Hemanth Mandapati, CEO of German startup Novo AI, was an early adopter of DeepSeek chatbots, switching to the Chinese AI model from OpenAI’s ChatGPT two weeks ago.
“If you have built your application using OpenAI, you can easily migrate to the other ones … it took us minutes to switch,” he said in an interview on the sidelines of the GoWest conference for venture capitalists in Gothenburg, Sweden.
DeepSeek’s emergence is reshaping the AI landscape, offering companies access to the technology at a fraction of the cost, according to interviews with more than a dozen startup executives and investors. It also has the potential to push other AI firms to improve their models and lower prices.
“There was an offer from DeepSeek which was five times lower than their actual prices,” Mandapati said. “I am saving a lot of money, and users don’t see any kind of difference.”
Europe’s tech startups have struggled to adopt new AI technology at the same pace as U.S. rivals, which have easier access to funding. But executives say DeepSeek could be a game changer.
“It marks a significant step forward in democratizing AI and leveling the playing field with Big Tech,” said Seena Rejal, chief commercial officer of British firm NetMind.AI, another early adopter of DeepSeek.
Analysts at Bernstein estimate that DeepSeek’s pricing is 20 to 40 times cheaper than equivalent models from OpenAI.
OpenAI charges $2.50 for 1 million input tokens – units of data processed by the AI model – while DeepSeek currently charges just $0.014 for the same number.
Venture capitalists invested nearly $100 billion in U.S. AI companies in 2024, compared with about $15.8 billion in Europe, according to data from PitchBook.
On January 22, U.S. President Donald Trump unveiled a $500 billion AI initiative called Stargate, a joint venture backed by OpenAI, SoftBank and Oracle.
Investment in Europe has been more modest.
Only France’s Mistral appears among the top foundational model developers, a list dominated by OpenAI, Meta, Anthropic and Google.
China’s DeepSeek gained attention after revealing in a research paper last month that training its DeepSeek-V3 model required less than $6 million worth of computing power from Nvidia H800 chips. It has since overtaken ChatGPT to become the top-rated productivity app on Apple’s App Store.
“This is a wake-up call that bigger isn’t always better,” said Fabrizio Del Maffeo, CEO of Axelera AI. “By making models more attainable to everyone, the total cost of ownership and barriers to building innovative tech are lowered, which can be a catalyst for the whole industry.”
While some analysts question whether DeepSeek’s training costs are as low as claimed, they agree they are significantly lower than those of comparable American models.
“I see DeepSeek as a tremendous opportunity for companies like ours,” said Ulrik R-T, CEO of Denmark’s Empatik AI. “It showed that we do not need huge budgets to achieve our vision.”
The price war may have already begun.
Last week, Microsoft made OpenAI’s o1 reasoning model available to all Copilot users for free, instead of the usual $20-per-month subscription fee.