The AI data center arms race continues to escalate! Can Meta (META.US) bet an additional $40 billion to dispel the theory of excessive computing power?

The Zhitong Finance App learned that Meta Platform (META.US) has promised to invest an additional $40 billion in its huge data center campus in Louisiana, which has brought the total estimated investment of the location above $250 billion. The company is continuing to expand its artificial intelligence computing footprint.

Meta announced Monday that it will expand the computing power of this project in rural Louisiana to at least 5 gigawatts at a cost of $50 billion. Previously, the company announced a $10 billion investment in the data center and surrounding communities.

The media reported in May that Meta plans to spend an additional $200 billion on the project, which will mainly be used to buy expensive computing chips to be placed on the nearly 4,000-acre campus. According to an anonymous source familiar with the matter, this brought the total estimated total of the location to at least $250 billion.

Other than that $500 million, Meta hasn't publicly disclosed any other expenses for the project.

CEO Zuckerberg has made aggressive investments over the past two years to build infrastructure such as data centers, which he believes is critical to making artificial intelligence super-intelligent. The company currently has 33 data centers built or under active development, and Zuckerberg has promised to invest at least $600 billion in infrastructure projects in the US over the next few years. Last week, Meta pledged $100 million to build its first data center in Canada.

This data center in Richland Parish, Louisiana is Meta's largest and most ambitious project to date, and the high cost has prompted additional financing from outside investors. Blue Owl Capital Inc., which owns 80% of the project, has sought billions of dollars from Wall Street to help support construction. Meanwhile, Entergy Louisiana (Entergy Louisiana) is spending billions of dollars to build 10 new gas-fired power plants to supply electricity to the park. While the data center will use 5 gigawatts of computing power, more than 2 gigawatts of this will be used to meet the campus's broader power needs.

Zuckerberg said in an interview last week that Meta is still seeking all the computing power it can get. But the founder is also considering whether to lease part of the production capacity to outsiders. The media reported earlier this month that Meta is developing a plan for a cloud infrastructure business. Related discussions include plans to sell access to “native” computing power similar to other so-called “new cloud” businesses (such as CoreWeave Inc.).

Meta said in a statement Monday that once it is operational, the Richland Parish data center will provide 1,000 jobs, which is double its previous employment promise. The company said it has so far awarded more than $1.6 billion in contracts to businesses in Louisiana since construction began at the site in December 2024.

Can an investment of 250 billion US dollars mitigate the “excessive computing power theory”

Judging from the current market game, Meta's “astronomical” investment of 250 billion US dollars is not only a strong shot for the bulls, but also a stimulant that triggers deeper anxiety among the bears. It was unable to completely “dispel” concerns, but it successfully moved the focus of market debates from “whether there is excess computing power” to a deeper proposition.

Previously, what the market feared most was “tech giants can't buy it” or “the AI boom is just a momentary impulse.” Meta's move of adding 40 billion dollars and rushing for a total of 250 billion dollars refuted this with the hardest financial commitment:

This proves that demand from tech giants for top computing power chips, high-speed storage (such as SanDisk and Micron's enterprise-grade SSDs), and advanced packages (such as TSMC) did not peak in a phased manner, but rather just needed infrastructure for several years.

Zuckerberg's revealed “computing power leasing” plan dispelled the market's ROI questions about Meta's “buying so many chips only to sell ads.” If computing power can be leased to the whole society like oil, then as long as the fires of global AI startups are not extinguished, Meta's inventory will always be taken over. Seen from this perspective, it has greatly mitigated the valuation crisis in the hardware industry chain.

One side is the high demand, and the other side is the horrible cost. Meta's investment, while proving insufficient computing power, has opened up a new battleground of concern:

Wall Street is beginning to wonder if the giants' “red-eyed” arms race will drag technology stocks into highly indebted “traditional power companies” or “real estate developers” even if they don't have enough computing power?

Strong as Meta, faced with such a huge funding gap, had to introduce Blue Owl to Wall Street to raise billions of dollars in structured financing. This reliance on external leverage makes risk-averse analysts wary of potential financial structural pressures.

This investment successfully alleviated the physical anxiety of the short-term stalling of the hardware supply chain (OEM, chips, high-end storage), and proved to the market that supercomputing power is still the industrial lifeblood of the digital economy.

But it didn't root out Wall Street's mental strain on AI monetization (ROI). It changed the question from “why buy so many chips” to “rent 10 power plants to burn money, when will I be able to earn back the first 250 billion”? In this rivalry between analysts and tech giants, the watershed is likely to be in the upcoming August earnings season.