DeepSeek Limits Access to Its AI Models Due to High Demand

Chinese startup DeepSeek, which made waves in Silicon Valley with its AI model R1—capable of advanced reasoning with lower computational costs—has announced restrictions on access to its API.

According to Bloomberg, the move comes as the company’s servers struggle to handle surging demand. DeepSeek stated on its official website that it is temporarily halting API credit top-ups to prevent system overloads and service disruptions. Previously purchased credits remain valid, with the company reassuring users: "Existing credits can still be used. We appreciate your understanding!"

DeepSeek, which had been offering its services at exceptionally low prices—raising concerns among competitors—also announced that it will discontinue discounts starting February 8. Since late January, its services have been operating at full capacity due to the rapid rise in popularity of its AI assistant, which the company claims is a serious competitor to OpenAI’s ChatGPT. Notably, DeepSeek emphasizes that developing its model required significantly less investment than its Western counterparts.

This claim has sparked scrutiny in the U.S., where officials have launched an investigation into potential export control violations. There are suspicions that DeepSeek may have acquired high-end Nvidia accelerators through intermediaries in Singapore, circumventing the U.S. ban on selling advanced chips to China.