The West Should Pay Attention

For most of the artificial-intelligence boom, the story seemed almost entirely American. OpenAI, Google, Anthropic, Meta and Microsoft dominated the headlines, the investment flows and the global imagination. The assumption was that the future of AI would be built in Silicon Valley, sold through subscriptions and APIs, and controlled by a relatively small number of enormously well-funded technology companies. China is now challenging that assumption in a way that is both technically impressive and strategically important. Rather than simply copying the American model of premium AI behind expensive paywalls, Chinese companies have increasingly embraced a different philosophy: make powerful artificial intelligence cheap, widely available and, in many cases, open enough for developers to download, modify and deploy themselves. The result is one of the most significant shifts in the global technology industry. DeepSeek, Qwen, Kimi, GLM, Doubao, MiniMax and other Chinese systems are no longer obscure alternatives for enthusiasts. They are becoming major platforms used by consumers, developers and businesses, while simultaneously forcing the rest of the industry to reconsider how much artificial intelligence should actually cost.

The most important word in this new competition is not “China.” It is “free.” The economics of artificial intelligence change completely when access becomes inexpensive enough that people stop rationing their usage. When AI costs £20 or £30 a month per service, people naturally ask whether they really need another subscription. When developers are charged every time their software sends a request to a powerful model, they design applications carefully around those costs. When businesses need millions of automated interactions, even tiny per-request charges can eventually become enormous bills. Chinese AI companies have attacked precisely this weakness. Free consumer access, extremely cheap APIs and downloadable models encourage experimentation on a scale that expensive proprietary systems cannot always match. Students can use AI constantly. Programmers can test thousands of ideas. Small companies can build products without burning through venture capital simply to pay API fees. Businesses can automate tasks that previously would not have been economically worthwhile. Free AI does not merely compete for existing users; it creates completely new kinds of usage.

DeepSeek became the symbol of this transformation because it demonstrated that a Chinese AI company could suddenly become part of the global mainstream conversation. For many people outside China, DeepSeek was the first moment when Chinese artificial intelligence stopped feeling like a distant domestic technology sector and became something they could personally use. Its combination of strong reasoning capabilities, aggressive pricing and open model releases challenged the assumption that the most advanced AI would necessarily remain locked inside American corporate platforms. But DeepSeek is only one part of a much larger ecosystem. Alibaba’s Qwen family has become hugely important among developers, while Moonshot AI’s Kimi, Z.ai’s GLM models, ByteDance’s Doubao, MiniMax and models from Tencent, Baidu and other Chinese companies have created an intensely competitive domestic market. That competition matters because no single company can relax. If one laboratory releases a cheaper model, another responds. If one improves coding performance, another improves reasoning. If one releases open weights, its competitors have pressure to become more accessible themselves. The speed of that cycle is extraordinary, and users ultimately benefit from it.

Qwen may be especially important because it represents something larger than a successful chatbot. Alibaba has built Qwen into an enormous family of models covering different sizes, languages and purposes. Developers can select relatively small models for inexpensive tasks and larger systems for difficult reasoning. Some models can be run locally or on private infrastructure, meaning businesses do not necessarily have to send every confidential document or customer interaction to an external AI provider. This is where Chinese AI becomes strategically disruptive. A closed chatbot is a product. An open-weight model can become infrastructure. Once a model is downloaded and integrated into thousands of applications, cloud services, devices and corporate systems, it becomes much harder to measure its influence simply by counting chatbot users. Its real impact is embedded throughout the software ecosystem.

This is also why Western companies should be paying close attention. The competition is no longer simply about which laboratory scores highest on an AI benchmark. It is becoming a struggle over who controls the underlying layer of intelligence used by the next generation of software. A developer deciding between a costly proprietary API and a capable open Chinese model may choose the latter even if it is slightly less powerful, because cost, control and flexibility often matter more than winning the final few percentage points on a benchmark. Businesses may prefer models they can host themselves. Governments may want systems that do not permanently depend on one foreign corporation. Device manufacturers may need AI that can operate locally without sending every request to the cloud. Startups may simply want to protect their margins. In each of these cases, cheap and open AI has a fundamental advantage.

The geopolitical irony is difficult to miss. The United States has spent years trying to slow Chinese advances in advanced computing through restrictions on semiconductor exports and other technologies, yet those pressures may also have encouraged Chinese companies to become unusually aggressive about efficiency, cost reduction and open distribution. Restricted access to the very best hardware creates an incentive to squeeze more performance from available resources. Competing against American companies with enormous financial resources creates an incentive to fight on different terms. If it is difficult to outspend the largest American AI companies, one alternative is to make your technology so inexpensive and accessible that developers everywhere begin using it voluntarily. Software can spread across borders far more easily than physical hardware. Once an open model has been downloaded, copied and integrated into products around the world, technological influence becomes extraordinarily difficult to contain.

None of this means that Chinese AI should be treated uncritically. Users should understand where their data is being processed, what information is retained and which privacy rules apply. Companies should distinguish between truly open-source projects and open-weight systems whose training process remains largely opaque. Political censorship can affect the behaviour of some China-hosted services, just as commercial moderation policies and legal requirements affect Western platforms. Businesses working with confidential or regulated information should evaluate models carefully regardless of where they were developed. These are legitimate issues. But they do not erase the larger reality that Chinese AI companies are producing capable technology that millions of people and developers increasingly find useful. Dismissing those systems merely because they come from China would be as strategically foolish as assuming that every Chinese product must automatically be superior.

The bigger story is that artificial intelligence itself may be becoming a commodity. We have seen this pattern before. Computer storage was once extraordinarily expensive. Bandwidth was scarce. Digital photography was costly. Powerful software required boxed licences costing hundreds or thousands of pounds. Over time, competition and technological improvement drove the marginal cost of these capabilities toward zero. AI may be travelling down the same road much faster. If dozens of companies can produce competent language models, the raw model becomes less valuable by itself. Value moves into applications, services, hardware, proprietary data, agents, distribution and user experience. In that environment, companies trying to maintain very high prices for basic intelligence may find themselves fighting the economics of the technology itself.

This could ultimately benefit everyone, including people who never use a Chinese AI product. Competition from DeepSeek, Qwen, Kimi and others places pressure on Western providers to cut prices, improve free tiers, release more efficient models and provide users with greater flexibility. A person who remains completely loyal to an American AI platform may still benefit from Chinese competition because that competition forces their preferred company to improve. This is one reason the rise of free Chinese AI should not be understood purely as a geopolitical threat. It is also an enormous competitive force. Technology markets are healthiest when no single group of companies can dictate prices indefinitely.

The next stage of the AI revolution will make this even more important. Today, most people still think of artificial intelligence as something they deliberately open: a chatbot, a website or an application. In the future, AI will increasingly disappear into everything else. Phones, cars, cameras, office software, industrial machinery, robots, search engines, medical devices and consumer electronics will contain models running quietly in the background. At that scale, cost becomes decisive. A manufacturer cannot necessarily afford to pay a premium API charge every time a device performs an intelligent action. The logical solution is local or self-hosted AI, and that is exactly the environment in which open and inexpensive models become extremely powerful.

China therefore appears to be making a very ambitious bet. It is betting that the future of artificial intelligence will not belong exclusively to whoever owns the smartest closed model. It may belong to whoever distributes capable intelligence most widely. Give developers the models. Make consumer access inexpensive. Lower API prices. Encourage enormous ecosystems to form around the technology. Allow thousands of companies to build on top of it. Instead of protecting every model as a precious secret, use models as a way of expanding technological influence.

That strategy could prove enormously consequential. Artificial intelligence began as a race to build the most impressive machine. It is rapidly becoming a race to determine who can make intelligence abundant. The winners may not necessarily be the companies charging the highest subscription prices or producing the most spectacular demonstrations. They may be the companies whose models quietly spread into millions of applications and billions of devices.

And that is why the popularity of free Chinese AI matters so much. China is not merely creating cheaper chatbots. It is helping redefine what people believe artificial intelligence should cost. Once users become accustomed to powerful AI being available for little or nothing, the expectations of the entire market change. The question facing the rest of the industry is therefore no longer simply whether Chinese AI can compete.

It is whether expensive AI can continue competing with free AI.