Historical note: this article covers DeepSeek’s January 2025 surge in the U.S. App Store. “Surpasses ChatGPT” refers to the free-app ranking at that moment, not a blanket conclusion that DeepSeek’s model was better than ChatGPT.
DeepSeek’s AI Assistant rose above ChatGPT in the U.S. free-app ranking on Apple’s App Store in January 2025, bringing the Chinese startup into mainstream attention.
The app used DeepSeek models including DeepSeek-V3. DeepSeek’s researchers said their training approach used Nvidia H800 chips and reported a training-compute cost below $6 million for V3. That figure was widely discussed at the time, but it should not be read as the total cost of building the company, developing every model or operating the service.
Reuters reported that DeepSeek’s rise prompted investors to question whether frontier AI would always require the level of hardware spending previously assumed.
The market reaction was immediate. Nvidia and other technology shares fell as investors reassessed AI infrastructure spending. That reaction reflected investor expectations at the time; it did not prove that demand for AI chips would permanently fall.
The useful lesson is to separate three different claims: an app-store ranking, model performance and development cost. DeepSeek’s January 2025 ranking established strong consumer interest. It did not, by itself, settle the other two questions.
