Image credit: Grok/X

American AI companies are grappling with the rise of DeepSeek, a Chinese AI startup whose disruptive models have drawn praise and skepticism alike.

Developers at leading U.S. firms are lauding the ingenuity of DeepSeek’s technology while also scrutinizing its claims, determined to defend their multi-billion-dollar investments in artificial intelligence against this low-cost alternative.

DeepSeek has shaken the tech world with the release of its AI assistant, which on Monday climbed to the top of Apple’s App Store in the U.S., surpassing OpenAI’s ChatGPT.

The startup’s model, trained on Nvidia’s H800 chips, reportedly cost less than $6 million to develop—a stark contrast to the billions spent by American tech giants. The news triggered a selloff in U.S. tech stocks as concerns about competition rippled through the market.

READ MORE: Nvidia plummets 14%

Deepseek tops Apple Store downloads

Meet Liang Wenfeng, the visionary behind DeepSeek

While some U.S. AI experts have praised DeepSeek’s strong research team and innovative approach, others have questioned the true cost of its development.

OpenAI CEO Sam Altman took to X to call DeepSeek’s R1 model “impressive,” particularly for its cost-efficiency. Nvidia also chimed in, saying that DeepSeek’s success highlighted the growing demand for its chips.

Major players are taking notice.

Snowflake, a U.S.-based software company, decided to add DeepSeek models to its AI marketplace after a surge in customer interest.

“We decided that as long as we are transparent with customers, we see no issue supporting it,” said Christian Kleinerman, Snowflake’s executive vice president of product. Internally, employees have called DeepSeek’s models “amazing,” underscoring the growing recognition of the startup’s achievements.

Meanwhile, U.S. developers are rushing to analyze DeepSeek’s V3 model, which was detailed in a December research paper. While the paper offered some insights into the model's architecture and performance, it left many questions unanswered, including the total cost of its development.

Industry experts believe that DeepSeek’s reported $6 million expense only covers the final training run, which utilized 2,048 Nvidia H800 chips. Earlier stages of development likely required a much larger investment, with some estimates suggesting costs exceeding $1 billion.

Despite skepticism over its budget claims, DeepSeek’s progress is undeniable. Industry insiders say that China has narrowed the AI development gap with the U.S. from 18 months to just six months.

However, concerns remain about whether DeepSeek can sustain its momentum. With its open-source strategy driving widespread adoption, the company could face challenges in securing enough chips to meet demand.

DeepSeek’s decision to open-source its models, including the R1, has drawn widespread acclaim. Venture capitalist Marc Andreessen called R1 “one of the most amazing and impressive breakthroughs I’ve ever seen” and praised its open-source availability as “a profound gift to the world.”

This move underscores the growing viability of open-source AI as an alternative to the proprietary, resource-intensive models developed by U.S. tech giants.

The rise of DeepSeek challenges longstanding assumptions that only the wealthiest companies with vast computing resources can dominate AI development.

With major U.S. tech firms set to report earnings in the coming weeks, the startup’s breakthrough has added pressure to prove that traditional models of AI innovation can still hold their ground in an increasingly competitive global landscape.

‍