Due to the increasing demand for computing power and advanced technology, foreign technology giants will continue to frantically build data centers in 2024.

From January to August 2024, Microsoft, Meta, Google, and Amazon invested a total of $125 billion in AI data centers, including AI capital expenditures, total data center operating costs, cash operating expenses, software, depreciation, and electricity costs.

Microsoft is spending the most on fixed AI assets, such as purchasing Gpus and other chips and maintaining AI data centers.

The four largest technology companies in the United States spent about $218 billion in the year

Amazon's AI capex is about $16 billion, with $8 billion spent on Gpus and other data center chips and $8 billion spent on other AI spending. The total data center operating cost is about $3 billion, of which $2 billion is for training, research and development, and $1 billion is for inference.

Meta's total AI capex is $23 billion, of which it will spend $11 billion on Gpus or other data center chips and $12 billion on "other AI expenses." Total data center operating costs are $4 billion, $2 billion for training and research and development, and $2 billion for inference.

The report puts Google's total AI capital spending at $29 billion. Google is spending $14 billion on Gpus and other data center chips, and another $15 billion on other AI spending. Total data center operating costs are $4 billion, $3 billion for training, research and development, and $1 billion for inference.

Microsoft, with a total AI capex of $40 billion, is a major investor in OpenAI. During the statistical period, Microsoft spent $20 billion on Gpus and other data center chips, and other AI spending was about $20 billion. Microsoft's total data center operating costs are about $6 billion, $3 billion for training and research and development, and $3 billion for inference.

Please note that the above statistics only cover the period from January to August 2024, if calculated by the full year, Microsoft, Meta, Alphabet, Amazon's investment is about 218 billion US dollars, to know that the investment in 2022 is only about 150 billion US dollars, 2023 less than 150 billion.

Mainly driven by AI large model training, the next step will start mergers and acquisitions

According to Dell Oro Group estimates, in the third quarter of this year, Amazon, Google, Meta, and Microsoft's investment in infrastructure increased by 81% year-on-year, mainly because of AI large model training, and enterprises are desperately building data centers.

In the third quarter of spending, AWS, Azure, and Google Cloud accounted for 80% of spending. In the three months ended Oct. 27, Nvidia's revenue roughly doubled, and revenue from CPU, memory, and storage vendors increased 90 percent. In the second quarter of this year, data center components industry revenue grew 127% to reach $54 billion. Hyperscaler's continued purchase of Gpus, coupled with increased demand for CPU servers, has led to the highest revenue growth in the industry's history.

In order to ensure the operation of AI, enterprises need to invest heavily in the construction of infrastructure, procurement of equipment, but also need to consume a lot of electricity, all of which bring huge demand for real estate, building materials, semiconductors, energy and other industries. The increase in energy demand is significant because a data center consumes as much electricity as a small city.

"If you look at the history of the industry, there have always been some historic upheavals, so you keep seeing the convergence of technology, semiconductors, data centers, Hyperscalers and power producers," said Stephan Feldgoise, co-head of M&A at Goldman Sachs.

Some companies are eager to grow quickly in nascent markets, so they are active in mergers and acquisitions. Overall, the current growth in the AI arms race is largely driven by capital spending.

Goldman Sachs believes that the big investment in data centers is only the first stage, and the second stage is mergers and acquisitions, which will kick in when winners start to emerge.

The numbers show that tech companies are investing in data centers that are jaw-dropping, but many still think they're not spending enough. Google CEO Sundar Pichai told analysts in August: "If you've been in a situation like this, the risk of underinvesting is much greater for us than the risk of overinvesting."

Tech ceos generally believe that investing in data centers is a bet on the future, and that even if the tech bubble bursts, data centers can prepare for the next tech breakthrough. (Knife)