Why China gives AI away for free

The war is beyond headlines

Muzamil opens by reframing what the AI conflict actually is. This is not a technical arms race. “The real story is about what is happening between us and AI,” he says. The conflict between the US and China over AI is not about who builds the smartest model. It is about who controls the infrastructure, who locks in users, and who owns the dependency relationship.

Why the US is panicking

The American AI companies like Anthropic have grown explosively. Anthropic’s annual revenue jumped from 87 million dollars in January 2024 to 1 billion dollars just 12 months later. By April 2026, that figure had reached 30 billion dollars in annualized run rate. These numbers look like success, but Muzamil argues they hide a fragile business model. The company is not actually profitable at these scales. The real cost lives in data centers and electricity. These costs are enormous and growing.

The US model works like this: make models so good and integrate them so deeply into workflows that people cannot leave. Once a company or developer builds entire systems on top of your API, they cannot switch. They are locked in. But this lock-in requires constant price pressure. As energy costs rise and compute becomes more expensive, US companies will need to raise API prices. That breaks the lock.

China flips the game

China’s strategy is the opposite. Instead of protecting its position through scarcity and pricing, China makes models free. This removes price as a barrier. Anyone in the world can use Chinese AI models for nothing. The goal is not to make money on the model. The goal is to make the world depend on Chinese infrastructure. If every startup and developer in Africa, South Asia, and Latin America runs their workloads on Chinese servers, China owns the relationship.

Muzamil explains the fundamental difference: “The American model is to make the technology so amazing that people keep coming back and become dependent. China is doing the opposite. They are asking, why not just give it away?” By removing cost friction, China shifts the game from who has the smartest model to who has the most available infrastructure.

Infrastructure is the real prize

The cost of building and running a large language model is not in the weights or the inference code. It is in the physical infrastructure. Data centers cost billions of dollars. Electricity is the largest operating expense. China is willing to subsidize these costs because owning the infrastructure is more valuable than owning the API revenue. Whoever controls the compute controls everything downstream.

Muzamil notes that this is already visible. AI companies that used to charge per token are hitting ceiling problems. Users are adopting AI so widely that marginal cost pricing becomes unsustainable. So companies have started raising prices quietly. Early users noticed: models are slower, tokens cost more, and bandwidth has limits. The lock-in works only as long as prices stay low. Once they rise, users have to find alternatives.

China has already built those alternatives.

What this means for Pakistan

Pakistan sits in the middle of a choice, though most policymakers do not see it yet. The country can build self-sufficient AI infrastructure now or become dependent on either US pricing or Chinese free compute. Muzamil is clear that this is not a small decision. If Pakistan does not own its own compute layer, it becomes a consumer of someone else’s infrastructure. The wealth and the control go elsewhere.

The window is not wide. The first principles behind AI are becoming visible. The competition is not about who is smarter. It is about who owns the infrastructure that makes intelligence cheap.

A call for first principles thinking

Muzamil closes with a challenge to think from first principles, not from headlines. Most discussions about AI get caught in correlation and miss causation. Why is China doing this? Why is the US doing that? If you understand the infrastructure game, you see both strategies make perfect sense. The US, which has the most expensive energy and highest labor costs, needs to extract high value from each computation. China, with cheaper electricity and state backing, can afford to give compute away as a tool of geopolitical leverage.

This is not about kindness or innovation. It is about owning the layer that everyone depends on. Understanding that changes how you read the news.

The homework for viewers

Muzamil does not claim to have all the answers. He calls for readers to engage critically. If you see it differently, write it out. If you have data that contradicts this analysis, share it. The goal is not to convince you of a single view. The goal is to think together about what is actually happening beneath the press releases and marketing.

This is the kind of thinking that matters for countries that want to stay competitive. Pakistan has the talent to build this. The question is whether it has the will to start before the choice narrows.