Satya Nadella says that Deepsek R1 came close to chat and surprised him
Microsoft CEO Satya Nadella has admitted that Chinese AI Startup Deepsek caught him from a guard with his powerful model, R1 – first he saw Openi’s chat seriously.
Listen to the story

Microsoft CEO Satya Nadella has admitted that Chinese AI Startup Deepsek caught him from a guard with his powerful model, R1 – first he saw Openi’s chat seriously. Talk to Bloomberg BusinessweakNadella revealed that when R1 launched earlier this year, its performance at an excerpt of the cost of openi forced Microsoft to take rapid action. “Openi has been ahead so far that no one really comes close,” he said. “Deepsek, and especially R1, was the first model I have posted some points.” The unexpected competition inspired Microsoft to integrate R1 in its Azure Cloud Services, yet heavy investment in Openai.
Deepsek’s R became a city talk earlier this year in January 1, when its open-source model gave almost matching results for approximately $ 36 per function compared to Open-$ 1,000. The price-to-demonstration difference was so striking that Nadella immediately ordered a 48-hour security review. Microsoft’s AI team, working around the clock, examined the Code of R1 and even reached the Dipsec engineers for north. Recalling Microsoft’s AI platform head Asha Sharma, “People didn’t sleep.” But instead of seeing R1 as a danger, Nadella saw an opportunity. Microsoft soon began offering R1 on Azure, making customers selected between Openai’s premium model and a budget -friendly option of Deepsek.
The rise of Deepsek R1 has added tension to Microsoft’s complex relations with OpenAII. While Microsoft has put billions in Openai, Nadella is now making its bets by supporting several AI models, including MII-2-like Microsek and Microsoft’s own in-house projects. Openai CEO Sam Altman has allegedly accepted the impact of R1, called it a “impressive model” that pushed his team to speed up the release. Even the beautiful Pichai of Google praised Deepsek’s work during an income call earlier this year.
Nadella’s quick adoption of R1 tells us some important where AI is going. It is no longer about making the most advanced models – it is about making AI work smarter and cheaper. “AI is rapid commoditis,” Nadella said, similarities with how coal and air travel over time became more efficient. The Azure of Microsoft now hosts 2,000 different AI models from giants such as Openai and Meta to scrapy startups. Message? Customer wants options, and Microsoft is not going to give a partner loyal to limit his stakes.
The rise of Deepsek R1 proves that in AI, no lead is safe. Even Nadella, who has chosen Arabs on Openai, had to accept: “When you have something that is good and cheap, you do not ignore it. You adapt.”