Amazon, Microsoft Keep Data Center Spending Spree Going
Edward LudlowAugust 6, 2024 at 7:05 AM EDT
Major technology companies are spending billions of dollars in capital expenditure and most of it is going for the same thing. But first…
Three things you need to know today:
• Palantir reports strong results, saying AI has “transformed” the business
• Dell is cutting sales jobs in a shakeup that creates a new AI-focused unit
• AI chip startup Groq scores a $2.8 billion valuation in its funding round
The party is far from over
While fears of a US economic slowdown pummeled global stock markets on Monday, hitting tech companies particularly hard, the decline won’t affect Big Tech’s plans to spend money — especially on more and better data centers.
Microsoft Corp., Meta Platforms Inc. and Amazon.com Inc. told us in their earnings reports last week that they’re raising spending, largely for data facilities in support of AI development, adding to a pipeline that was already in full flow for new capacity.
Microsoft said capital expenditures hit $19 billion in the quarter and will increase in the next fiscal year. Meta tweaked its capex to $37 billion to $40 billion for the full year, raising the low end of an earlier range by $2 billion. And Amazon, the market leader in cloud computing, said it spent $30.5 billion in the first half of the year and pledged to exceed that figure over the next six months. Much of that money is going to data centers, the unglamorous warehouse-like buildings filled with rows and rows of powerful computer servers, routers, cabling and cooling systems.
These mind-boggling figures are the cost of doing business at the cutting edge. The technology powering these AI server farms is expensive to build and operate. Even before the recent earnings reports, analysts, real estate investors and chip companies were calling the boom. Tech industry researcher 650 Group calculated that the top 5 US hyperscalers — the companies most aggressively spending on servers, including the aforementioned trio — spent$105 billion in 2023 and forecast the figure rising to $187 billion in 2028.
There are other places to look for clues on growth. Digital Realty, a real estate investment trust focused on financing data centers, shared slides in its earnings showing global capacity will jump to 83,600 megawatts by 2028 from about 34,000MW last year. That measure of electricity usage is commonly used as a rough representation of how much processing power a data center can accommodate.
It’s important to point out that names like Digital Realty and Equinix have had really big data-center development pipelines for years. The growth isn’t a new phenomenon. It’s just been supercharged by AI, like so many other aspects of the tech business.
Away from AI, there are other factors at work. The majority of the world’s companies haven’t migrated their computing workloads to the cloud. There’s a large addressable market of new customers for providers such as Amazon, Microsoft and Alphabet Inc.’s Google to host other people’s data and software. As more firms migrate their business information to the cloud, there will need to be more data centers. Artificial intelligence is the catalyst driving companies to not just build more, but also modernize the facilities that already exist.
If you scan the transcripts of the earnings calls, the word “infrastructure” appears more frequently than “data center.” What executives, and the analysts posing questions to them, mean is indeed data centers. But that comes with a much longer shopping list: land, hardware, semiconductors, cooling systems and labor.
Data centers can range from, say, 100,000 square feet to millions of square feet with tens of thousands of servers all lined up. The energy consumption can range from 20MW to more than 100MW per facility. That’s a lot of juice. And there’s not room enough in this newsletter to delve into the growing concern around where we’ll obtain the energy to power all these server farms cropping up globally.
The absolute market leader and main beneficiary of this data center boom is, of course, Nvidia Corp. Chief Executive Officer Jensen Huang likes to talk about how he sees the modern computer not as a desktop in your living room or a laptop in a cafe but a 300-pound box nestled neatly in a row of humming servers.
“The world’s trillion dollars of data centers should be modernized with accelerated computing,” Huang said at the company’s investor day in late June.
And to enforce the idea this is happening (out with the old and in with the new), you can look at those that are missing out. Intel Corp. reported horrible results, largely because its variety of general-purpose central processing units are being left behind by graphics processors better suited for churning through vast data troves to train AI models.
There is concern that the industry is already overbuilt and if things slow down we’ll be left with excess capacity. But the companies at the heart of the demand, chipmakers like Nvidia and Advanced Micro Devices Inc., are confident that AI as a driver of data center spending has legs to run.
“We are seeing an environment where everyone needs more data center AI infrastructure,” AMD CEO Lisa Su told me on Bloomberg Technology last week. Let’s see how long that environment lasts.
The big story
Google lost a major antitrust case when a US judge ruled the company was an illegal monopoly in the market for internet search engines. The Justice Department had argued that Google’s payments to Apple and Samsung, estimated to total more than $25 billion annually, for placing its search engine as the default option on smartphones and web browsers effectively blocked competitors.
Get fully charged
Nvidia is having trouble fabricating its next-generation AI chip.
Bankwatch 