Dane Vahey of OpenAI made a statement that doesn’t sound real but somehow is: the cost of processing a million tokens—the amount of computational power needed to generate something semi-coherent in AI—has dropped from $36 to $0.25 in just 18 months. Let’s stop here for a moment, because that’s almost too big to compute. That’s like saying an iPhone now costs the same as a McDonald’s cheeseburger. It’s not just a reduction in cost; it’s an economic collapse of precedent.
In the span of a year and a half, AI went from being something with a substantial price tag—whether you’re using it for language generation, predictive models, or simply trying to make a chatbot that doesn’t lose its train of thought halfway through a conversation—to something so cheap that you could essentially drown in it. But this isn't just about AI suddenly getting a Musgrave-style bulk discount on computation. This is about how technology, in general, seems to be entering a bizarre, death-spiral loop where costs drop so fast and so hard that they lose all meaning. This has happened before—but never at this speed, and never with a technology that has the potential to rewrite the entire way we think, create, and operate.
Before AI, the closest thing we had to a technological revolution that completely restructured the way the world worked was the steam engine. The steam engine didn’t just replace the horse and buggy—it set the horse on fire and told everyone to get used to it. Suddenly, you could move cargo across oceans and continents at a fraction of the time it used to take, and for a fraction of the cost. But even that wasn’t immediate. It took decades for steam power to really saturate society, and in the meantime, entire industries had to be reimagined.
When James Watt figured out how to make the steam engine commercially viable, it didn’t make horses obsolete overnight. It took years for infrastructure to evolve, for steamships to be built, for railroads to snake across landscapes. The costs didn’t drop overnight; they gradually eroded as companies scaled up and engineers learned how to make engines more efficient. Compare this to AI. We’ve seen an astronomical cost reduction in a matter of months. Not years. Not decades. Less time than it takes for the latest iPhone to become outdated.
Moore’s Law: The Oldest Cliché in Tech
Everyone loves to talk about Moore’s Law, which predicted that the number of transistors on a microchip would double roughly every two years, resulting in exponential growth in computing power. For years, this was the tech industry’s gospel. You could basically bet your 401(k) on the fact that computers would get twice as fast and twice as cheap every couple of years. But Moore’s Law was also predictable; it was a safe bet.
Here’s the kicker: AI is demolishing Moore’s Law. We aren’t seeing a gradual doubling of performance every two years; we’re seeing performance explode in months while costs plummet to near-zero. In a sense, AI doesn’t care about your two-year doubling cycle. AI is like that person who shows up to the gym for the first time and goes from benching the bar to benching 300 pounds in one session. It doesn’t make sense, but it’s happening anyway.
Let’s get some perspective. In the early 2000s, cloud computing began to take off. It was supposed to be the great democratizer of data, allowing anyone with an internet connection and a credit card to access the same computing power that used to only be available to large companies with deep pockets. And it worked—sort of. Prices did drop, but they didn’t fall off a cliff. You could store your data in the cloud for cheaper than buying hard drives, but there was still a tangible cost. AI is doing something different. It’s not just lowering the bar; it’s removing the bar altogether.
The Internet Was Free, Until It Wasn't
Remember when the internet was "free"? Not in the financial sense—you still had to pay for dial-up or broadband—but in the sense that it was an open, uncharted wilderness where anyone could start a blog, build a website, or upload videos without worrying about monetization or data mining. The infrastructure was there, and it was up to you to figure out what to do with it.
But now, everything on the internet is either behind a paywall or being monetized through your data. We’re realizing the "free" internet was never really free; we were just too naive to notice the long con being played by data brokers and advertisers. And maybe AI feels a bit like that right now. Sure, the cost per million tokens is plummeting, but what happens when it’s practically free to generate content at scale? When you can flood the internet with more articles, images, and ideas than any human could ever hope to sift through?
The optimistic take is that AI will make everything easier and more accessible. The pessimistic take is that we’re heading into an era where the sheer volume of content created by AI will drown out human voices entirely.
The Post-Scarcity Economy (Sort of)
There’s this old idea in economics known as "post-scarcity," which suggests that at some point, technology will advance so much that resources become essentially infinite. It’s a utopian concept. You want food? Print it. You want energy? Harness the sun. You want a million words written on any given subject? Just ask AI.
This drop in cost feels like a step toward that post-scarcity world, but the problem with post-scarcity is that it only works if everyone agrees on the rules. When costs drop to near-zero, as we’re seeing with AI, it doesn’t mean the market for ideas or content is suddenly equal. It just means that everyone has the same access to cheap computation, and the people who figure out how to leverage it first are going to dominate the landscape.
In some ways, AI is like a modern-day printing press. The printing press made books affordable, which was a radical shift in human history. But it wasn’t just about making books cheaper—it was about making ideas cheaper. AI is doing the same thing, except instead of democratizing the written word, it’s democratizing intelligence, creativity, and problem-solving. We’ve gone from needing thousands of hours of human labor to get something done, to being able to generate it in seconds for a fraction of a penny.
Dane Vahey’s stat is a marker of where we’re headed, but the real story is what happens next. AI isn’t just cheap; it’s becoming ubiquitous. And when something becomes ubiquitous, it stops being remarkable. The question isn’t whether AI will change everything; it already has. The question is how we navigate a world where the cost of intelligence is trending toward zero. Do we end up with a society that’s better because of it, or do we drown in an endless sea of machine-generated content with no soul?
The answer, as always, is both.
Add comment
Comments