The release of DeepSeek’s R1 model rattled Western tech investors — for good reason. It took just four months for the Chinese generative-artificial-intelligence lab to release a model that could match ChatGPT at its own game of “chain of reasoning” inference.

It’s not just a performance issue. So-called fast-followers such as DeepSeek seem poised to sustain an aggressive price war against AI pioneers. OpenAI came up with the chain-of-reasoning technique, allowing DeepSeek to focus on replicating it with greater efficiency. DeepSeek happens to be a Chinese company, but a comparably efficient model released elsewhere would have had the same effects.

Taken together, these facts suggest that the economics of generative AI (GenAI) has reached a tipping point. Leading GenAI labs and their investors can continue to push the frontier of model capabilities as they have done so far, at significant R&D costs — but can no longer count on the private market paying for that effort.

In other words, AI capabilities continue to require large, upfront R&D spending, but the underlying economics of the industry don’t seem to provide a stable path to recouping those investments. Why would users pay for more expensive, leading-edge technology when the market offers affordable, good-enough alternatives? The frontier of AI is where the action lies for researchers, but few users and businesses will need to continuously pay for the “latest and greatest.”