AI "power shortage" ignites virtual power plants! Sunrun partners with Tesla to establish a 16GW power plant, with stock prices soaring by 31% at one point

The Zhitong Finance APP noted that American residential solar supplier Sunrun (RUN.US) announced a partnership with Tesla (TSLA.US) and Renew Home to provide electricity for data centers and utility companies across the United States. The company's stock price surged by as much as 31%, closing up 12.57% at $14.42 on Wednesday.

These companies stated on Wednesday that the agreement will establish a platform to provide 16 gigawatts (GW) of electricity by integrating hundreds of thousands of home batteries operated by Sunrun and Tesla, along with 8 million thermostats managed by Renew Home—enough to meet the electricity needs of approximately 12 million American households.

They indicated that this electricity capacity will be offered to artificial intelligence (AI) data center developers on a first-come, first-served basis. The power supply will also be provided to the largest grid operator in the United States—PJM Interconnection.

The boom in artificial intelligence has triggered a surge in electricity demand, placing heavy pressure on the grid. The collaborating companies stated that the current electricity infrastructure is designed based on peak demand that occurs only a small portion of the year, leading to expensive power sources being idle for long periods.

A recent report from the Brattle Group found that better utilization of the existing grid could reduce electricity costs in the U.S. by $110 billion to $170 billion over the next decade and accelerate the connection speed of data centers.

This year, residential solar suppliers have faced declining sales as the lucrative solar panel tax credit policy expired. In response, home energy suppliers have adjusted their strategies to find ways to profit from the soaring electricity demand from data centers.

U.S. Data Centers in Power Struggle

As the iteration of AI chips and training of large models grow exponentially, U.S. AI data centers are facing an unprecedented electricity shortage crisis. This "gold rush" triggered by AI is rapidly evolving into a brutal "power grab."

According to forecasts from the International Energy Agency (IEA) and several Wall Street institutions, electricity consumption by data centers across the U.S. is expected to double in the coming years. Taking the largest data center hub in the U.S.—Northern Virginia (within the PJM grid coverage area)—as an example, the waiting time for new data centers to be powered has extended to several years.

To break the electricity bottleneck, tech giants and energy companies are showcasing their ingenuity. On one hand, companies like Microsoft are turning their attention to nuclear power, reviving old nuclear plants through long-term power purchase agreements; on the other hand, as demonstrated by the latest collaboration between Sunrun and Tesla, utilizing distributed residential storage and virtual power plants (VPP) to share peak loads is becoming a new path to alleviate grid pressure.

The reason virtual power plants (VPP) are being revisited is that they do not require building power plants or laying lines, relying solely on software to integrate home storage, temperature control, and EV charging stations into a "reverse power plant," which can be operational in just a few months According to Wood Mackenzie, by 2025, the number of active Virtual Power Plants (VPPs) in North America will reach 1,940, a year-on-year increase of 33%, with a total capacity of 37.5 GW, and 25 leading buyers each securing over 100 MW. The driving force is no longer utility companies, but data center developers who are stuck in the grid interconnection queue.

The most typical signal is the three-year, 100 MW agreement signed between Google and VPP operator Voltus—Voltus aggregates residential storage, smart temperature control, and flexible industrial and commercial loads in the PJM region, forming a "Bring Your Own Capacity" (BYOC) model funded by Google.

In the short term, electricity has replaced computing chips as the primary ceiling determining the expansion speed of the AI industry in the United States. Whoever can secure a stable power supply first will gain an absolute advantage in the arms race of the second half of AI