Earlier this week, Chinese artificial intelligence startup DeepSeek sparked a sell-off of global tech stocks that lost the AI-driven stock rally $1 trillion in value.
The Hangzhou-based company launched its open-source DeepSeek-R1 AI reasoning models the previous week, declaring them on par with those from OpenAI and Anthropic. In December, the startup released DeepSeek-V3, an AI model it said cost just $5.6 million to train and develop on just under 2,050 of Nvidia’s (NVDA+0.26%) reduced-capability H800 chips. V3, according to the startup’s technical paper, performed comparably to OpenAI’s GPT-4o and Meta’s (META+1.61%) Llama 3.1 models, which use a much larger quantity of Nvidia’s more powerful chips.
DeepSeek’s cheaper-yet-competitive models have raised questions over Big Tech’s big spending on AI infrastructure. Unsurprisingly, the Chinese startup made its way into Big Tech earnings calls this week, and the chief executives of Microsoft, Meta, and Apple took an optimistic tone while discussing what DeepSeek’s models could mean for the industry.
Here’s what Big Tech is saying about DeepSeek.
Despite launching an investigation this week over whether a group associated with DeepSeek obtained data output from OpenAI’s AI models in violation of the startup’s terms of service, Microsoft launched DeepSeek-R1 on its Azure AI Foundry (MSFT+0.85%) and GitHub.
Microsoft chief executive Satya Nadella said during the company’s earnings call on Wednesday that developers will also be able to run the startup’s distilled reasoning models on Copilot+ PCs soon.
In response to an analyst’s question about DeepSeek, Nadella said he thinks it “has had some real innovations.” As AI reasoning models such as those from DeepSeek and OpenAI get “commoditized” and “broadly used,” Nadella said customers will be “the big beneficiaries” of that software cycle.
“When token prices fall, inference computing prices fall, that means people can consume more, and there will be more apps written,” Nadella said.
Meta chief executive Mark Zuckerberg, who touts open-source AI models as “good for the world,” said during the company’s earnings call on Wednesday that DeepSeek did “a number of novel things” with its open-source models that the industry is “still digesting,” and that the startup has made “advances that we will hope to implement in our systems.”
“That’s part of the nature of how this works,” he said, “whether it’s a Chinese competitor or not.”
However, Zuckerberg said it’s too early to have a strong opinion on what DeepSeek means for the trajectory of Meta’s plans for AI infrastructure and capital expenditures.
Earlier this month, Zuckerberg said Meta is planning to invest between $60 billion and $65 billion in capex on AI this year. The company is also building a data center with a capacity of more than two gigawatts, he said — a site that could cover a large part of Manhattan. Meta plans to bring about one gigawatt of compute online this year, Zuckerberg said, and will end the year with more than 1.3 million graphics processing units, or GPUs.
While Apple (AAPL+0.01%) stock avoided the DeepSeek sell-off, Apple chief executive Tim Cook was asked by an analyst for his “worldly perspective” on the AI startup and its implications for the iPhone maker’s business.
“In general, I think innovation that drives efficiency is a good thing,” Cook said. “You know, that’s what you see in that model.”
However, Apple is facing another problem in China, with its revenue being down 11% year over year for the quarter ended in December.
The post DeepSeek rattled Big Tech. Here’s what the CEOs of Apple, Meta and Microsoft are saying appeared first on Quartz.
