upvote
I'm expecting it to somewhat collapse. I don't know if it'll go back to pre bubble prices (here's to hoping), but I do expect a pretty sharp decline around 2030... probably not before then.

Basically everyone that makes memory is building new fabs, meanwhile I'm not sure how much longer AI datacenter demand for ram will last. I think the decrease in AI ram demand and the new fabs will likely coincide leading to a collapse in pricing.

That is, of course, assuming the memory manufacturers don't pull their favorite trick and collude.

reply
If there's a decrease in AI ram demand, it will not be because the models get better. Models getting better will increase RAM demand, because it grows the part of the economy that models are useful for. Classic Jevon's Paradox.
reply
I'm expecting the drop for a couple of reasons.

For 1, AI datacenter builders have said that they have more equipment than they have places to put them. Leaving a ton of hardware shelved while you wait for datacenters to build out is bad business to say the least.

For 2, I think we are nearing saturation for the usefulness of AI. I certainly could be wrong, but I don't really foresee there to be a bunch of new exciting usages of AI that will ultimately justify the continued buildout.

reply
Re 2: Even if that was true penetration is still very low. If all companies start using AI at scale the demand will skyrocket.
reply
the better AI becomes the more non-programmers will be able to use it to build apps, and because they don't know how to program themselves they will use FAR more tokens than us to achieve the same thing. As long as AI is good enough for them to build whatever they desire there will be a lot of demand imo.
reply
I’ve been thinking about this question for months: with access to good LLMs and unlimited tokens, how many non developers will end up making software?

I’m sure some people will. But most people I know don’t think about software like that. They don’t know what use cases are, or ask themselves how their life could be simpler with different software. LLMs only make software if you prompt them. Most people don’t frame the problems they have in life as “I’m missing software to solve problem X”.

I think most people will never prompt LLMs to write software. Just like I don’t have a 3d printer, because I don’t think about how problems in my life could be solved with little plastic doodads.

reply
My analogy is: almost everyone can to go Home Depot and grab all the plumbing supplies they need to do a lot of jobs. And people do - Home Depot does great business.

But plumbers are still in very high demand, and charge really high prices for what is in many cases really simple stuff.

For a lot of folks, the idea just never enters their mind that they can solve simple problems themselves. Most people don’t want to spend the time, undertake the learning, have the possibility of doing it wrong, risk their property, insurance/code issues, etc. And this is an absolutely reasonable position to take! We can’t all be specialists at all things.

All that to say: people will long think that software is something that is made by someone of a special role. Even long after that’s not necessary anymore. That’s not great job security, but it’s what I’m going with for now!

reply
I've met at least one person that has gone pretty far on a somewhat complicated project who has literally zero programming experience. he did it for fun.

I also know someone that thought about making a SaaS for his specific niche he works in, he asked me if I ever used Claude then told me his project.

Yes, most people will never do it because they're not the type of people that are self-motivated and courageous enough to dive into something with confidence to see how it goes.

But probably, as more people build stuff and it makes headlines that X person built Y service and is making Z$$ per month and he can't even write a hello world to save his life, a lot more people will try it out.

reply
> X person built Y service and is making Z$$

If person X didn’t contribute anything but a prompt, he probably wont be able to charge much.

reply
> Yes, most people will never do it because they're not the type of people that are self-motivated and courageous enough to dive into something with confidence to see how it goes

Or they’re just not interested. Just look around to see how many things that people are doing that you have no interest into. That’s the same disinterest people have towards programming and sys admin.

reply
> For 2, I think we are nearing saturation for the usefulness of AI. I certainly could be wrong

Unless some other breakthrough in AI happens, like applying AI to some other large market.

Overall i think that AI is here to stay, if anything it will become more common.

Maybe in ten years we’ll all have some 256gb ai machine at home, like we most (all?) of us have AC.

reply
I think LLMs will start being etched into custom chips. There’s already some companies experimenting with it. The benefits are enormous - you get better performance, lower memory usage and no need to load a model in ram. Companies would be able to sell local LLMs while keeping their models proprietary.

If that happens, hopefully the ram price comes back to earth.

reply
I’d love to buy a tiny swappable coprocessor that slots in next to my cpu and can take this stuff on. Swap it out every 8 months as you can benefit from something better. Hell, sign me up for a subscription!
reply
Yes, but a right now a lot of the RAM buy is for demand that doesnt exist yet. Big tech is projecting AI demand and making sure they have enough compute to fulfill that, so AI demand would have to increase even more than they expect, otherwise theyll build their datacenters and then slow down the amount they buy.
reply
I'm going with 2029 just so that it can be exactly 100 years from the Great depression. ;)
reply
100%. We're probably on the cusp of over-capacity, a glut, cheap RAM, bankruptcies, and shortages.
reply
The memory companies report that they're sold out through 2027... so it might be a while
reply
That's only if the AI companies continue their build outs.

How many people actually use fable over opus? How far are we up the diminishing returns curve, and will their customers even care?

reply
The stock market crash caused by AI companies halting their planned build outs may in practice cause RAM to remain unaffordable.
reply
How does a stock market crash increase RAM prices (when RAM demand goes down)?

Or would it affect people's purchasing power. I guess to some extent it does for people with significant investments in the stock market, but that's not a huge percentage of the people wanting to buy RAM or devices with RAM

reply
The second and third and fourth-order exposure to AI spending is maybe bigger than you think. The blast radius if it pops could be quite large. Patrick Boyle had an interesting video discussing this recently https://www.youtube.com/watch?v=wTiYaWFP59Q
reply
I don’t think the effects on the hardware market of open weight models triumphing has been reflected upon enough. It’s not clear that it will be less impactful if every enterprise decides to build for predominantly on-premise inference. In fact, before the question is ultimately decided, we’re probably heading towards a few years where hobbyists, enterprises and ‘hyper scalers’ are all competing for certain parts in common. Ram booked through 2027 sounds about right.
reply
Companies are avoid risks, and at this stage, sometimes, I feel that all cloud providers are just buying RAM that they would have bought anyway. Now it's ensure that no company will be willing to pay x digits just for cards. One of my clients is stuck in "we are doing things on premise but we are too cheap to spend a few 100k in cards but we don't want to go on clouds.
reply
Last thing i heard was that large companies are starting to seriously spend on ai infrastructure (think nvidia clusters rather than tokens from openai or anthropic) to fine-tune and run custom models.

This might become the driver for memory and chips demand in the near future.

But i don’t know, of course, we’ll i guess.

reply
Fable would have a much bigger userbase if Anthropic didn’t remove a zero data retention option because muh safeguards
reply
Or if it didn’t constantly kick people off fable. It won’t talk to me about gardening or computing.

It will happily walk me through all the legal and illegal ways to systematically kill rodents.

At least it is on-brand for something used by militaries that want to target civilians, I guess.

reply
Or, they're price fixing, as they have many times before, and are currently being sued for [2].

[1] https://en.wikipedia.org/wiki/DRAM_industry_price_fixing

[2] https://moginlawllp.com/dram-antitrust-suit-ai-memory-supply...

reply
I for one can't wait. Current prices are absolutely insane.

However a problem exists with scalper bots. They're going to fight tooth and nail to keep prices artificially high even when supply increases. Same with the hyperscalers blowing "free money" at eating supply as a strategic position against smaller competitors.

reply
Scalper bots need customers to sell to; their business model falls apart in the face of sufficient supply.
reply
That sufficient supply is maybe on the distant horizon. In the meantime everyone is being fleeced because the scalpers have the only supply available. Even if there's a supply glut tomorrow the scalpers have the infrastructure in place to buy out retail channels keeping prices high.
reply
Looks like I'll be sporting my 5800X3D + DDR4 + 9070 XT gaming box for a little while. Too bad the CPU's ST is slower than my MB Air m4.
reply
I built a new PC about two years ago, and I probably got it at the last possible opportunity for a while. CPU and motherboard have come down by maybe £100 in between the two of them, but a 7900 xtx (or any other 24GB GPU) for under £1,000 now seems like a bargain, and £180 for 64GB of DDR5 makes me feel like an old man talking about the halcyon days.
reply
I built my current PC the day that the AM5 platform released, for about $6k not including the 3090 I moved over from the previous rig.

If I sold just the two sticks of RAM in it right now, it’d pay for nearly half of the total cost.

reply
I bought a motherboard from Newegg in late 2024. It came with 16GB of RAM as a free "gift". The same RAM goes for $534 on Newegg today. It was $437 in January. Prices are still creeping up for this "throwaway" memory.

I still have it, but never installed it because I only use ECC memory in my desktops. I'm saving it for a future Desktop in case memory prices never fall again.

reply
Welcome to the club. If you're _really_ competitive in cs2, I'd swap out to a 9800x3d setup, but it's still a maybe. Very little reason to upgrade right now other than to run LLMs.
reply
It will never make sense to me to run Llms locally unless I had 50k. I don't even think that would compete with price perf of a remote llm and getting business done. And that's completely ignoring that sol/fable level is not local
reply
Yeah, instead I have unsubbed from WoW this week. The atrocious state of the game's code and scripting is unbearable.
reply
Absolutely, I am in a hold pattern for the next few years. I suspect in another 3-4 years you will be able to build an absolutely stacked machine at a reasonable price. Probably won't go back to what was before but it will be decent.
reply
If it will happen in 100 years it will practically never happen (for us). Even 25 years would be a lot, it's half of a career.

Could you provide more details about the Epyc + DDR4 server?

reply
Cycles tend to be 5-10 years long not 25. Without knowing anything else I would expect a fab you seriously start planning today will be at full capacity in about 5 years. Nobody serious likes delays - in particular the banks don't like loaning money that won't at least start paying off. They know it takes some time to design a building - but factories typically are standard buildings so once you know about the size you can get it done fast - I expect 1 year to have the building done is the worst case (and it can be done in 3 months possibly if your project management is good - after interest this is cheaper than the 1 year). It takes time to build and install the specialized machines that go inside - this is the largest problem, but you typically order them first and then plan the building around the needed space and when they will arrive. Then you need 6 months to setup the inside of the building. From there it is just ramp up time.

The above is a standard project management problem. We do this for lots of industry all the time. There is every reason to think you can get a new factory running in 5 years.

Note that I said 1 factory above. Some of the special machines we don't have the ability to make them fast enough to do 2 (I don't know the real number!) new factories in 5 years. Existing factories are using most of the special machine capacity to replace machines that wore out on the way - this can be corrected as well, but it adds another year and the expenses are much larger. Realistically though 1 new factory is likely enough.

reply
If you don't understand that the AI driven memory boom has totally broken the cycle you are going to lose a lot of money. There is infinite demand for Intelligence and that translates directly to chips.
reply
Sounds like you are telling us it’s a new paradigm?

https://transportgeography.org/contents/chapter3/transportat...

reply
"If you don't understand that the AI driven memory boom has totally broken the cycle you are going to lose a lot of money."
reply
Fortunately there's a near infinite demand for boosterism too.
reply
> The DRAM market is cyclical.

...

> I don’t think anyone truly knows when, but it will happen.

Do you know what cyclical means ... ?

reply