AI Is Getting Cheaper—Why That Could Worry Nvidia

Cheaper AI helps customers do more, but may weaken the spending growth needed to support the industry’s enormous infrastructure buildout.

Falling model prices are changing the economics of AI. Businesses can route routine work to faster, cheaper models and reserve frontier systems for tasks that truly need deeper reasoning. That can deliver more work without a matching increase in the customer’s bill.

This is excellent news for buyers, but a harder financing picture for infrastructure providers. The enterprise-spending data discussed in the video shows modest recent growth despite substantial demand. If unit prices fall faster than usage rises, suppliers need far more activity just to produce the revenue growth expected from their capacity commitments.

Model orchestration is one sign of the shift: use a reasoning model to plan, then a less expensive model to execute. The pattern is spreading beyond software development into customer support, where many requests need a quick, accurate answer rather than frontier-level reasoning.

The key question is not simply whether AI usage is growing; it is whether each data center can sell enough compute at prices that cover its costs. Nvidia’s reported supply commitments make that distinction consequential. Export controls and the preservation of experienced engineering teams are additional factors shaping the sector’s ability to grow reliably.

Source