TPM Reader EB emailed today to tell me something that hadn’t come across my radar: the cost of computer memory is going absolutely through the roof. Just do a Google search for something like rising cost of computer memory and you’ll see a ton of articles. To give you a sense of scale the cost increases are approaching 200% year over year and as much as 30% for certain kinds of gaming RAM recently in one week. The cause is what you’d expect: the insatiable demand for memory created by the AI server farm buildout. I buy computer memory too but I don’t think I’ve tried to buy recently enough to be aware of the surge.
I told EB that I continue to find all of this surreal.
I’m familiar with both the techno-optimist nutball theories of the AI boom as well as those I consider reasonable and fact-based. It’s obvious that we’re in a bubble. But that doesn’t really answer the most important question. There was an internet bubble that gave us the dot com bust. But that didn’t mean the internet was fake or stupid. That was just a shaking out on the path to the 21st century world we live in today. There was a big railroad build out bubble too. Bubbles are almost inevitable in the early stages of transformational technology. Railroads and the internet totally transformed society. But this little data point, the cost of computer memory at the consumer and gamer level captured things for me in a new way. Say what you will about valuations, theories, AI-driven sky-high equities market numbers – this small fact point brought it home for me. There’s never been anything quite like this. You didn’t have other industries scrambling to find steel during the railroad boom. There was really nothing like this in the late 90s. Yes, things were crazy. There were hot-shot companies, wild theories of the future, whole industries popping up around the web browser. But the boom didn’t drive these wild scarcities for things that are basic building blocks of early 21st century society. That’s new, and it’s different. (And yes, I know there are similar things happening with water and electricity. There are many versions of this. I’m just saying what pushed my perception of the situation into a different place.)
So let’s not talk about a bubble, which is inevitable. Maybe I should call it a fakeness bubble, something that is not just turbulence and creative destruction on the path to building out a societal transformation but some more fundamental misallocation of resources that was based on bad facts and over-concentrated decision-making power. Is it a fakeness bubble?
Whatever the future of AI might be you cannot separate this situation from the wealth concentration and platform dominance of the big tech companies. By which I mean how the decisions of a relatively few people can have such a vast impact on the resource allocation decisions of the whole society, of the whole world. Nor can you separate it from the fact that you’ve got a half dozen or so mature, dominant companies in what amounts to a kind of pre-World War I great powers arms race for dominance. The whole thing is driven less by the prospect of new profits for the tech megaliths than the prospect of annihilation if they lose the race for AI dominance.
As I said above, I don’t know enough to make any strong statement about AI, what its value is, where this is all going. But this kind of resource allocation seems of an order of magnitude, breadth and unanimity that is in a different category than anything I can think of in the past. And it’s all bound up in the ultra-power and ultra-wealth of Silicon Valley.