written 23 August, 2026
published 30 August, 2026
Most everyone has now heard about Artificial Intelligence (AI) and the massive data centers required to support it.
Boosters believe it is the wave of the future, solving problems previously unsolvable, revolutionizing the economy, with potentially trillions in profit for investors. AI helps wildfire cameras be more efficient, improves medical diagnostics, has designed new drugs, and even new viruses. The possibilities are amazing.
Detractors point to science fiction like the "Terminator", bioterrorism, disinformation, and the potential destruction of the middle class from massive unemployment. Political deep fakes are getting much better, being created more quickly, destroying trust. Models have "escaped" their test bed. AI lies, and writes its own code.
Whether for good or ill, data center planning has exploded. The trillions invested keep the stock market expanding. Over 2,000 domestic data centers have been proposed. Believing this will be lucrative, people rush to get in on the fiscal elevator. The "fear of missing out" is a herd mentality driving an economic feeding frenzy. But this may be another economic bubble, which could crash like others have before.
Putting aside the determination of how AI could help or destroy our economy, the reality of what is required to build and operate these data centers has to be considered. They demand massive amounts of money, land, cooling water, and electricity.
The existing electrical grid is inadequate to supply the sudden increase in power demanded, causing regional power price jumps and degraded grid reliability. Grid expansion is expensive and time consuming. Additionally, the large transformers essential to grid nodes, each one uniquely designed for that location, take years to fabricate and deliver.
But AI is a rush to be first, and can't afford to wait, so many data centers are now planning energy production systems co-located on their site. A few are expecting to be fusion powered, although this technology is still at the laboratory level. Others plan to use small modular reactors, which are still being developed, so cost and longevity are yet to be determined, let alone reliable delivery times.
Less speculative plans will use banks of natural gas or diesel generators. But the natural gas fired turbine wait list is already five years and growing, and the war on Iran has perturbed the global natural gas market. While domestic natural gas prices are still holding steady, our most productive gas fields are experiencing production declines, and may have peaked. Domestic diesel prices have risen 50 percent due to the war.
The war also has a fiscal impact on data centers. Middle East fossil fuel production has been very lucrative for decades, and those excess funds have been reinvested in U.S. stocks and bonds, subsidizing our ever-growing national debt.
The war has significantly reduced traffic through the Strait of Hormuz, currently running at 10 percent of prewar levels, so that petro-dollar cash flow has diminished. In addition, the president's erratic foreign policy has degraded trust in the America dollar, further stressing the funding of our national economy.
AI is still looking for a profitable product to sell, and most companies have yet to turn a profit, so investors are getting nervous. Data centers demand an enormous number of expensive, very complex computer chips. As investment money gets tighter, the chip companies are "loaning" money to the data center developers to keep their industry solvent, further eroding investor confidence. While more money is required to keep AI development moving, we find the global economy less willing or able to contribute further.
Climate driven droughts are increasing, affecting agriculture and power generation. Data center water demands make things worse. Public backlash has risen dramatically, cutting across partisan lines, driven by concerns about noise, water, electricity, and economics, forcing about half the proposed data centers to be canceled.
In addition to these structural hinderances to data centers, the war still has no solution. The strait is still constrained. We continue to burn up finite strategic reserves. The diminished energy capacity of the world has yet to hit with full force, now estimated to arrive in September. When the issue is no longer just rising prices, but also limited availability, the economy will probably shrink. At that point, if AI is a bubble, it will likely pop. With so much of the economy focused on AI, some people think the housing crash of 2007 will become just a preview of coming attractions.
This could all happen before the midterm elections. What an interesting time to be alive!
