Summary of Key Points
The expansion of computing power in the AI industry, such as building data centers and using large numbers of GPUs, has led to a surge in electricity demand. In the past, the cost of this additional electricity was shared by ordinary residents. Now, both the federal and state governments in the United States are legislating to require tech giants (such as Microsoft and Google) to bear the costs themselves. Residents are strongly opposing the construction of data centers in their communities due to rising electricity prices, prompting tech companies to build their own power plants to bypass the public utility grid. This not only affects the speed of AI infrastructure development but also presents investment opportunities (as existing data centers become scarce) and raises concerns about China's "East Data West Computing" strategy.
I. Why Has AI Suddenly Become a "Power Hungry Industry"? – From Chip Shortages to Electricity Challenges
Previously, the AI industry was concerned about a lack of chips (such as GPU production capacity); now, the biggest problem is a shortage of electricity. Data centers, which are the heart of AI computing power and contain numerous GPUs that consume a lot of energy, have become major power consumers. For example, Loudoun County in Virginia has 199 operational data centers, accounting for 40% of the state's total electricity consumption (compared to less than 5% in 2010). According to Bloomberg, electricity wholesale prices in areas with concentrated data centers have increased by 2-3 times, leading to higher residential electricity costs—in some places, prices have risen by 58% to 94% over five years.
In short, the constraint for AI has shifted from "how small can chips be made" (Moore's Law) to "where can we get enough electricity" (Ohm's Law), and the competition in computing power is now about who can provide sufficient power supply.
II. Who Bears the Cost of Rising Electricity Prices? – From Shared Costs to Giant Companies Bearing the Brunt
This is essentially an old issue of who should pay for the consequences: Tech companies that build data centers need to expand the electric grid, and in the past, the cost was spread among all residents, including you. But now, residents are not tolerating it—why should they bear the double-digit increase in their electricity bills when tech companies earn billions?
For instance, residents in Virginia will have to pay an additional $16 per month for electricity. Surveys show that only 7% of Americans support the construction of data centers in their communities, while 48% are strongly opposed. Within three months, $98 billion worth of data center projects were delayed due to community protests (such as Google's $1 billion project in Indianapolis being forced to be canceled).
Legislation is changing this: The federal Taxpayer Protection Act requires tech companies to fund the construction of their own power grids, and Virginia has introduced a tax on data center electricity consumption (0.011 dollars per kilowatt-hour). New York has also suspended approval of new data center projects for a year, preventing tech giants from exploiting public resources for free.
III. Increasing Legislative Restrictions: Federal and State Governments Are Slowing Down AI Growth
From the federal to the state level, restrictions are being implemented:
1. Federal Level: The bipartisan Taxpayer Protection Act aims to cut off the hidden subsidy chain that allows residents to subsidize tech giants (although it is still in the legislative process, the signal is clear).
2. State Level: Virginia's electricity consumption tax will take effect on July 1st (the first of its kind in the country), and New York has suspended approval of large-scale data center projects for a year (freezing 28 projects with a total demand of 9682 megawatts).
3. Electionary Factors: Midterm elections are approaching, and both parties are using electricity prices as an issue in their campaigns. Voters don't need to understand AI; they will vote against any increase in electricity costs, so no one is speaking up for data centers.
These policies directly affect the speed of AI infrastructure development: New projects are harder to approve, and expansion is slowing down.
IV. Tech Giants' Countermeasures: Building Their Own Power Plants
Tech companies are seeking to become self-sufficient by investing in their own power sources:
Microsoft, Google, and Amazon are all investing in their own power generation—reactivating nuclear power plants, building natural gas facilities, and exploring geothermal energy, as well as small modular reactors (SMRs). Future data centers may become integrated complexes that combine power generation with computing facilities, reducing their reliance on the public utility grid.
However, there is a time lag: The demand for AI computing power will increase significantly in 1-2 years, while it takes at least 2-3 years to build new power plants (nuclear plants even longer). In the short term, the electricity gap will not be filled, meaning the pace of AI growth could be constrained by energy supply.
V. Investment Opportunities and Implications for China
1. Investment Opportunities:
- Existing data centers that are already built and connected to the grid will become scarce, leading to increased prices for existing assets (the older they are, the more valuable they become).
- Tech companies that have invested in their own power generation (such as those with nuclear or natural gas plants) will have an advantage and won't be restricted by the public utility grid.
- Companies with large data centers in Virginia and New York will face higher costs due to taxes and additional fees for using the grid.
2. Implications for China:
China's "East Data West Computing" strategy, which involves building computing power centers in the western regions, may encounter similar electricity challenges and community opposition in 2-3 years. China needs to plan its power supply in advance to avoid conflicts with local residents over the cost of AI infrastructure expansion.
In Conclusion
AI has not changed the fundamental laws of electricity consumption, but it has accelerated the exposure of the conflict between rapidly growing industries and public infrastructure. While regulatory adjustments are slow, when they happen, they can have a significant impact. The rising electricity costs of ordinary citizens are becoming a key factor that could shape the future of AI.