Wall Street went into panic mode about two weeks before Chinese startup Deepseek released an artificial intelligence system that appears to be fundamentally more efficient than its American competitors.
Investors who have put trillions of dollars into tech stocks over the past few years worried if the tens of millions of dollars tech companies were spending on new data centers suddenly looked like comics.
However, the biggest tech companies have revealed in their recent revenue report that they believe there may not be anything too much when it comes to new data centers.
Amazon hinted on Thursday that its capital expenditures (numbers that include other items such as data center construction and warehouses) could exceed $100 billion this year. Microsoft said spending could exceed $80 billion. Alphabet said it would spend $75 billion, reaffirming its plan to reach capital expenditures of as much as $65 billion.
Together, they were able to spend about $100 billion more on these projects than last year.
Executives urged patience. The problem now, they said, is that customers want more AI than companies can supply. And the only way they can meet the demand is to build as quickly as possible.
Meta's CEO Mark Zuckerberg told employees at a company-wide meeting last week, according to a recording obtained by the New York Times. . “The competition is good,” he added. “But we need to make sure we win.”
Here are some important points to understand this spending moment:
High-tech companies need more data centers.
Many companies say they are limited by the chips, land and electricity supply needed to build data centers and are racing to make them more open. Microsoft, Alphabet and Amazon all said that capacity could lead to high cloud computing sales. Cloud services are the typical way that AI is delivered to customers.
Alphabet saw “demand beyond the availability of capacity,” Alphabet's financial chief Anat Ashkenazi told investors. “So we'll deal with it and work hard to bring more capacity online.”
Microsoft has said it has been restricted for a while, and previously told investors that pressure would be eased earlier this year. However, when reporting its latest revenue last week, executives told investors it might take until summer to get enough capacity to meet full demand. The stock fell approximately 5% in after-hours trading after reporting.
They say higher efficiency will increase AI use and demand
Many people consider it a very expensive and power-hungry place where advanced AI systems are being developed, but it is also a place where AI is being deployed. These are two different steps. Rather than training the models that support ChatGpt, you need to ask ChatGpt for recipe suggestions.
The deployment of AI is known in the industry as “inference.” That's what tech companies are saying more and more, their businesses are booming.
As costs go down, “AI will be much more ubiquitous,” Microsoft CEO Satya Nadella told investors last week.
Amazon CEO Andy Jassy told investors Thursday. The world infused with AI, all apps said it might be difficult to speculate, “This is the world we're always thinking about.” That vision is speculating at the heart of it, he said.
He argued that lowering the cost of inference follows a previous pattern of technical trends. When the system is deployed cheaply, customers said they would be “thrilled to be able to build something else they always thought cost was blaming. Previously, and they usually have more in total. You will have to spend the rest.”
Companies say they have to think for the long term.
Cloud providers are used to provide customers with the illusion of endless supply. This means you'll just have to juggle with enough data centers online to stream the video you need or answer chatbot queries. But they can't even build it in advance, locking billions of dollars that can be deployed elsewhere. Balancing these two – especially when it comes to ensuring the land, chips and power in your data center, is one of the major challenges facing businesses.
Executives argue that they can adapt how investments are used between building and deploying AI models and serving their own core businesses and their customers' businesses. Nadella said Microsoft's infrastructure is “pretty troublesome.” Ashkenazi said Google is also flexible. For example, “reuse ability” is possible to provide services to Google searches rather than cloud customers.
Zuckerberg said that while Meta was studying Deepseek and how it produced its efficiency, he said that heavily invested in data centers would be a strategic advantage over small, agile competitors.
“We serve over a billion people, and because it's a lot of people, a lot of our fleet is directed towards reasoning about running,” he told employees.
Regardless of the explanation, reducing profits – even the flashy profits of a large tech company rarely thrills investors. All companies saw their share prices fall after their earnings report.
Nico Grant and Mike Isaac contributed to the report from San Francisco.
