some of their architecture that they're putting out is truly good and novel and making a difference. And well, Number 3, which you can't overlook, is the recent kind of US sanctions that have been handed down largely because of Anthropic to all model providers here in the US, which is delaying— we've seen reports maybe like a 30-day, 45-day or more, whereas these companies would have been pushing these models out to the public a little faster. But now that we live in permission slip AI land, you know, it's a combination of those 3 things that have kind of closed this gap. That was much wider before. But it's the benchmarks. We do have to talk about the benchmarks. All right. So if we look at the Artificial Analysis Index, which we talk about— sorry, the Artificial Analysis Intelligence Index, which we do talk about a lot on this show, you know, now all of a sudden, you have KIMI K3 in the same tier, right, almost as Claude Fable 5 and GPT-5.6 SOL. So on this benchmark, which is probably the best single overall benchmarks because it is a conglomerate, right? Claude Fable 5 has a 60, GPT-5.6 Sol has a 59, and Kimi K3 has a 57, right? And there is usually always like at least a 10% drop-off, uh, right, in terms of the AA index. Now it's like a 5% drop-off. Right. Which is not that big of a drop. So here's though where it's changed, especially the past like couple of models with ZAI's GLM 5.2. This is something that, yeah, you could run it a slowed down, a watered down version of that model if you had a super powerful consumer PC, right? Or you couldn't actually run it at full speed. Nothing like you could run online, but people out there that spent maybe way too much money on their consumer setups, they were able to run essentially watered-down versions of GLM 5.2. Not anymore, right? Because now it seems like the next step that China is taking, they want to compete on the frontier, the frontier frontier. And as the US models, the state-of-the-art models get more and more capable, well, it gets harder for, you know, a small enough model to close that gap. Because now what we're seeing is these Multiple trillion parameter models. So open source, right, especially through 2025, generally meant, okay, you could get a quant version of this model, right, which, you know, only activates certain parameters. So think about 2, you know, for context, right, the GPT-4 family of models was like 2 trillion parameters. So now you have Open source models that are bigger than that, right? Which is crazy. So these multiple trillion parameters, you can't run them. You literally need a small data center. Or like I said, if you are an enterprise company that has access to compute, yeah, you're going to need a whole rack of NVIDIA GPUs to actually run these things. So this does even change, I think, how most enterprise leaders should be viewing open source. It's almost like you should be saying like, what kind? like consumer open source or enterprise open source, because there really wasn't that distinction. There really wasn't that, you know, multiple tiers, you know, because if someone chained together a couple Mac Studios a year ago, you could probably run a quant, a quant version of these, you know, Chinese open source models. But they're not Exactly cheap anymore either. So with that distinction or moving away means they're not exactly cheap. And so for me, when I'm looking, right, if I'm a business leader, well, I am, right? But I'm not necessarily making enterprise decisions at Fortune 500 companies, although I do advise those type of companies. Before, even 2 or 3 months ago, You say, yeah, look at open source models as an alternative. Today, I don't really know why people should. And maybe that's largely because of what OpenAI has been able to do, right? In terms of cost per intelligence index, right? Which I think is just as important as the overall intelligence from artificial analysis. But this essentially says how much Are you paying to get these tasks done? Right. Because all of these benchmarks, right, artificial analysis runs it and they say, here's how much it actually costs to get all of this work done. And what we've seen is, well, Anthropic models— this is not one where you want to have a big lead. You want to have a bigger bar chart. No, that's bad. Right. Anthropic models are ridiculously expensive. So as an example, it costs $2.75 per task. Whereas OpenAI's models are much cheaper, up to a third cheaper. So their smaller version, GPT-5.6 Terra, is $0.82. And what's interesting here is actually the Chinese models, Kimi K3 and Quen 3.7, actually cost more than GPT-5.6 Terra. And the Quen costs more than the mid-tier. Oh, sorry, I don't have this one on the benchmark here. But it, right, when you have a leading state-of-the-art model and it's costing about the same or even less than the open-source model. So it's like, okay, maybe there's a couple hundred companies in the US that can actually go out and run this themselves without paying API prices. Because if you are paying API prices at this point, you would probably just should be using OpenAI. Because when it comes to price per task, which is what is ultimately going to matter, they're winning, right? Or maybe you are looking at Groq and Meta Muse Spark, right? You remember on the show last week, I said it was a really bad week for Anthropic with all of these new open source models coming out, but then also with Groq and MetaMuse Spark 1.1 coming with some pretty good models that people weren't expecting, especially on the coding and software engineering side. But similarly, artificial analysis has this quadrant when it's cost per task and the essentially intelligence or the artificial intelligence— artificial analysis intelligence index score. So essentially you want to be in the upper left-hand corner, which means You have the smartest model at the cheapest cost, and none of the Chinese open source models are in that quadrant anymore. That obviously it resets as new models come out and the, the, the medians and the averages all change. But right now, the 3 companies in there are OpenAI, Grok, and Meta. No open source Chinese models where this is actually a quadrant that they used to dominate, which is why I think 6 months And a year ago, it made a lot of sense for business leaders to be looking at Chinese open-source models, especially when they were a little bit more tameable in terms of what you can do without having a multimillion-dollar compute setup. But it's just not the case anymore. So for me anyways, even though I know that the war is going to rage on, if I'm making decisions, I'm looking at these charts and saying, well, there's maybe not that big of a reason for us to look at these models unless you do want or need that open nature, which I know some companies obviously do. But it's no longer just the cheap API prices that reveal the true cost of using these models, right? Because you have these aggressive Chinese prices that shows their strategy is not just competing for money, right? Because there's other subsidies, cloud cross-selling competition, and also still these lower prices force enterprises to question the expensive closed model defaults. So I know that the models from the past week, right, specifically Kimi K3 and Quen 3.8, they may not still fit in the traditional open source cheap Chinese models to use paradigm, but There are still those models. They will still continue to be developed, and I think that there will be a place for them. But it comes to distillation. We can't get to you know twenty plus minutes and not talk about distillation. So distillation is essentially where, for the most part, Chinese companies, you know, take it's just like they copy the questions and they copy the answers for a lack of a better term from the big AI model providers. So it's just like they copy the answers on the tests. They take all the hard work from the American. Companies and use it to train their own models, right? And that accelerates the progress, but you still can't explain, uh, you know, how they get there, um, just through distillation. That's not enough because they have had some great advancements, uh, the Chinese companies, in architecture and just their overall execution. So AI model distillation isn't universally illegal. It's highly frowned upon, and it is, um, you Sparking some backlash politically and economically with export controls. I think we're going to continue to see those ramped up now here for the rest of 2026. But these alleged Chinese practices do violate contracts and have led to major US national security actions. So what is the US doing to stop this model distillation? Well, They're trying to cluster traffic and, you know, detect and, you know, catch large-scale offenders and, you know, block them. But it's pretty hard because it's— they're just playing whack-a-mole. A US House committee did, though, just unanimously back a bill to sanction foreign actors for extracting US models. But laws in the US are a slow-moving machine, right? So it could be many months or multiple quarters or even more than a year before before something like that ever becomes law. But these published open weights cannot be recalled, so businesses can still adapt them anyways, right? So when models go open source or open weight, it's kind of like the cat's out of the bag at that point. So the US can try to restrict access, but that could potentially backfire because Washington can— yeah, they can pressure, uh, you know, via export controls or otherwise, the chips, the cloud access procurement sanctions, and all these other things. But the restrictions may just protect the technology while Raising the cost for American companies. Like I said, if you are restricting these models, but companies already have the weights, you may just be keeping money ultimately from other ecosystems that could power the American enterprise, right? So I think the writing though is kind of on the wall. And I think it's actually Microsoft's— some of their recent actions that show, I think, what we might be looking at when it comes to how should large enterprises be looking at or using these Chinese models. So reportedly Microsoft was evaluating DeepSeek for cheaper Copilot co-work model routing. So not for all of Microsoft Copilot, right? And this is just according to reports. I don't believe Microsoft has confirmed anything yet. And they were looking at this for a backup because what they found, Microsoft, that the Copilot co-work was actually very cost intensive and they did started to bill it, you know, at a token rate and no longer just including usage like they did when it was in beta. So that is telling you everything that you need to know. They haven't switched to it, but Microsoft, yes, that Microsoft, the one that has huge stakes, um, in OpenAI and Anthropic, they are looking at a model like DeepSeek. For all the, you know, the time that DeepSeek's name has gotten dragged through the mud for, you know, some alleged shady practices, they're still looking at them to use them. And I think ultimately You have to look at your balance sheet. You have to look at the dollars and the cents and make it make sense. And in this case, Microsoft, right? And the other thing, again, if you are a large company and you can, you know, essentially download the weights to these models and you can fine-tune them and make sure that they perform up to a certain standard. But I think the future workloads, like in this Microsoft example, could just be routed by cost, capability, security, geography, and regulation, right? I've been a huge advocate over the years for, who knows, maybe we'll get there eventually. But, you know, the mixture of models, much different than a mixture of experts, right? That's using a single model. And then the model kind of calls on the different experts and the different parameters in this dense model, right? Mixture of models is similar to model routing, but it's just using, right? I think Perplexity's Model Council, Microsoft has something similar, but it's, well, when you put a prompt out there and there's just a router and it might send a simple prompt to one open-source model, right? That's a good example. Or it might send a complex prompt to 100 different models and then an orchestrator model to go in and collect all the information. And maybe half of those 100 models are open source. But I think if anything, the shift though does weaken loyalty to one model and rewards flexible architecture. I do think that is probably the future where enterprise leaders need to be focusing on. So as we wrap up, let's talk about that. What is coming next and what business leaders should be doing now? Well, I would do this. I'm going to say expect controlled openness. All right. Whether it is China placing export on their open models, which I know kind of goes against the very reason they put them out there in the first place, but that's another Another topic for another episode, but I would expect controlled openness. So either China is going to restrict probably the US from using their maybe most frontier models, and/or the US may also restrict the usage of some of these Chinese open models as well, especially as now these models are getting more and more capable. More and more autonomous. And when it comes to cyber, things are getting a little bit scary-ish, right? I think in 6 months, a year, that's when things are going to be getting like real scary in terms of these autonomous models, as they can technically get smaller, faster, more capable, and, well, more open source. That's when these exports are going to— or these export controls are going to come in. So expect both nations to probably just share older models, but maybe guard their best. But what you need to do is inventory where your current Chinese open source models are running and then route those tasks by cost, security, and policy risk. And then you need to document your sourcing and be ready to swap out a model or to, you know, go to your plan B. when and if those government restrictions harden. All right. I hope this episode was helpful. But like I said, whether you know it or not, there is a good chance if you're using open source models, that part of your plan might need to be modified pretty soon because the war between the US and China when it comes to AI is heating up. So I would expect a lot of back and forth racket. So don't get stuck in the middle. Don't waste hours every single day toiling over it. Just tune in to us here on the show, uh, and the newsletter at youreverydayai.com because we're always gonna keep you up to date. I hope this one was helpful. Thank you for tuning in. Hope to see you back tomorrow and every day for more Everyday AI. Thanks, y'all.