DeepSeek’s emergence is changing the AI landscape, giving businesses access to the technology at a fraction of the cost, according to Reuters interviews with more than a dozen startup executives and investors. The Chinese AI model also has the potential to spur other AI companies to improve their own models and lower costs.
"DeepSeek offered five times lower than their actual price. I saved a lot of money and users didn't see any difference," said Hemanth Mandapati, director of German startup Novo AI, who switched from OpenAI's ChatGPT to DeepSeek two weeks ago.
Tech startups in Europe struggle to adopt new technology at the same pace as their counterparts in the U.S., where funding is easier to come by. But executives say DeepSeek could be a game-changer.
“This is a significant step forward in democratizing AI and leveling the playing field with Big Tech,” said Seena Rejal, chief commercial officer at UK company NetMind.AI, one of the first companies to adopt DeepSeek.
Analysts at Bernstein estimate that DeepSeek is 20 to 40 times cheaper than comparable OpenAI models. OpenAI charges $2.50 per million input tokens—units of data processed by an AI model—while DeepSeek currently charges $0.014 for the same number of tokens.
Regulators have raised concerns about whether DeepSeek copied OpenAI’s data or censored answers that could be misleading about China, an issue that is currently being investigated in several European countries.
“While the future of DeepSeek as a business is difficult to predict, the structural implications appear to be quite far-reaching,” said Sanjot Malhi, a partner at venture capital firm Northzone.
According to data from PitchBook, nearly $100 billion was invested by venture capitalists in AI companies in the US in 2024, while in Europe the figure was around $15.8 billion. On January 22, US President Donald Trump announced the $500 billion AI project Stargate - a joint venture supported by OpenAI, SoftBank and Oracle.
AI investment in Europe remains more modest, with only France’s Mistral making the list of leading AI models, alongside names like OpenAI, Meta, Anthropic and Google.
China's DeepSeek made headlines when it announced that it cost just $6 million to train its DeepSeek-V3 model on Nvidia H800 chips. Since then, the AI tool has also surpassed ChatGPT to become the highest-rated productivity app on Apple's App Store.
“This is a wake-up call that bigger is not always better. By making models more accessible to everyone, the total cost of ownership and barriers to building cutting-edge technology will be reduced, which could be a catalyst for the entire industry,” said Fabrizio Del Maffeo, CEO of Axelera AI.
Reuters commented that the price competition may have begun.
