The Price of Intelligence

Three pricing signals in one day.

Alibaba raised AI compute prices by up to 34%. Micron reported margins that would make a luxury brand jealous. Nvidia restarted H200 production for China, because even in a market shaped by export controls, the demand for AI chips is too large to leave revenue on the table.

Everyone talks about how AI is getting cheaper. GPT-5.4 nano at $0.20 per million tokens. Open-source models that run on laptops. Inference costs dropping by an order of magnitude every eighteen months. The narrative is clear: intelligence is commoditizing, and access is democratizing.

The narrative is wrong. Or rather, it’s half right in a way that makes it completely misleading.

The Two-Price Economy

There are two AI economies now, and they have opposite pricing dynamics.

The intelligence economy is deflationary. Models get better and cheaper. Today’s frontier is tomorrow’s commodity. OpenAI releases nano pricing that undercuts Gemini Flash-Lite. Anthropic competes on Haiku. Meta gives everything away for free. In this economy, prices fall and the value accrues to the people who use intelligence, not the people who sell it.

The infrastructure economy is inflationary. Memory costs are rising. Cloud compute is getting more expensive. Alibaba isn’t cutting prices to win AI workloads — it’s raising them because demand exceeds supply and customers have nowhere else to go. HBM4 is sold out for all of 2026. Data center construction costs are measured in tens of billions. Power contracts are becoming the binding constraint on where you can even build.

The mistake is looking at only one of these economies and extrapolating.

If you only look at model pricing, AI is getting magically cheap. If you only look at infrastructure pricing, AI is getting alarmingly expensive. The truth is both. At the same time. And the tension between them is the defining economic feature of this moment.

What Alibaba’s Price Hike Really Means

Alibaba’s T-Head AI chips are going up 5-34%. Cloud Parallel File Storage is going up 30%. Bloomberg reported it and the stock went up — not despite the price increases, but because of them.

Think about what that signals. In a normal market, raising prices by 34% would be a competitive death sentence. Customers would switch. In this market, Alibaba’s customers can’t switch because there isn’t enough supply elsewhere. The pricing power itself is the news.

This is the same dynamic Micron is experiencing. Memory has become a seller’s market for the first time in the industry’s history, not because of cartel behavior or artificial scarcity, but because AI’s appetite for silicon is growing faster than the world can build fabs.

And it’s the same dynamic driving Nvidia’s return to China. The H200 is restarting production because China’s AI demand is so large that even under export controls, the business case is overwhelming. Jensen Huang said orders are already coming in. The Wall Street Journal reported that U.S. policy now allows H200 sales under special conditions.

Export controls were supposed to freeze China’s AI ambitions. Instead, they created a licensing regime where the chips still flow but under tighter terms. Competition tightened, but it didn’t stop. The demand is too structural.

The Lemonade Stand Fallacy

The common pitch for AI goes like this: intelligence is cheap, so everyone can build a lemonade stand. The barrier to entry is falling. A solo developer with an API key can build what used to require a team of fifty.

This is true at the application layer. And it’s irrelevant at the infrastructure layer.

Because while the lemonade is getting cheaper, the lemons are getting more expensive. The compute. The memory. The power. The networking. The data pipelines. Every physical input that makes cheap intelligence possible is going up in price, not down.

The net effect is not “AI gets cheap.” The net effect is “the value chain reshapes.” Application-layer margins compress because anyone can build the same product with the same cheap API. Infrastructure-layer margins expand because the physical inputs are scarce and getting scarcer.

Micron at 74.4% margins. Alibaba raising prices by a third. Nvidia finding ways to sell into every market on Earth regardless of geopolitics.

The price of intelligence is falling. The price of the infrastructure that makes intelligence possible is rising.

That’s not a contradiction. That’s the economy.