The tech world is buzzing about AI cost-efficiency. Some big brains are comparing it to coal, suggesting that cheaper AI will lead to massive usage. But hold up! This ‘less is more’ idea has some serious flaws. Let’s break down why the logic isn’t as solid as it seems.
Here’s a quick rundown of what we’ll be covering:
- The surprising comparison of AI to coal.
- Why cheaper AI doesn’t guarantee success.
- Real-world examples that debunk the ‘less is more’ theory, including fracking, solar panels and genetic sequencing.
The Coal Paradox: Is AI the New Fuel?
Tech leaders are drawing parallels between AI and coal. They’re referencing the Jevons Paradox, a concept that shows how increased efficiency can actually lead to *increased* consumption. Think about it: when steam engines got better, they used less coal per train trip, but because it was cheaper, people ran *more* trains, leading to a net increase in coal use. It sounds good, right? If AI gets cheaper, we’ll use it more, which should be great for business. But that’s not the whole picture.
Why Cost Efficiency Isn’t Always the Answer
The problem with blindly applying the coal analogy is that it assumes that demand is always there. Just because something becomes cheaper doesn’t mean the world will beat a path to its door. Some analysts and execs like Microsoft’s Satya Nadella have stated the rise in efficiency will make AI use explode, but historical examples shows us that it might not be the case.
Let’s take a look at some examples where efficiency didn’t guarantee a gold rush:
The Fracking Fiasco
Hydraulic fracturing (fracking) made it much cheaper to extract oil and gas. The result? The US became the world’s top crude oil producer. But the increased production didn’t lead to a surge in demand, and lots of fracking companies went belly up. The Institute for Energy Economics and Financial Analysis reported that in 2019 about 42 fracking companies went bankrupt, with about $26 billion in debt.
So, just because you can do something cheaply doesn’t automatically mean everyone will want it in huge amounts.
Solar Panels: Cheap but Highly Competitive
Solar panel costs have plummeted, and demand has soared. Global solar power capacity increased ten-fold between 2013 and 2023. Awesome, right? Well, not entirely. Because solar panels are mostly interchangeable, competition is fierce. Manufacturers are struggling with razor-thin margins. Even with record sales, the big players in the industry are not seeing massive profits, according to the National Renewable Energy Laboratory. Cheap can be great for consumers, but not always for business. It can lead to commodification and reduced profitability.
The Genetic Sequencing Squeeze
The cost of sequencing a human genome has gone down from about $100 million in 2001 to around $200 today, thanks to companies like Illumina, which controls the market for advanced sequencing machines. It should be a cash cow, right? Despite the technology being very useful and having huge potential, and being a near-monopoly, it’s not making astronomical profits. Why? Demand for genetic testing is still developing, and there’s growing competition. According to LSEG data, analysts expect Illumina’s revenue in 2026 to reach only $4.7 billion, not much different than in 2021.
Even if you’re the market leader, you still have to run fast just to stay in place, even in markets where you’d expect a lot more growth.
The AI Reality Check
So, can AI avoid the same traps? Maybe. It has the potential to be incredibly useful and make a lot of money. But there’s no guarantee. Just because something is cheap doesn’t mean it will be profitable. In the words of Dario Amodei, CEO of Anthropic, the market might thin out as the tech gets more difficult to develop and only few companies dominate, as novel techniques are still rapidly emerging. If the market is dominated by only a few or even one company, it does not guarantee extraordinairy profit margins.
The Bottom Line
The idea of cheaper AI leading to massive profits is appealing, but it’s not a sure thing. We need to learn from the past and not assume that ‘less is more’ always equals success. The future of AI depends on more than just cost-cutting, and it’s crucial for investors and businesses to avoid some of the pitfalls of previous technological advances. The key is not only to make it cheap, but also to make it incredibly valuable, hard to copy, and with a sustainable demand.