Big Tech keeps its AI data data center alive

Wall Street entered into panic mode approximately two weeks after the new Deepseek company launched an artificial intelligence system that seemed to be radically more efficient than its US competitors had built.

Investors who had pumped billion dollars in technological actions in recent years worried if the tens of billions of dollars that technological companies spent on new data centers suddenly seemed exaggerated comics.

But larger technological companies made it clear in recent gain reports that believe there can be no exaggeration when it comes to new data centers.

amazon implied on Thursday that its capital expenses, a figure that includes the construction of the data center and other items such as warehouses, could exceed $ 100 billion this year. Microsoft said its expense could exceed $ 80 billion. Alphabet said he would spend $ 75 billion, and a goal reaffirmed the plans so that the capital expenditure reached up to $ 65 billion.

Combined, they could spend approximately $ 100 billion more than last year on these projects.

Executives urged patience. The problem at this time, they said, is that customers want more than companies can supply. And the only way in which they can meet demand is to build as much as they can as quickly as possible.

“Every time I see someone else to do something better, I say: 'Ugh, we should have done that,” said Mark Zuckerberg, Meta Executive Director, employees at a meeting of the entire company last week, according to a recording obtained by the New York Times. “The competition is good,” he added, “but we must make sure to win.”

Here are some key points to understand this happy moment for technology:

technology companies need more data centers than they have.

Many of the companies say that they are limited by the supply of chips, land and energy necessary to build data centers, and are running to open more of them. Microsoft, Alphabet and amazon said they could have had higher cloud computing sales if they had the capacity. Cloud services are the typical form of ai that is delivered to customers.

Alphabet saw “demand that exceeds our available capacity,” Anat Ashkenazi, Alphabet Finance Head of Finance, told investors. “So we will work hard to address that and make sure we bring more online capacity.”

Microsoft has been saying that it has been limited for a while, and previously told investors that the pressure would relieve earlier this year. But last week, when he reported his last profits, executives told investors that he could take until summer to obtain enough capacity to meet the complete demand. Their shares fell around 5 percent in trade away after the report.

They say that greater efficiency will expand the use and demand of ai

While many people think of data centers such as enormously expensive and hungry places of energy where advanced ai systems are developed, they are also where ai unfolds. Those are two different steps: to train a model that supports Chatgpt, instead of asking ChatgPT with a recipe suggestion.

The implementation of ai is known as “inference” in the industry; It is where, technological companies say more and more, their booming businesses.

As costs lower, “the ai ​​will be much more omnipresent,” said Satya Nadella, executive director of Microsoft, to investors last week.

Andy Jassy, ​​Executive Director of amazon, told investors on Thursday that, although a world where each application was infused with ai could be difficult to understand: “This is the world we are thinking about all the time. ” That vision, he said, has inference in his nucleus.

He argued that reducing inference costs would follow the pattern of previous technological trends: as the systems become less expensive to deploy, said Jassy, ​​customers “It was completed with what else they could build that they always thought it was expensive before, and they generally end up spending much more in total. “

Companies say they have to think about the long journey.

Cloud suppliers are accustomed to giving customers the illusion of an endless supply, which means that they must juggle with sufficient online data centers to transmit the video you want or respond to your Chatbot consultation. But they cannot build too much anticipation, blocking billions of dollars that could be implemented elsewhere. Balancing these two, particularly when securing land, chips and energy for data centers can take years, is one of the huge challenges faced by companies.

Executives have argued that they can adapt how investments use, between the construction and implementation of ai models, and between serving their own main businesses and those of customers. Mr. Nadella said Microsoft's infrastructure was “quite fungible.” Mrs. Ashkenazi said Google was also flexible. It could, for example, the “reuse capacity” to serve Google instead of customers in the cloud.

Zuckerberg said Meta was studying Depseek and the ways in which he created efficiencies, but that investing largely in data centers would be a strategic advantage against a small and agile competitor.

“We serve more than one billion people, that is only many people, so more and more of the fleet will be used towards inference,” he told employees.

Regardless of the explanation, it is unlikely to reduce the profits, even the striking profits of the largest technology companies, excite investors. Each company saw the price of its shares fall after its profit report.

Nico Grant and Mike Isaac San Francisco contributed reports.