OpenAI triggered this ordeal with its order of 900K DRAM wafers per month on october 2025, making everyone panic because it was around 40% of the world production at the time.
We all know that memory providers know what near bankruptcy feels like when they over invest, but right now SK Hynix have 80% margins (ie TSMC ~60%) and choose to pay their shareholders this year the quivalent of ALL their revenues in 2025. It's now common knowledge thanks to Samsung that the rate of factory buidling is not maxmium to portect their margins. And memory providers have already been fined in 2000 for price fixing.
Supply is short because they all choose to make it that way.
Demand has increased for high-bandwidth memory. That increases prices and quantity.
The increase in quantity has happened. Quantity sold has increased. Fabs have increased production and shifted away from other products including consumer products.
Prices increase when there is an increase in price of a substitute in production.
New fabs are planned, consistent with profit-maximizing behavior increasing production in response to increased prices, but they take time.
Sam Altman can eat my entire ass.
Careful what you wish for, Sam may actually be willing to do it.
mainly because the thing about open weights is folks will just immediately strip it back out.
But I can guarantee you that openai/anthropic/other will absolutely jump at the chance to monetize users by feeding them paid recommendations.
1. Local inference can cause a large under-utilization of resources from users that do not run local inference hardware at capacity over the hardwares lifespan. This is under-utilization and under-utilization is not optimal from an economics perspective.
2. The concentration of inference hardware to singular providers can aggregate power to single systems or organizations, and their incentives may be at odds with individual users.
The same power dynamic and economics struggle has existed nearly everywhere for all of history. For example, if you want to decide to own or rent a piece of heavy construction machinery for business, you will find a long list of do-not's on the rental contractual terms, but you are more than welcome to buy your own and operate your own machinery at several times the price for the hours used and use it however you want.
Must why people in this HN thread are throwing out so many ad-hominems instead.
Also, I'd be very surprised if the current rush to build datacenters everywhere is optimal usage of resources. How many companies are building datacenter commons and sharing with other companies to optimize resource usage?
> Two statements can be true
As I general point, because I keep seeing this line on HN: Of course they can in the general case, but more often than not, the reason why this line is written is because there is some tension or contradiction between the statements. In that case, understanding where the tension lies is IMO more useful than shrugging and deciding to "hold two contradictory thoughts in your head" like in 1984.
Sometimes it feels like a rhetoric device to "rescue" an opinion when the opposite opinion is hard to argue against.
I actually use both gpt and local qwen and I personally think the question tends to be of what one needs ( speed? privacy? lack of corporate silliness? ).
Still, it is sad that we can predict the future so well based merely on what happened before.
Edit: I certainly hope I don't have to explain to you why the above is relevant. No market participant would dare to argue this if the competition was healthy.
This is, very literally, capitalism at work. It's the CEO of a massively valuable company jawboning his self-interest.
These massively valuable companies are the ones who want you to "own nothing and to be happy". That's the core tenet of the rental/"sharing" economy.
I use my RAM how the hell I want and nobody else better damn well think about telling me otherwise.