James Carville posted “The economy stupid” sign in Bill Clinton’s 1992 Little Rock campaign headquarters. He wanted the campaign to focus on what mattered.
The data-center debate could use some focus too. Among other things data centers train and run artificial intelligence models. There is a seemingly irrational race to build the best large language model. Amazon is in the race.
The race seems irrational because the cost is huge by historic benchmarks. And because AI prices from existing models are cratering. And because AI prices from new models may not justify their cost. The price is what matters.
Will Amazon win the race or be an also ran? Will Amazon buy Entergy’s electricity as contracted regardless? Will Entergy’s residential customers pay more if it doesn’t? Entergy says not to worry. All’s good. We can see the future. Small customers will pay $2 billion less over the next 20 years than they will otherwise due to Amazon’s data centers.
That’s a 20-year prediction for a business that’s changing faster every year. An AI engagement that paid $20 18 months ago pays seven cents today. You have to sell something better, produce it cheaper, or both to survive. Hyperscalers are betting on both.
Amazon, Meta, Microsoft, and Alphabet are the big four US hyperscalers. They are in an AI arms race. They spent over $400 billion in 2025 and plan another $725 billion in 2026. Most of the spending was for new models. Cash flow covered most of their earlier spending. Now they are adding debt and financial risk to technology risk.
They are spending more in one year than the total cost in today’s dollars of the Apollo Moon Shot, the Marshall Plan to rebuild Europe after WW2, and the Manhattan Atomic Bomb project. That’s a lot of eggs in the same technology basket.
There are other players too. They include the Stargate consortium (SoftBank, Open AI, Oracle, and Abu Dhabi’s MGx) and its $500 billion and Elon Musk and his pocket change. Then there’s China with its robust electric grid, plenty of electricity, and abundant brainpower. Those are big competitive advantages.
Will the market be big enough for everyone to be profitable? Maybe. But maybe not if super model prices fall too fast. And that’s likely if their AI supply exceeds demand.
Reminds me of my years in the fertilizer business. It’s notorious for feast and famine cycles of under supply and high prices followed by capacity additions, over supply, and low prices. Our modern plants made us low-cost producers, and we had good profits when demand exceeded supply. But when supplies were long, we ran the plants at a loss to avoid shutdown and startup costs while waiting for prices to recover. Fortunately, our traders who had no plant investments made money then buying and selling fertilizer from others who did.
Today some smart money is betting that buying and selling AI will be more profitable than owing it. Apple is one of the richest companies on earth. It could be in the AI arms race. It’s not. It pays Google about a billion a year to buy AI for Siri. Perplexity, an AI search company, does the same thing. It has a model it uses to buy AI from the cheapest model that best fits the job and sells the answers. I use it to fact check my articles.
Why the seemingly irrational race for super models? It’s likely to lead to mutually assured excess supply. That’s happened before. The Nash Equilibrium explains the madness. This game theory concept was Hollywooded up in A Beautiful Mind starring Russel Crowe. The memorable scene is four college boys (hyperscalers) in a bar eyeing a beautiful blond (super model) that only one or maybe none of them could win.
The theory says each hyperscaler’s spending is rational if other hyperscalers keep spending too — even though the results may not good be for anyone. If Amazon stops spending and the others don’t, it may lose access to chips, engineers, data center sites, customers, etc. — and lose competitive standing. Nash’s theory does not predict a good outcome. It just explains why no competitor can quit the race if the others don’t.
Then there’s the China wild card. China makes the race a national security threat as well as profit threat. This may explain free models from China.
Have irrational races happened before? Yes. The 1996 - 2000 telecom and fiber optic race. It cost competitors about $500 billion in today’s dollars. No company wanted to be left behind. The resulting technology created the internet economy. But competitors spent too much on too much capacity. Prices cratered. The S&P communications equipment makers index fell 86% — $793 billion in market value evaporated.
AI can change the world and still destroy capital and bankrupt investors. Fiber optics did. Railroads did. Is this a wash, rinse, and repeat cycle? Keep your eye on the price.
The classic Dr. Strangelove movie is about the hydrogen bomb race. Neither Russia nor the US could quit it. The memorable scenes are Major Kong (Slim Pickens) astride the bomb whipping his cowboy hat and the mushroom clouds that doomed everyone.
Major Kong reminds me of hyperscalers whipping an AI slide deck. And Entergy whipping: “We can see the future and it’s all good.” As Yogi Berra is said to have said: “the future ain’t what it used to be.” The present ain’t like the past either. Things have changed a lot since Mississippi got in the data center game.
I am not opposed to data centers. I’m not a Luddite. I am opposed to Entergy’s misleading customers about unknowable future costs. And to the secret deal that makes residential customers more vulnerable to technology and market risk — and Entergy’s monopoly less vulnerable.
AI is the wave of the future. It’s a generational disruptive technology. It will probably transform our economy, culture, and lives — in ways we can’t predict. There will be winners and losers. The market will sort them out if left to work. Our politicians and central planners don’t have a good track record picking technology winners with taxpayer money.
They don’t have skin in the game. Maybe they should quit.
Kelley Williams, a Northsider, is chairman of Bigger Pie, a Jackson-based think tank promoting free markets and government efficiency.
