Hazy Memory

If I was Micron and everyone was hating on my company for making life just a little more unaffordable, I might try looking for a scapegoat, too.

But given how little the RAM folks have stuck their necks out in the year of the RAM crisis, this quote from a recent Wall Street Journal story was nonetheless revealing:

In an interview Wednesday night, Micron Chief Business Officer Sumit Sadana said the company couldn’t make investments during the memory market’s last downturn, when Micron’s gross profits went negative, in part because certain customers took advantage to pay rock-bottom prices.

“We told a couple of the customers who were being very aggressive with pricing at that time that this is not constructive,” he said, without naming Apple, adding that low prices discouraged capital investments. “A lot of the industry investments got shut down in 2023 because of really poor pricing and really poor margins.”

On the surface, the logic checks out. RAM takes a long time to build, and when the cycle of supply and demand prevents further reinvestment, the result is that demand becomes completely impossible to anticipate on the surface.

So yeah, hand one to the memory companies. I’m sure when you have a product everyone wants that is extremely difficult to produce and lacks the elegance of anything Apple does, it can be difficult to eyeball your long-term memory demand. (On the other hand, when your profits are surging … it makes it hard to be charitable.)

This, of course, led to the last really huge memory crunch, dating to the late 1980s, when it caused problems for Nintendo cartridges in particular. Back when I wrote about it, I gave it an ending that aged like sour milk:

Megabyte for megabyte, that means it’s unlikely we’ll ever see a 1988-style shortage ever again. We can’t make better oil, but we can make better RAM.

Yes, it turns out, we can make better RAM. Problem is, sometimes the demand is so much that we can’t account for it with traditional measurement tools. We can't always see the run on tulips before it happens.

If you look at Apple's business model, it is essentially a 50-year bet that it can maximize profit by owning as much of its production process as possible. RAM and storage are, at this point, two of the biggest, most expensive components that Apple does not make in its phones and computers, having created its own ARM-based SoCs. Apple hands those chips to a manufacturer, TSMC, to build for hire, but they are in control of the supply chain.

Part of what drove the rise of Apple Silicon was the very same pressure to rely on a partner with different priorities from Apple. A surface read suggests that this is a fairly similar situation, but it falls apart when you look closer. Unlike oil, RAM does get better over time, as does solid-state storage. But it’s a difficult market to do well. Here’s a video of Micron’s own process of turning sand into silicon:

Doesn’t seem particularly easy, does it? And a good case against vertical integration might be Apple’s primary mobile competitor, Samsung, which also doubles as a major supplier. Samsung makes memory chips, but the market is so screwed up that Samsung isn’t even willing to sell some of those chips to itself. Even with the understanding that the company likely has a better understanding of its own supply needs than anyone else, Samsung is favoring the demand from corporate buyers over its own product lines.

(It’s so bad that Apple is trying to talk up the Trump administration to get some RAM from a Chinese supplier that is otherwise banned from the U.S. market.)

In theory, Apple would vertically integrate everything to the best of its ability. But RAM is not an area where they can really do something different from the rest of the crowd, unified memory excepted. It would cost them a lot of money and time to do it properly, and it might not even turn into anything useful in the end. It’s worth noting that Apple recently had one of those, and it did not work out for them. Apple has the budget to do it, but it’s easy to look at their battered frenemy Intel and see that being a component supplier is not all wafers and clean rooms.

To put it another way, vertical integration is in so many ways Apple’s moat, but the tech that Micron and other big memory suppliers offer can’t be vertically integrated—and they know it. It’s going to make for tough negotiations in the future, because Apple is no longer the biggest company buying all the RAM.

And we, as consumers, suffer.