SAN FRANCISCO – Days after the overnight success of Chinese chatbot DeepSeek wiped a trillion dollars off the value of U.S. tech stocks, software CEOs are telling investors that the artificial intelligence revolution is still on track.
On earnings calls Wednesday, the leaders of Meta and Microsoft said they still plan to spend tens of billions of dollars this year on new chips and data centers for AI. They said the huge investments can still pay off in a world where upstarts like DeepSeek give away cutting-edge AI software that others can build on.
Both CEOs also indicated that they expected capable AI software to be increasingly available at low cost or free – a trend that could put pressure on companies dependent on charging others to access private AI models.
“DeepSeek has had some real innovations. … Obviously, now all that gets commoditized,” Microsoft CEO Satya Nadella said on an investor call Wednesday after his company posted its latest quarterly results.
Nadella said wider access to AI models would encourage more use of the technology, increasing demand for Microsoft’s cloud business. “People can consume more, and there will be more apps written,” Nadella said. “This is all good news as far as I’m concerned.”
Nadella also said his company, OpenAI’s primary partner and largest investor, is hosting DeepSeek’s R1 AI model for cloud-computing customers to try out or build upon.
Meta CEO Mark Zuckerberg was also upbeat on an investor call after his company’s quarterly results Wednesday. The way DeepSeek released the AI model powering its blockbuster chatbot app for others to download and modify without charge validated Meta’s decision to give away its own AI technology, he said.
Zuckerberg said that the latest version of Meta’s free Llama AI model is competitive with leading proprietary rivals, but that the next release will take the lead. He also said Meta is trying to learn from DeepSeek’s technology.
“There’s a number of novel things that they did that I think we’re still digesting,” Zuckerberg said. “And there are a number of things … that we will hope to implement in our systems.”
Meta has set up “war rooms” internally to analyze DeepSeek’s technology – a measure the company often takes when competitors such as OpenAI, Anthropic or Google release new AI models, according to a person familiar with the matter who spoke on the condition of anonymity to discuss private company initiatives.
DeepSeek, founded by AI engineers in China who previously worked on algorithms for trading stocks, said technical breakthroughs allowed it to build an AI chatbot with capabilities similar to the best offered by U.S. companies for a fraction of the cost.
The company’s app went viral this past weekend, sending shock waves through Silicon Valley, Wall Street and Washington.
Tech stocks fell on Monday as investors questioned whether U.S. companies had lost their lead in the AI race and wasted some of the billions of dollars being invested in AI data centers and chips to stock them. Shares in Nvidia, the semiconductor company whose chips are vital to AI projects, dropped nearly 17 percent.
By market close Wednesday, the tech-heavy Nasdaq composite index had recovered some of Monday’s losses but was still down compared with last week.
DeepSeek’s technology earned widespread recognition from AI insiders this week, but some whose apps were displaced by the Chinese start-up in app store rankings said they could stay ahead.
“Deepseek’s R1 is an impressive model, particularly around what they’re able to deliver for the price,” OpenAI CEO Sam Altman said Monday in a post on X. “We will obviously deliver much better models and also it’s legit invigorating to have a new competitor!”
OpenAI said Wednesday that the company had evidence that DeepSeek violated its terms of service by using responses from the ChatGPT developer’s AI offerings to help train its own chatbot, undermining the idea that Chinese developers had built a highly capable AI all on their own.
The stock market often sells tech stocks when an innovation comes out that could force companies to lower prices, said Andrew Feldman, CEO of AI computer chip company Cerebras. But such sell-offs inevitably prove to be short-sighted, he said, because lower prices increase the possible uses of software, and demand shoots up. “The first thought was, this makes the market smaller. In every instance, that was wrong.”
AI analysts have also questioned DeepSeek’s claim of spending only a few million dollars to train one of its earlier AI models, saying that figure didn’t account for other costs such as buying or renting computer chips or running the software after it was finished.
With numerous AI chatbots now available, many U.S. companies are targeting more sophisticated applications of the technology.
Zuckerberg said on Meta’s investor call that he expects it will be possible next year to build an AI tool that “has coding and problem-solving abilities of around a good mid-level engineer,” adding that it could become “one of the most important innovations in history.”
Microsoft and Meta both said they will charge ahead with deep investments in AI hardware and data centers.
Meta advised investors that its capital expenditures will grow in 2025 in part because of AI spending, to between $60 billion and $65 billion.
Microsoft will continue with its previously disclosed plan to spend as much as $80 billion on AI infrastructure this year. Nadella said it was necessary to provide computer resources for other companies to use AI tech to build their own applications. “AI will be much more ubiquitous,” he said.
Meta’s stock rose 2.3 percent in aftermarket trading, while Microsoft’s fell 4.6 percent. The software giant reported lower than expected revenue in its cloud division, though Nadella said on the investor call that the company’s AI-related revenue was growing quickly.
