DeepSeek makes a steep model price cut permanent

The price war heats up

DeepSeek confirmed it would permanently keep a large discount on one of its models, a move that pressures rivals competing on the cost of running AI workloads. By making the lower price permanent rather than promotional, the company is betting that volume and adoption matter more than premium margins.

For developers and businesses, cheaper model access lowers the cost of building and scaling AI features, which can accelerate experimentation and deployment across products.

Inference economics

As capable models proliferate, the industry is increasingly competing on price per token, the basic unit of AI usage. When several providers offer comparable quality, cost becomes a primary differentiator, and aggressive pricing can shift market share quickly.

This dynamic squeezes margins across the sector and forces providers to find efficiencies in how their models run, from hardware utilization to software optimization.

What it means for the market

Falling prices broadly benefit customers and can expand the overall market by making AI affordable for more use cases. At the same time, they raise hard questions about how providers fund the heavy cost of training new models if the revenue from running them keeps shrinking.

The outcome may favor players with the scale, efficiency or alternative revenue streams to sustain low prices, while pressuring those who depend on premium pricing to survive.