CEO of Google DeepMind, Demis Hassabis, has dismissed claims that Chinese AI startup DeepSeek developed its artificial intelligence system for less than $6 million, saying it is “exaggerated and a little bit misleading.”
The statement comes after DeepSeek made headlines last month for releasing a popular chatbot and AI model at a fraction of the cost reportedly spent by U.S. rivals like DeepMind and OpenAI.
Speaking at the Artificial Intelligence Action Summit in Paris, Hassabis told Bloomberg Television that DeepSeek’s reported $5.6 million expenditure likely only covers the final training round of its AI model, which represents a small portion of the total development cost.
“They seem to have only reported the cost of the final training round, which is a fraction of the total cost,” he said.
DeepSeek’s claims under scrutiny
DeepSeek’s announcement that it trained its AI model using older Nvidia Corp. chips has raised eyebrows among researchers and industry experts. U.S. authorities have reportedly opened an investigation to determine whether the Chinese startup circumvented a chip ban by purchasing semiconductors through Singapore.
- Additionally, Bloomberg News reported that OpenAI and Microsoft Corp. are probing whether a group linked to DeepSeek used a process called “distillation” to obtain data from OpenAI’s models.
- Distillation involves one AI model extracting outputs from another for training purposes. Hassabis suggested that DeepSeek may have relied on Western models for distillation but did not provide specific details.
No “Silver Bullet” in AI development
Hassabis also refuted the idea that DeepSeek’s emergence has disrupted the economics of AI development.
“We don’t see any new silver bullet technologies. DeepSeek is not an outlier on the efficiency curve,” he said.
He emphasized that Google’s AI model, Gemini, is more efficient than DeepSeek in terms of training-to-performance and cost-to-performance ratios.
“We just don’t talk about it very much,” he added.
Continued investment in AI
Despite the buzz surrounding DeepSeek’s cost-effective approach, major Silicon Valley players like Alphabet Inc. remain committed to significant investments in AI.
- Last week, Alphabet announced plans to spend $75 billion in capital expenditures in 2025, with a focus on its cloud-computing division and AI services like Gemini, which is being integrated into Google’s search and other products.
- Earlier in January this year, Microsoft also announced plans to invest approximately $80 billion in AI-enabled data centers to train AI models across the globe.
- The Vice Chairman and President of Microsoft, Brad Smith, disclosed this in a blog post highlighting the company’s commitment to AI as part of a broader strategy to respond to China’s daring moves in the tech space.
