A wave of large technological companies has declared quarterly results in recent weeks – including six of the so -called magnificent companies – shortly after The Chinese AI startup Deepseek rocked the stock market while investors considered its new models of cheap artificial intelligence as a worrying sign for American technological actions.
On the calls for profits with investors, the leaders of these companies quickly praised the models of artificial intelligence of Deepseek, rejected them or tried to avoid the subject completely.
The reactions of the leaders of American technology for sale – who had an impact on most of their actions – went from the defensive to excited. While most have agreed that Deepseek News is a sign that IA costs will finally drop, they have reaffirmed their commitments to spend massive sums for capital expenses and other investments for IA infrastructure in 2025, despite a lack of clarity at the moment when payment for this expense comes.
Here’s what they said.
Microsoft (Msft) CEO Satya Nadella quickly adopted Deepseek, mentioning the company in its opening remarks on a post-benefit call on January 29.
Nadella underlined Microsoft’s decision to put the latest DEEPSEEK AI model on her developer platforms, Azure AI Foundry and Github, adding that he had gone through “an automated red team, a safety integration of the contents of content and safety digitization “. He said customers could soon run the Deepseek models locally on Microsoft IA PCs.
“I think Deepseek had real innovations,” said Nadella, adding that he sees the AI to become “commodity”.
“For a hyperscaller like us, a PC platform supplier like us, it’s good news for me.”
Microsoft’s president and chief executive officer Satya Nadella speaks during the Microsoft event on May 20 in Microsoft in Redmond, Washington, May 20, 2024. (Photo by Jason Redmond / AFP) (photo of Jason Redmond / AFP via Getty Images) ·Jason Redmond via Getty Images
Meta (Meta) – that a technological analyst recently described as “The most well -placed company to enjoy the generator AI»Given its advertising activity – saw its actions climb on the beginnings of Deepseek on its new AI model called R1, the actions increasing by almost 2% on the day of the news.
CEO Mark Zuckerberg was relatively nonchalant about the Deepseek frenzy.
When asked if the potential of AI models at a lower cost would affect Meta’s capital expenses, the CEO declared in a following appeal the latest quarterly results of the company: “I don’t know – it is probably too early to really have a strong opinion on what it means for the trajectory around infrastructure and capex and things like that.”
Meta PDG Mark Zuckerberg Essiaries Orion AR glasses at the annual Meta Connect event at the company’s headquarters in Menlo Park, California, United States, September 25, 2024. Reuters / Manuel Orbegozo ·Reuters / Reuters
“I also think of the light of some of the recent news, the new competitor, Deepseek de China … This is one of the things we are talking about is that there will be an open source standard Global scale, “said Zuckerberg.
“And I think that for our kind of national advantage, it is important that it is an American standard.”
Asked for his “worldly perspective” on “the deep situation”, Apple (Aapl) CEO, Tim Cook, said that in a post-benefit, called it on January 30 that “innovation that stimulates efficiency is a good thing” and noted that the iPhone manufacturer adopts a “approach very cautious and deliberate of our expenses ”.
When asked what the low cost model of Deepseek for Google means (Goog), CEO Sundar Pichai said that although the Deepseek team “did a very, very good job”, he thinks that Google’s Flash Gemini models are better.
“I would say both our 2.0 flash models, our Flash 2.0 thinking models, these are some of the most effective models, in particular by comparing the V3 and R1 of Deepseek.”
“And I think that many is our strength of complete development (Google manufactures its own personalized AI chips as well as the AI models and the software that executes them), end -to -end optimization, our cost obsession by Request, “added Pichai.
“I think that part of the reason why we are so excited by the opportunity of AI is that we know that extraordinary use cases, because the cost of use actually will continue to Descend, which will make user cases more possible. “
Amazon (Amzn) CEO Andy Jassy said he was thinking that Deepseek’s models will not start an slowdown in AI investment.
“Sometimes people hypothesize that if you are able to reduce the cost of any type of technological component, in this case, we are really talking about inference (execution of AI models), which goes in a way cause less total expenses to technology.
Jassy underlined the aggressive expenditure of the company to develop its cloud infrastructure in the early 2000s, even if the costs have dropped.
The CEO of Amazon, Andy Jassy, speaks during an opening address at AWS Re: Invent 2024, a conference organized by Amazon Web Services, at Venetian Las Vegas on December 3, 2024 in Las Vegas, Nevada . (Photo of Noah Berger / Getty Images for Amazon Web Services) ·Noah Berger via Getty Images
He is right, at least for the moment.
Meta, Alphabet, Amazon and Microsoft said that in their calls of calls according to which, despite the anticipation that training and inference costs of artificial intelligence will drop, they will be down Spend 325 billion cumulative dollars in 2025An increase of 46% compared to the previous year. Amazon is the group’s largest expenditure.
Investors were not satisfied: Amazon’s stock dropped by 4% on Friday According to the comments of the leaders according to which they expect to increase capital expenditure by 35% to more than $ 100 billion.
AMD (Dmla) The CEO Lisa SU has expressed its conviction that new innovations such as Deepseek models will not face investments in AI, stressing the infrastructure project of Stargate IA recently announced, supported by SoftBank (SoftBank (Sftby), Oracle (Orcl), and Openai.
“All these initiatives require massive quantities of new calculations and create unprecedented growth opportunities for AMD in our businesses.”
The CEO of AMD, Lisa SU, speaks on stage during the 2024 per year dinner at Current At Chelsea Piers on December 11, 2024 in New York. (Photo of Noam Galai / Getty Images for time) ·Noam Galai via Getty Images
Deepseek’s innovations were perhaps Qualcomm (Qcom) CEO Cristiano Amon, who argued that such developments could speed up the calendar so that AI knows a new wave of demand for smartphones and PCs.
“Deepseek-R1 and other similar models recently demonstrated that AI models are developing more quickly, becoming smaller, more capable and effective and now capable of operating directly on the device,” said Amon.
“While we are entering the era of AI inference, we expect that, although the training continues in the cloud, the inference will work more and more on the device, rendering the ‘IA more accessible, customizable and efficient.
The CEO and President of Qualcomm, Cristiano Amon, attended the first day of the web summit in Lisbon, Portugal, November 12, 2024. (Photo by Rita Franca / Nurphoto via Getty Images) ·Nurphoto via Getty Images
Arm (ARM) CEO René Haas also argued that the designer of flea architecture based in the United Kingdom would benefit from an AI at a lower cost for consumption devices, even by taking a blow to Nvidia.
“Also a product as wonderful as (nvidia) Grace Blackwell, you could never put it in a mobile phone, you could never put it in headphones. You can’t even put it in a car. But the arm is in all these places.
Laura Bratton is Yahoo Finance journalist. Follow it on Bluesky @ laurabratton.bsky.social. Send him an email to laura.bratton@yahoinc.com.
