Micron locks in historically high memory prices for five years
theregister.com
theregister.com
Now, with AI, software becomes less able to demand the margins it once did.
Meanwhile, the history of low margins of hardware have created a situation where there are so few players able to demand now software-style high margins.
Hardware has always been valuable but was unable to capture it's value. Those days might be over.
I hope this encourages people who would build software companies to look to hardware. A lot of fun challenges there. Deeply technical, interesting ones. And now solutions will pay.
They are in saas metrics territory in terms of margins, this is insane.
Does anyone really think that there is any agreement in the world that will keep companies paying $1000 for a product priced at $20 on the market? The larger the gap the larger the incentiv to break the agreement.
This agreement puts a floor on the price that Micron can sell their chips at, and a ceiling on the price that tech companies buy them.
Micron wants the floor, so they can invest in more fab capacity without going bankrupt if there's a memory glut.
Tech companies want a ceiling so they can keep selling their products even if there's a shortage.
If there's a glut, the tech companies will just resell these chips at a loss, or take a bath on their $22 Billion of deposits. It doesn't "lock in high prices" for consumers, just the producers.
A higher price floor for longer creates an incentive for competitors to bring more capacity online. The eventual supply glut will be even bigger.
The cyclicality of the semiconductor business (especially for simpler chips) is a result of the fact that inventory can build up. That hasn't changed and it never will.
So clearly 16 large buyers consider it likely that prices will go even higher. How likely? >10% chance? Likely enough to sign an agreement.
Turns out Apple fucked Micron in 2023 so they couldn't expand capacity: https://9to5mac.com/2026/06/25/micron-exec-suggests-apples-a...
Doubt it. Has it EVER happened before?
In all seriousness, the payoff of a real competitor not in the cartel entering at some time in the next five years would be huge. They would have business through the busts because people would go to them first. The challenge will be fighting corruption every step along the way. They would have to keep a sharp legal team on staff for all the litigating necessary to defend against anti-competitive practices and even then would only succeed in a legal and political environment accepting of anti-corruption enforcement.
It lives with some of the other things I'm grieving due to the AI boom, like Apple's car project.
Just don't buy consumer electronics for the next 3 years. You might even be able to buy a non-smart TV in the near future because of RAM prices!
> To date, five manufacturers have pleaded guilty to their involvement in an international price-fixing conspiracy between July 1, 1998, and June 15, 2002, including Hynix, Infineon, Micron Technology, Samsung, and Elpida.
It is history; we have not learned; we are doomed to repeat it.
[0] can actually be anyone
if the chinese DRAM is any good, prices should start stabilizing... I wouldn't get my hopes up too high on them going back to the before times level anytime soon, however.
By giving them stability of cash flows, the AI companies are enabling them to make those investments and to ramp up production. That's a good thing, not a bad thing. Over time it should ease the squeeze on chips.
For decades now we have seen the expectations that software businesses (and in particular FANGs) have pushed any hardware margnins to be more and more like commodities, while they were extracting all the value.
Decades ago, when memory production still existed in many countries, no such margin increases would have been possible.
Even now, this would not have been possible without the US government actively suppressing competition in the memory market, by sabotaging the Chinese memory producers.
The so-called "sanctions" against the Chinese memory producers have started some years ago precisely in the moment when Micron was threatened to lose market share to the Chinese producers (e.g. when Apple was considering to switch to them as providers). Based on the "Cui prodest?" principle, it is extremely likely that Micron was the entity who lobbied the US government to sabotage the Chinese memory producers, creating the environment where companies like OpenAI could successfully drive the memory prices to record levels.
And software isn't?
Because software you build it once and can resell it as much as you want, a stick of RAM is only sold once.
> The same was reflected in engineering salaries, with software engineering salaries being often a multiple of hardware engineering ones.
Arguably software is much more difficult to build because it is never complete.
Yes exactly, and theoretically if there was a functioning market that should have driven prices way down, because the marginal cost to enter the market is so low. However, the non-functioning anti-trust laws as well as the expansion of ip laws as allowed a few enormous corporations to essentially control the market and keep prices up.
> Arguably software is much more difficult to build because it is never complete.
There are hardly any hardware companies where the hardware is ever complete. Disregarding the fact that many hardware products contain some form of software that needs to be updated. In pretty much any field hardware companies need to continue developing new revisions/improvements on the hardware to stay ahead of their competitors, however those revisions/improvements are significantly more complex to put into reality.
What? Software is not expansive, most of it is free. Its just that the same code can be used by more than one user, unlike hardware. The high profit margins come from economies of scale which is greater than whats possible with hardware.
> There are hardly any hardware companies where the hardware is ever complete. Disregarding the fact that many hardware products contain some form of software that needs to be updated. In pretty much any field hardware companies need to continue developing new revisions/improvements on the hardware to stay ahead of their competitors, however those revisions/improvements are significantly more complex to put into reality.
Most of the deployed hardware is complete. The RAM in my PC is complete but the software running on it needs to be updated every month or so.
They know that as new factories come online the current shortage will be resolved and their pricing power will disappear. I was equally critical when Microsoft used these types of contractual provisions.
Why not just sell on the open market, and let traders and financiers and all their prediction models give you the best possible price?
You could argue strategic deals like this allow the manufacturers to make the massive capital investments in manufacturing capacity to increase supply.
Otherwise it might be simply too risky - and in the end you end up with lower supply and higher average price.
Moreover, the AI companies have not bought anything with their own money, but with the money of naive investors who believe that their money will be used by the AI companies to buy things out of which they will be able to get the most value.
So for now, this is strictly only speculation, which has driven the prices sky high. It remains to be seen who will really get any value (besides Micron, NVIDIA and the like, who have got good money for their products).
Money was supposed to be a means by which it is recorded what someone has given to others, so that they may receive equivalent resources in return. But now money has retained this function only for employees and other low-income categories.
> where some people have access to huge fictitious resources
The "huge" adjective seems to be the core of the problem.