We’ve been in a global memory shortage for about a year now, causing many of our systems and PC parts to drastically increase in price. Since then, it’s very much been a guessing game how long it will all go on for. The primary cause of the shortage has been the insatiable demand of AI data centres, so we’ve naturally assumed if that dies down or if memory production can match it, all will be well again.

We might worry that the Tesla and SpaceX CEO’s proclamation that “Robot RAM needs will exceed everything else on Earth” will prevent this. But it’s not something we should worry about for a very long time, as even if the amount of RAM robots use increases and we start producing way more robots, they will still demand far, far less than AI data centres currently do.

There’s an element of flippancy to Musk’s comments, of course, given he’s saying that AI data centers will be housed in space and therefore robots will be the biggest guzzlers of RAM on the planet—this doesn’t imply that robots will require more RAM than AI.

In the context of what Musk is replying to on X, it’s nevertheless clear he’s saying they will demand a lot. Musk’s comments are a defence of some Counterpoint research that reportedly claims that by 2030, humanoid robots will demand 20x more DRAM than they do right now.

P Equity Research has shared a chart from Counterpoint which lays out this predicted increase in demand. Apparently, the current 8,300 Tb worth of DRAM demanded by “humanoid robot brains” will increase to 172,000 Tb by 2030.

Aside from the obvious fear of a Skynet future, the worry here is that this is another thing that could put severe strain on the memory market and keep the price of our PC components sky-high. There were already predictions that memory supply will still fall short of demand into 2030 and even beyond, so adding a potential 20-fold increase in demand from humanoid robots into the mix can’t bode well, right?

Well, let’s not be so hasty. The current global DRAM capacity is estimated to sit at about 40 exabytes (EB). One exabyte is equal to 1,000,000 terabytes (TB), which is equal to 8,000,000 terabits (Tb). 40 EB is 320,000,000 Terabits.

So, when Counterpoint predicts robots will demand 172,000 Tb of DRAM by 2030, it’s predicting that robots will demand about 0.05% of the current global DRAM capacity. That’s a miniscule sliver of what AI currently demands and what memory makers actually produce. By 2030, the sliver will seem even smaller.

So, I’d say moving our data centres to space would pretty much have to be the requirement for such humanoid robots to have any meaningful impact on the memory shortage. The ball is still firmly in the court of the AI bubble and how that all pans out.