There is no shortage of capital, companies or innovation. What is missing is continuity: when a project takes a decade and the rules change every four years, uncertainty becomes the biggest risk of all.
Published on July 18, 2026 · by Alex · Artificial intelligence · Data centers · Geopolitics
While China accelerates the construction of data centers, factories and power grids, the United States has just made a decision that could change the pace of the technology race. And the striking part is that almost nobody is talking about the consequences.
Because the underlying problem is not a lack of money, companies or innovation. The problem is time: AI infrastructure takes a decade to build, and the rules of the game in the United States can change every four years.
| China | United States |
|---|---|
| Coordinates banks, utilities, local governments and industry in one plan | Each actor decides separately, with different incentives |
| The plan holds for years without changing | Every election can shift priorities |
| Execution speed that is hard to match | More checks, more uncertainty |
| Mistakes, overcapacity and more accumulated debt | Less overbuilding, but also less speed |
Worth saying plainly: the Chinese model is not free. It makes mistakes, creates overcapacity and piles up debt. The comparison is not "one good and one bad" — it is which kind of risk each side takes on.
China's risk is overbuilding. America's risk is not finishing what it starts. For a company that has to commit billions over ten years, the second weighs more: it can lose government support midway, and nothing hedges that.
New York became the first state in the country to approve a moratorium of up to one year on new large-scale data centers. Governor Kathy Hochul put it in place by executive order on July 14, 2026: during that period the Department of Environmental Conservation will not issue discretionary permits for such facilities while an environmental impact statement is prepared.
Three stated reasons: the effect on residents' utility bills, the consumption of water and other natural resources, and the stability of the power grid. A data center at this scale consumes enormous amounts of energy and water, and that bill is not paid only by whoever builds it.
The reaction was immediate. Investors such as Bill Ackman pointed out that China is not slowing down data center construction, and that the United States has to win the race for artificial superintelligence if it wants to keep its economic and technological lead.
Does the United States need less regulation to compete, or clearer rules that allow faster building without giving up protections for people?
It is not a rhetorical question, and the answer is not obvious. A one-year moratorium can cost ground in a race measured in fifteen years. But building infrastructure that drives up millions of people's electricity bills creates a backlash that also stalls projects — only later, and worse.
And one data point adds pressure: Chinese models are now competing with the best American ones at a much lower cost. If the technical edge narrows, what decides the race is precisely the ability to build.
Many agree the challenge is not to copy the Chinese model. It is to stop every change of government from restarting strategic projects that need decades to pay off.
Because the AI race no longer depends only on who develops the best model, but on who can build the necessary infrastructure before everyone else. And that could be the fight that decides technological leadership for the next decade.
None of this is a buy or sell recommendation. See our legal notice.
Because it coordinates banks, utilities, local governments and industry inside a single plan that holds for years. In the United States each actor decides separately and every election can shift priorities. The Chinese model is not free: it makes mistakes, creates overcapacity and piles up debt.
On July 14, 2026, Governor Kathy Hochul signed an executive order imposing a moratorium of up to one year on new large-scale data centers. During that period no discretionary permits are issued while an environmental impact study is prepared. New York was the first state in the country to do this.
Because AI infrastructure takes a decade to build and the rules in the United States can change every four years. For a company committing billions over ten years, the risk of losing government support midway through the project is one nothing hedges against.
This article is the written version of the Saturday analysis.
Watch the video on TikTok · Follow the channel
