The economic reality behind the AI boom: Two data centers in California have been idle for years due to "power shortages".

Generative AI brings unprecedented computing demands, but in California, USA, where Nvidia's headquarters is located, two completed data centers have remained idle for years due to “waiting for power.” As the Silicon Valley power grid struggles to cope and power equipment becomes outdated, the AI boom is exposing another invisible crisis: electricity is becoming the biggest bottleneck for the American tech industry.

The empty data center in the hometown of Nvidia: built but with no electricity available.

Santa Clara, California, is the headquarters of the global AI chip leader NVIDIA, yet it is now experiencing a strange phenomenon.

Two data centers built through the investment of the real estate investment trust Digital Realty Trust and the digital infrastructure developer Stack Infrastructure have been completed for a long time but have been idle for years due to insufficient power supply from the public utility Silicon Valley Power, SVP.

The company's spokesperson Janine de la Vega stated, “SVP is investing $450 million to upgrade the power grid system, which is expected to be completed by 2028.”

(OpenAI as a national treasure? When investment myths collide with economic realities, will AI become a national arms race?)

AI will double the use of electricity, and the power grid has become an invisible pain point in the American economy.

Bloomberg predicts that the electricity consumption for AI computing in the United States will double before 2035. However, the aging power system, slow infrastructure upgrades, and cumbersome permitting processes in various states make electricity the biggest obstacle to industry expansion.

Dominion Energy, an energy company in Virginia, has pointed out that the waiting period for data center power connections has been extended to 3 to 7 years. Amazon (Amazon)'s Oregon division has further sued Buffett's Berkshire Hathaway (Berkshire Hathaway), claiming that its utility company has refused to supply sufficient power.

In an era where AI, cloud computing, and electric vehicles are competing for energy, electricity is becoming a scarce resource, replacing chips. As former Google CEO Eric Schmidt stated in April this year:

AI may ultimately consume 99% of the world's electricity, growing faster than we can build new power plants.

Why is Silicon Valley particularly difficult? The low-latency battleground of AI.

Although regions in the central United States, such as Texas and Louisiana, have no difficulty in expanding large AI parks, data centers must be close to densely populated areas to reduce latency for applications like high-frequency trading, autonomous driving, and real-time cloud services.

Bill Dougherty, Vice President of Data Centers at the real estate brokerage firm CBRE, pointed out: “Some demands have to be close to the city, like in California, but power constraints prevent them from going online.”

In other words, the advancement and development of AI is in positive conflict with the reality of energy.

(Jensen Huang warns: China will surpass the United States in the AI race, with energy and regulation becoming key factors)

Despite substantial capital investment, developers are still constrained by power limitations.

Even if developers hold vast sums of money, they cannot escape the reality of electricity.

Digital Realty has over 300 data centers worldwide, with an average cost of up to $13.3 million per megawatt, and even higher in Silicon Valley. Stack's parent company, Blue Owl Capital, has also announced a $50 billion investment plan, including the construction of a new campus in collaboration with Meta and Oracle (.

However, Blue Owl CEO Marc Lipschultz candidly stated: “The real competition is not capital, but whether you secured the right land and sufficient power two years ago, while also knowing how to navigate regulatory hurdles.”

The Next War of AI Infrastructure: Seizing Electricity, Not Land

One can imagine that the opportunities in the AI era do not lie in servers and chips, but in a stable power grid and energy supply. Looking at the power supply bottlenecks across various states in the U.S., the tech industry is facing a crisis.

The Economic Reality Behind the AI Boom: Two Data Centers in California Remain Idle for Years Due to “Power Shortages” Originally Appeared on Blockchain News ABMedia.