Micron CEO Says Memory Supply Will Be Much Tighter in 2027 and 2028 Than in 2026
Thread
Unofficial Hacker News client; not affiliated with Y Combinator.
Micron CEO Says Memory Supply Will Be Much Tighter in 2027 and 2028 Than in 2026
Unofficial Hacker News client; not affiliated with Y Combinator.
ocd · · focus · HN ↗
It makes me consider whether there would be retaliation in later years against the US intellectual property system in countries other than China. While RAM is a tangible product, the areas it serves aren't really.
malfist · · focus · HN ↗
tzs · · focus · HN ↗
Where does he say he's going to make sure prices stay high?
zehaeva · · focus · HN ↗
tzs · · focus · HN ↗
The CEO is saying that that is the situation we are in.
Demand has increased massively over a short time. Supply cannot increase as rapidly because for big demand increases it involves building new factories and equipment which takes significant time.
The CEO says that their new stuff will start coming online in 2028 but they can't forecast when supply will catch up.
ffsm8 · · focus · HN ↗
or less, as you seem to forget what this whole discussion is about.
If they dont make sure the prices dont go up, they go up.
Once theyre up, and expected to stay up for the foreseeable future, competitors see a gigantic opportunity. Suddenly, the large amount of investment necessary to the ball rolling is worth it.
Once this opportunity is taken advantage of, the original market position of the company being able to freely set whatever price they wanted is eroded forever -- and in this particular case: it opens them up to being out-competed by the new competitor(s).
In this case those will will likely be coming out of china and be state backed, so theyll likely operate at a loss for maximum damage to get as much of the market as they can get.
in 5 years, Micron is likely going to go cry for state support to socialize their upcoming losses.
and before you say "but there are other competitors already": you should be aware that theyve previously been convicted of running a cartel. And the way theyve been behaving is exactly how they were behaving back then. While unproven, it is more then just likely that theyre still price fixing (and have been for years)
tzs · · focus · HN ↗
ffsm8 · · focus · HN ↗
the price didnt just go up for consumers, they themselves have deeply dipped into the opportunity to maximize profits.
its so hilariously outlandish that eg SK Hynix has to pay out >$450k to union employees (per employee - not in total!) because they have a contract that a percentage of profits get shared as a bonus across the company. this payout is usually not particularly high.
rcxdude · · focus · HN ↗
ffsm8 · · focus · HN ↗
the consumer price barely matters for the perspective of the company evaluating wherever they can enter the market, because thats just a tiny fraction of it. And the largest piece of the market buys directly from the producer, SK Hynex, Samsung and Micron.
For consumers wanting to buy RAM sticks the difference is academic, i give you that. But for the capital that decides wherever to invest millions if not billions into getting a factory up and running - the price they can set for the b2b sales is the more important factor.
If it wasnt already proven to be basically whatever number they feel like setting... the opportunity would be a lot less obvious.
selectodude · · focus · HN ↗
The idea that companies should have priority price controlled access to memory puts us even deeper into a situation where we can never afford local compute ever again.
tombert · · focus · HN ↗
I actually like Clojure more but even with GraalVM AOT it still tends to take significantly more memory than even naive Rust code (e.g. using Clone everywhere). If you’re looking willing to spend time abusing the borrow checker you can often get memory usage down to like 5% of what you’d get out of Java stuff.