Microsoft CEO Satya Nadella surprised by DeepSeek R1’s performance against ChatGPT

Microsoft CEO Satya Nadella admitted that DeepSeek's R1 AI model caught him off guard with its performance comparable to OpenAI's ChatGPT at a fraction of the cost. 

Satya Nadella

Image Credit: Getty

Microsoft CEO Satya Nadella has admitted that Chinese AI startup DeepSeek caught him off guard with its powerful model, R1 — the first he’s seen to seriously challenge OpenAI’s ChatGPT.  Speaking to Bloomberg Businessweek, Nadella disclosed that Microsoft was compelled to respond quickly when R1 was introduced earlier this year due to its performance at a fraction of OpenAI's price.

"OpenAI has been so far ahead that no one’s really come close," he stated. "DeepSeek, and R1 in particular, was the first model I’ve seen post some points." Despite its significant investments in OpenAI, Microsoft decided to include R1 into its Azure cloud services due to the unexpected competition.

DeepSeek's R1 made headlines earlier this year in January when its open-source model produced results virtually identical to ChatGPT's — but for only $36 per job, against OpenAI's $1,000. The price-to-performance disparity was so significant that Nadella immediately requested a 48-hour security study. Microsoft's AI team, working around the clock, examined R1's code and even contacted DeepSeek's experts for assistance. 

"People didn't sleep," said Asha Sharma, Microsoft's AI platform leader. However, rather than viewing R1 as a danger, Nadella saw an opportunity. Microsoft quickly began selling R1 on Azure, allowing users to pick between OpenAI's premium models and DeepSeek's more affordable option.

The rise of DeepSeek R1 has added tension to Microsoft’s complicated relationship with OpenAI. While Microsoft has invested billions in OpenAI, Nadella is now hedging his bets by supporting numerous AI models, including DeepSeek and Microsoft's own in-house efforts such as MAI-2. 

OpenAI CEO Sam Altman has allegedly acknowledged R1's significance, describing it as "an impressive model" that prompted his team to accelerate releases. Sundar Pichai of Google applauded DeepSeek's efforts during an earnings conference earlier this year.

Nadella’s quick adoption of R1 tells us something important about where AI is heading. It’s not just about building the most advanced model anymore — it’s about making AI work smarter and cheaper.