I don't buy the central thesis of the article. We won't be in a supply crunch forever.
However, I do believe that we're at an inflection point where DC hardware is diverging rapidly from consumer compute.
Most consumers are using laptops and laptops are not keeping pace with where the frontier is in a singular compute node. Laptops are increasingly just clients for someone else's compute that you rent, or buy a time slice with your eyeballs, much like smartphones pretty much always have been.
I personally dropped $20k on a high end desktop - 768G of RAM, 96 cores, 96 GB Blackwell GPU - last October, before RAM prices spiked, based on the logic that hardware had moved on but local compute was basically stagnant, and if I wanted to own my computing hardware, I'd better buy something now that will last a while.
This way, my laptop is just a disposable client for my real workstation, a Tailscale connection away, and I'm free to do whatever I like with it.
I could sell the RAM alone now for the price I paid for it.
We won't be in a supply crunch forever. We'll have a demand crunch. The demand of powerful consumer hardware will shrink so much that producing them will lose the economics of scale. It 've always been bound to happen, just delayed by the trend of pursuing realistic graphics for games.
People who are willing to drop $20k on a computer might not be affected much tho.
> People who are willing to drop $20k on a computer might not be affected much tho.
They probably won't, but those willing to drop $3-10k will be if the consumer and data-center computing diverge at the architectural level. It's the classical hollowing out the middle - most of the offerings end up in a race-to-the-bottom chasing volume of price-sensitive customers, the quality options lose economies of scale and disappear, and the high-end becomes increasingly bespoke/pricey, or splits off into a distinct market with an entirely different type of customers (here: DC vs. individuals).
Not exactly the same (it's about power rather than price). But close enough that when you said it, I thought, "oh! there is something like that." There's also more fundamental economics laws at play for supply and demand of a resource / efficiencies at scale / etc. Given our increasing demand of compute compared increasing supply (cheaper more powerful compute). I expect the supply will bottleneck before the demand does.
I rarely doge a chance to shit on Microslop and its horrible products, but you don't use a browser? In fact, running all that junk in a single chromium instance is quite a memory saver compared to individual electron applications.
>we're at an inflection point where DC hardware is diverging rapidly from consumer compute.
I thought the trend is the opposite direction, with RTX 5x series converging with server atchitectures (Blackwell-based such as RTX 6000 Pro+). Just less VRAM and fewer tensor cores, artificially.
Where is the divergence happening? Or you don't view RTX 5x as consumer hardware?
> I personally dropped $20k on a high end desktop - 768G of RAM, 96 cores, 96 GB Blackwell GPU - last October, before RAM prices spiked […]
768GB of RAM is insane…
Meanwhile, I’ve been going back and forth for over a year about spending $10k on a MacBook Pro with 128GB. I can’t shake the feeling I’d never actually use that much, and that, long term, cloud compute is going to matter more than sinking money into a single, non-upgradable machine anyway.
Your battery is going to suffer because of the extra ram as well.
I don't know your workloads, but for me personally 64 GB is the ceiling buffer on RAM - I can run entire k8s cluster locally with that and the M5 Pro with top cores is same CPU as M5 Max. I don't need the GPU - the local AI story and OSS models are just a toy for my use-cases and I'm always going to shell out for the API/frontier capabilities. I'm even thinking of 48 config because they already have those on 8% discounts/shipped by Amazon and I never hit that even on my workstation with 64 GB.
> Your battery is going to suffer because of the extra ram as well.
No, it won't. The power drain of merely refreshing DRAM is negligible, it's no higher than the drain you'd see in S3 standby over the same time period.
How is this going to work? You need uncontrolled compute for developing software. Any country locking up that ability too much will lose to those who don't.
My home server has 512GB RAM, 48 cores, my 4 desktops are 16 cores 128GB, 4060GPU each. Server is second hand and I paid around $2500 for it. Just below $3000 price for desktops when I built them. All prices are in Canadian Pesos
Before this price spike, it used to be you could get a second-hand rack server with 1TB of DDR4 for about $1000-2000. People were massively underestimating the performance of reasonably priced server hardware.
You can still get that, of course, but it costs a lot more. The recycling company I know is now taking the RAM out of every server and selling it separately.
It seems like you largely agree with the article - people shall own nothing and be happy. Perhaps the artificially induced supply crunch could go on indefinitely.
Also, I wonder how many of us, even here on HN, have the ability to spend that amount of money on computer for personal use. Frankly I wouldn't even know what to do with all the RAM - should I just ramdisk every program I use and every digital thing I made in the last five years?
Anyhow, I suppose for the folks who can't afford hardware (perhaps by design), one ought to own nothing and be happy.
People spend a lot more than that on a car they use less, especially if they're in tech.
The RAM choice was because I have never regretted buying more RAM - it's practically always a better trade than a slightly faster CPU - and 96GB DIMMs were at a sweet spot compared to 128GB DIMMs.
That, and the ability to have big LLMs in memory, for some local inference, even if it's slow mixed CPU/GPU inference, or paged on demand. And if not for big LLMs, then to keep models cached for quick swapping.
I don't share the same 1:1 opinion with regards to the article,
but it is absolutely clear that RAM prices have gone up enormously.
Just compare them. That is fact.
It may be cheaper lateron, but ... when will that happen? Is there
a guarantee? Supply crunch can also mean that fewer people can
afford something because the prices are now much higher than before.
Add to this the oil crisis Trump started and we are now suddenly
having to pay more just because a few mafiosi benefit from this.
(See Krugman's analysis of the recent stock market flow of money/stocks.)
Open source efforts need to give up on local AI and embrace cloud compute.
We need to stop building toy models to run on RTX and instead try to compete with the hyperscalers. We need open weights models that are big and run on H200s. Those are the class of models that will be able to compete.
When the hyperscalers reach take off, we're done for. If we can stay within ~6months, we might be able to slow them down or even break them.
If there was something 80-90% as good as Opus or Seedance or Nano Banana, more of the ecosystem would switch to open source because it offers control and sovereignty. But we don't have that right now.
If we had really competitive open weights models, universities, research teams, other labs, and other companies would be able to collaboratively contribute to the effort.
Everyone in the open source world is trying to shrink these models to fit on their 3090 instead, though, and that's such a wasted effort. It's short term thinking.
An "OpenRunPod/OpenOpenRouter" + one click deploy of models just as good as Gemini will win over LMStudio and ComfyUI trying to hack a solution on your own Nvidia gaming card.
That's such a tiny segment of the market, and the tools are all horrible to use anyway. It's like we learned nothing from "The Year of Linux on Desktop 1999". Only when we realized the data center was our friend did we frame our open source effort appropriately.
> We need open weights models that are big and run on H200s.
We have this class of models already, Kimi 2.5 and GLM-5 are proper SOTA models. Nemotron might also release a larger-sized model at some time in the future. With the new NVMe-based offload being worked on as of late you can even experiment with these models on your own hardware, but of course there's plenty of cheap third-party inference platforms for these too.
This article inspired me to look and see what this computer is. Apparently it is a "AMD Athlon(tm) II X2 250 Processor" from 2009. So 17 years old. It has 8 GB of DDR3 memory and runs at 3 GHz. It currently has OpenBSD on it, but at least one source thinks it could run Windows 10.
The fact that I didn't know any of this is what is significant here. At some point I stopped caring about this sort of thing. It really doesn't matter any more. Don't get my wrong, I am as nerdy as they come. My first computer was a wire wrapped 8080 based system. That was followed by an also wire wrapped 8086 based system of my own design I used for day to day computing tasks (it ran Forth). If someone like me can get to the point of not caring there is no real reason for anyone else to care.
The general take here seems to be "everything eventually passes". That isn't always true. I wonder how many people have a primary computing device that they don't even have full control over now (Apple phones, tablets...). Years ago the concept of spending over $1k on a computer that I didn't even have the right to install my own software on was considered ridiculous by many people (myself included). Now many people primarily consume content on a device controlled almost entirely by the company they bought it from. If the economics lead to a situation where its more profitable to sell you compute time than sell you computers then businesses will chose to not sell you computers. I have no idea if that is what ends up happening.
I'm also very skeptical of "everything eventually passes" as it pertains to hardware prices. Right now, prices are high because supply can't keep up with demand. But if/when supply increases to meet demand or demand decreases, there's no reason for companies to drop prices now that consumers have become accustomed to them.
Not necessarily. Many people grew up with PCs and laptops but now mostly use their phones, because outside of specific jobs or hobbies, everyday computing needs are heavily optimized for mobile-first.
(A large factor here is, obviously, the cloud. With photos, documents, e-mail, IMs, etc. all hosted for cheap or free on "other people's computers", the total hardware demands on the end-user computing device is much less. Think storage, not just RAM.)
It's true even in tech; half a year ago I switched my phone to a Galaxy Z Fold7, and I haven't used my personal laptop since then, not once. I have a separate company laptop for work, and I occasionally turned on my PC, but it turns out that a foldable phone is good enough to do everything on personal side I'd normally use a laptop for. So here I am, with my primary compute device I don't have full control over - and yes, I'm surprised by this development myself, and haven't fully processed it yet.
A long article begging the question when the last paragraph or two countered the panic of the beginning. Two Chinese firms are ramping up production of consumer RAM/SSDs because they see a market opening as the existing producers move to selling to enterprise/hyperscalars.
There have been memory chip panics before, the US funded RAM production back into the 80s/90s in competition with Japan at the time.
The AI boom/"hyperscale" currently is almost exactly like the dotcom boom.
It's already starting to shake down. Anthropic is occupying the developer space, OpenAI has just exited the video/media production space. More focused and vertical market AI is emerging.
The current vortice of money between OpenAI <-> Microsoft <-> Oracle <-> NVidea <-> Google <-> etc etc is going to break.
> Two Chinese firms are ramping up production of consumer RAM/SSDs because they see a market opening
Yes but these Chinese firms are a tiny share of the overall RAM/SSD market, and they'll have the same problems with expanding production as everyone else. So it doesn't actually help all that much.
Let me be the devils advocate here.
Ok, let's say you optimize that TODO list app to only use 16 mb of RAM. What did you gain by that? Would you buy a smartphone that has less RAM now?
We can't ever escape the market forces? You're right, of course if software gets less bloated, vendors will "value-optimize" hardware so in the end, computers keep being barely usable as they are today.
That's crazy talk. What will you ask for next? Add functionality to make apps at least as good/capable as they were in the 1990s and early 2000s? And then? Apps that interoperate? Insane.
More seriously and more ironically, at the same time, we've now reached a strange time where even non-programmers can vibe-code better software than they can buy/subscribe to - not because models are that good, or programming isn't hard, but because enshittification that has this industry rotten to the core and unable to deliver useful tools anymore.
I've seen comments on here before that went somewhere along the line of "adults don't care about RAM prices." HN is no stranger to siding with the oppressors.
Just to mention one thing, helium -which is a necessity for chip production- is a byproduct of LNG production. And 20% of that is just gone (Qatar) and the question is how long it will take to get that back. So not only a chip shortage because of AI buying chips in huge volumes but also because production will be hampered.
Tongue in cheek: we urgently need fusion power plants. For the AI and the helium.
Well, as long as they can make electricity too cheap to meter, we can get helium from somewhere. Mine it from LNG sources currently untapped due to EROI < 1, or ship it from the goddamn Moon - ultimately, every problem in life (except that of human heart) can be solved with cheap energy.
That's another lifetime-limited thing -- the helium leaks out, and you cannot (for practical purposes) stop it or even meaningfully slow it down. When it's gone, the drives are dead. And the helium leaks by calendar-days, it doesn't matter whether the drive is powered on or off.
Non-helium hard drives are basically limited by their bearing spin hours. If one only spins a few hours a week, it'll probably run for decades. Not so with helium.
Gonna sit on my half-empty tank for party balloons from my daughter's birthday, maybe we'll be able to sell it to pay off mortgage quicker than the helium itself escapes the tank.
Articles entire thesis looks like it can be completely de-railed if one activity happened: ai infrastructure firms cease to be able to secure more capital.
Is that likely? History says it's inevitable, but timeframe is an open question.
> ai infrastructure firms cease to be able to secure more capital
If this does occur, unfortunately it isn’t like any of the production capacity is going to immediately shift or be repurposed. A lot of the hardware isn’t usable outside of datacenter deployments. I would guess a more realistic recalibration is 2-3 years of immense pain followed by gradual availability of components again.
yeah 3 years sounds reasonable to me, less than one asset depreciation cycle in business. Pain for you and me, but just a bump in the road for the accounts dept.
My computer, and I think all threadripper systems, has registered ECC DDR5 RAM which I think is the same type used in AI datacenters. Well one half of it, the other half being HBM memory used on video cards, which is soldered to them and non-upgradeable. But the main system memory from a used AI server can become your main system memory.
So that becomes the next question -- will we see an ecosystem of modifications and adapters, to desolder surplus and decommissioned datacenter HBM and put it on some sort of daughterboard with a translator so it can be used in a consumer machine?
Stuff like that already exists for flash memory; I can harvest eMMC chips from ewaste and solder them to cheaply-available boards to make USB flash drives. But there the protocols are the same, there's no firmware work needed...
I grabbed an upgrade at the end of last year because my ~10 year old workhorse is starting to show signs of aging. Despite 16 gigs of RAM having lasted me thus far I decided to bite the bullet and get 32; so I expect this new machine to last me another 10 years (although I now have a full SSD, whereas my old workhorse had an SSD for the OS and a hybrid drive for /home, so we'll see whether or not it will actually last).
I do not see this from an infinite shortage point; I see this from a locked down hardware point. Old hardware is hackable, new hardware mostly not. That is for me where the real pain is and why I just buy old computers and phones that are rootable.
We are in a renaissance of computing right at this moment. If expand our definition of computers outside of screens and traditional input devices, microcontrollers are capable of so much more, with so much less (energy consumption | ram | storage).
The tipping point for MCUs was WiFi - which not only allows you to speak multiple protocols (UDP/Zigbee/HTTP/etc) and have audio IO, but also P2P communication and novel new form factors. There's been incredible progress with the miniaturisation of sensors and how we're able to understand and perceive our environment.
So yes, whilst traditional hardware is getting more expensive and locked down, there's a strong counter movement towards computing for everyone - and by that I also mean that there's going to be less abstraction in the entire stack. Good times ahead!
> These days, the biggest customers are not gamers, creators, PC builders or even crypto miners anymore. Today, it’s hyperscalers. …
> These buyers don’t care if RAM costs 20% more and neither do they wait for Black Friday deals. Instead, they sign contracts measured in exabytes and billions of dollars.
Does all this not apply to businesses buying computers for their employees?
Hold onto your hardware. Hold on to your existing software and the current version. Don’t upgrade without a specific need. None of the “progress” is actually helpful to hackers and I’m not sure it’s even helpful to typical users. There’s enough information being given to and slurped by others, don’t make it more effective.
My PC has an Intel Xeon from 2007, a GPU from 2010, and 4GB of RAM.
It’s enough for web browsing and can handle 1080p/60fps video just fine.
For gaming, I have a dedicated device - a Nintendo Switch, but I also play indie PC games like Slay the Spire, Forge MTG, some puzzle games e.g. TIS-100.
Linux with i3 is fast and responsive. I write code in the terminal, no fancy debuggers, no million plugins, no Electron mess.
It’s enough for everything I need, and I don’t see a reason to ever upgrade. Unless my hardware starts failing, of course.
In order to go from 360p video 15 years ago to 4K HDR today, I have upgraded from a 2mbps 802.11g WiFi on a 1366x768 display to a 200mbps connection on 802.11ax and a 55 inch 4k television.
The experience is quite immersive and well worth the upgrade that happened very progressively (WiFi 5 1080p then WiFi 6/7 4K).
At the same time, we had cheap consumer gigabit ethernet, and still have cheap consumer gigabit ethernet. 2.5 is getting there price-wise, but switches are still somewhat rare/expensive.
In such a future the iPhone and android ecosystem is dead? Because a single $1k phone is a hell of a computer. So if you can still buy a phone you can still get a computer. Local AI aside these are very capable.
I see they are offering to macos for iphone pro and ipad pro next years with subsc. ? or via upgrade with price I mean it's now possible more than ever
uBlock Origin has prevented the following page from loading:
https://xn--gckvb8fzb.com/hold-on-to-your-hardware/
This happened because of the following filter:
||xn--$document
The filter has been found in: IDN Homograph Attack Protection - Complete Blockage
That whole feature is kind of paragraph 22. No legit/popular site uses it so users don't expect national characters in domain names, so no one actually hosts sites using "xn-" domains.
Capitalism at work. There is more value to be generated by moving resources to data centers for the moment. This isn't some me be insensitive or anything. It's the same people who are buying iPhones and PCs who are demanding more compute for AI.
There could be a swing in the future where people will demand local AI instead and resources could shift back to affordable local AI devices.
Lastly, this thesis implies that we will be supply constrained forever such that prices for personal devices will always be elevated as a percentage of one's income. I don't believe that.
haha, all of a sudden I see a tab "waifu pillow" on Amazon, and think I have a split personality that runs searches in between consciousness shifts, and then I come back to a funny message.
I feel like this is just the bubble talking. I'm pretty naive here, but at some point suppliers will adjust so they can take money from data center builders and consumers, just like pre-bubble.
> Hardware is going to be expensive for awhile but its not as dire as the article makes it out to be.
At the same time, the article’s argument that the value of personal computer ownership is only going to rise, in terms of the value of speech, not strictly in terms of the value of lunch, is important to call out.
I’m glad I held on to my 2009 MacBook, for example, as it still functions today as an active part of my homelab, at an amortized yearly cost of practically the price of taking a nice steak dinner once a year.
It's probably closer to the suppliers don't think this will last and are ramping slowly if at all so they're not left holding the bag.
The US is headed for a cataclysmic crash at this point and it's not clear what will trigger it, but all those companies pushing underpriced tokens and Rust ports of existing tools by agents aren't going to survive it.
It's a thought provoking article and I felt the pain when I shopped around for a new GPU lately to replace a 4090 I thought was faulty (eventually a cleaning of the PCIe connector solved those crashes). I bought it at the end of 2022 and three and a half years it seems like we've gone backwards, not forward on GPUs available for end users. They cost more and do less.
But also consider that PCs have been an anomaly for very long. I don't think there's an equivalent market where you, as a consumer, can buy off-the-shelf cutting-edge technical pieces in your local mall and piece them together into a working device. It's a fun model, for sure, but I'm not sure it's an efficient model. It was just profitable enough to keep the lights on, thanks primarily to a bunch of Taiwanese companies in that space but it wasn't growing anywhere and the state of software is a mess.
Apple the PCs collective lunch before DCs did. So have gaming consoles. So I weep for consumer choice but as things become more advanced maybe PCs and their entire value chain don't make a lot of sense any more.
Obviously at the end there will still be consumer devices, because someone needs to consume all of this AI (at least people are thrown entirely out of the loop, but then all those redundant meat sacks will need entertainment to keep them content). We have the consumer device hyperscaler Apple doing rather OK even with these supply crunches although I'm not sure for how long.
Micron is killing its Crucial consumer brand, not supplies to consumer brands who use its chips. Hynix never had a consumer brand for RAM I don't think?
I refuse, I'll buy when I need to and can hold on for a few months if prices become insane. This means I'll spend less on hardware then what I could, if I wanted to buy max mpro or latest framework I just will not, because prices are too mad and g o for a cheaper version.
whatever happens it's crazy and hope AI madness is worth it
For laptops, I always spring for the lowest amount of ram + hdd/ssd, and then instantly upgrade this from local after-market sources. However, this wouldn't work for apple devices (Hence I don't own any Apple devices).
For example, my current Thinkpad T14-gen5, was bought with 8GB ram and 256GB NVME, and then upgraded to 64GB ram and 2TB NVME, for the same price as 16G/512G would have cost at Lenovo. And then I still have the 8GB/256GB to re-use/re-sell.
AI companies driving RAM prices up is, in my opinion,
theft from the common man (and common woman). Sure,
you can say that in capitalism, those who pay more
benefit the most, but no system, not even the USA,
has a purely driven capitalistic system. You still have
transfer money, public infrastructure and what not. So
private companies driving up the prices, such as for
RAM, is IMO also theft from common people. And that should
not happen. It can only happen when you have lobbyists
disguised as politicians who benefit personally from
helping establish such a system. The same can be said
about any other prive-upwards scaling that is done via
racketeering.
I just realized that this blog site is pretending to be malware. I opened the tab and was constantly switching between the blog and writing this HN comment (I deleted the rest of the comment after realizing it) and was wondering where the tab went and kept opening it over and over again, then I realized that it completely rewrote the tab title with NSFW content (one of the title contained the world "nudes" with a faked amazon favicon) and when you reopen the tab, it shows you a black overlay with a message intended to induce shock if you ever bother to read it (I didn't read past the first sentence so I don't know what it was actually about).
Can dang/a moderator please ban the domain from HN? Even if its not exactly malware, it's pretending to be malware to grab your attention and it's obviously intending to fill your browser history with inappropriate content, which didn't work on my browser because I opened the blog in a private browser session. The operator clearly doesn't run his blog in good faith.
The whole point is to grab your attention and bully you to turn off JavaScript. It links to another page: https://disable-javascript.org/
I opened the tab on my work laptop and having NSFW title and icon in the office is unacceptable, I understand the intent but the implementation and this way of forcing people to do something is ridiculous. I do not own or control this machine, I trust the links of the frontage of HN to be somewhat safe and not put me in an uncomfortable position.
Yes, the site not necessarily malware but a dark pattern and that’s not how you teach your average day-to-day user.
It doesn't write anything extra to the browser history. How about actually checking before exaggerating. If you are bothered by a single wrong title with the right URL, well... I think something else is wrong.
You are also completely speculating on the intent. Less drama please.
Owning hardware is great. But I get the impression that some people view owning petty hardware as some liberty panacea.
You might have a DVD collection, ten external drives, three laptops, and a workstration. You may still for all intents and purposes be wholly dependent on cloud computing, say, because that it is the only practical way to run whatever AI-driven software three years from now.
Edit: That’s an example. It goes beyond AI. and...:
I disagree. There is in fact a non-zero chance that we will get good enough models that are MOE optimized for desktop size hardware that can do a lot of the same things as the SOTA models. Im certainly crossing my fingers that the open-weights models continue improving. Engram from Deepseek for instance seems very interesting for a compute to memory offloading perspective.
I'm not sure why people are upset. This is how Capitalism is supposed to work - resource allocation towards the most productive (in terms of Capital) usage.
Those who are best able to use a resource are willing to pay the most for it thus pricing out unproductive usages of it.
This is pure Capitalism.
If one is in general against Capitalism, yes, one can complain.
But saying "I want free markets" and "I want capitalism", but then complaining when the free markets increase the price of your RAM is utterly deranged.
Some will say "but Altman is hoarding the RAM, he's not using it productively". It's irrelevant, he is willing to pay more than you to hoard that RAM. In his view he's extracting more value from that than you do, so he's willing to pay more. The markets will work. If this is unproductive use of Capital, OpenAI will go bankrupt.
And the RAM sellers make more money, which is good in Capitalism. It would be irresponsible for them to sell to price sensitive customers (retail), when they have buyers (AI companies) willing to pay much more. And if this is a bad decision, because that AI market will vanish and they will have burned the retail market, Capitalism and Free Markets will work again and bankrupt them.
Survival of the fittest. That is Capitalism. And right now AI companies are the fittest by a large margin.
AI and Capitalism are the exact same thing, as famously put. We are in the first stages of turning Earth into Computronium, you either become Compute or you will fade away.
However, I do believe that we're at an inflection point where DC hardware is diverging rapidly from consumer compute.
Most consumers are using laptops and laptops are not keeping pace with where the frontier is in a singular compute node. Laptops are increasingly just clients for someone else's compute that you rent, or buy a time slice with your eyeballs, much like smartphones pretty much always have been.
I personally dropped $20k on a high end desktop - 768G of RAM, 96 cores, 96 GB Blackwell GPU - last October, before RAM prices spiked, based on the logic that hardware had moved on but local compute was basically stagnant, and if I wanted to own my computing hardware, I'd better buy something now that will last a while.
This way, my laptop is just a disposable client for my real workstation, a Tailscale connection away, and I'm free to do whatever I like with it.
I could sell the RAM alone now for the price I paid for it.
People who are willing to drop $20k on a computer might not be affected much tho.
They probably won't, but those willing to drop $3-10k will be if the consumer and data-center computing diverge at the architectural level. It's the classical hollowing out the middle - most of the offerings end up in a race-to-the-bottom chasing volume of price-sensitive customers, the quality options lose economies of scale and disappear, and the high-end becomes increasingly bespoke/pricey, or splits off into a distinct market with an entirely different type of customers (here: DC vs. individuals).
My phone has 16gigs of ram and a terabyte of storage, laptops today are ridiculous compared to anything I studied with.
I'm not arguing mind you, just trying to understand the usecases people are thinking of here.
Running Electron apps and browsing React-based websites, of course.
I wonder if there’s a computer science law about this. This could be my chance!
https://en.wikipedia.org/wiki/Wirth%27s_law
Not exactly the same (it's about power rather than price). But close enough that when you said it, I thought, "oh! there is something like that." There's also more fundamental economics laws at play for supply and demand of a resource / efficiencies at scale / etc. Given our increasing demand of compute compared increasing supply (cheaper more powerful compute). I expect the supply will bottleneck before the demand does.
I thought the trend is the opposite direction, with RTX 5x series converging with server atchitectures (Blackwell-based such as RTX 6000 Pro+). Just less VRAM and fewer tensor cores, artificially.
Where is the divergence happening? Or you don't view RTX 5x as consumer hardware?
768GB of RAM is insane…
Meanwhile, I’ve been going back and forth for over a year about spending $10k on a MacBook Pro with 128GB. I can’t shake the feeling I’d never actually use that much, and that, long term, cloud compute is going to matter more than sinking money into a single, non-upgradable machine anyway.
I don't know your workloads, but for me personally 64 GB is the ceiling buffer on RAM - I can run entire k8s cluster locally with that and the M5 Pro with top cores is same CPU as M5 Max. I don't need the GPU - the local AI story and OSS models are just a toy for my use-cases and I'm always going to shell out for the API/frontier capabilities. I'm even thinking of 48 config because they already have those on 8% discounts/shipped by Amazon and I never hit that even on my workstation with 64 GB.
No, it won't. The power drain of merely refreshing DRAM is negligible, it's no higher than the drain you'd see in S3 standby over the same time period.
It wasn't my primary motivator but it hasn't made me regret my decision.
I hummed and hawed on it for a good few months myself.
How is this going to work? You need uncontrolled compute for developing software. Any country locking up that ability too much will lose to those who don't.
See a $1100 GPU on eBay, but it’s in the US? Actually a $1900 GPU.
A colleague were just talking about how well he timed the purchase of his $700 24GB 3090.
Before this price spike, it used to be you could get a second-hand rack server with 1TB of DDR4 for about $1000-2000. People were massively underestimating the performance of reasonably priced server hardware.
You can still get that, of course, but it costs a lot more. The recycling company I know is now taking the RAM out of every server and selling it separately.
Apple hardware is incredibly overpriced.
Also, I wonder how many of us, even here on HN, have the ability to spend that amount of money on computer for personal use. Frankly I wouldn't even know what to do with all the RAM - should I just ramdisk every program I use and every digital thing I made in the last five years?
Anyhow, I suppose for the folks who can't afford hardware (perhaps by design), one ought to own nothing and be happy.
The RAM choice was because I have never regretted buying more RAM - it's practically always a better trade than a slightly faster CPU - and 96GB DIMMs were at a sweet spot compared to 128GB DIMMs.
That, and the ability to have big LLMs in memory, for some local inference, even if it's slow mixed CPU/GPU inference, or paged on demand. And if not for big LLMs, then to keep models cached for quick swapping.
I don't share the same 1:1 opinion with regards to the article, but it is absolutely clear that RAM prices have gone up enormously. Just compare them. That is fact.
It may be cheaper lateron, but ... when will that happen? Is there a guarantee? Supply crunch can also mean that fewer people can afford something because the prices are now much higher than before. Add to this the oil crisis Trump started and we are now suddenly having to pay more just because a few mafiosi benefit from this. (See Krugman's analysis of the recent stock market flow of money/stocks.)
Open source efforts need to give up on local AI and embrace cloud compute.
We need to stop building toy models to run on RTX and instead try to compete with the hyperscalers. We need open weights models that are big and run on H200s. Those are the class of models that will be able to compete.
When the hyperscalers reach take off, we're done for. If we can stay within ~6months, we might be able to slow them down or even break them.
If there was something 80-90% as good as Opus or Seedance or Nano Banana, more of the ecosystem would switch to open source because it offers control and sovereignty. But we don't have that right now.
If we had really competitive open weights models, universities, research teams, other labs, and other companies would be able to collaboratively contribute to the effort.
Everyone in the open source world is trying to shrink these models to fit on their 3090 instead, though, and that's such a wasted effort. It's short term thinking.
An "OpenRunPod/OpenOpenRouter" + one click deploy of models just as good as Gemini will win over LMStudio and ComfyUI trying to hack a solution on your own Nvidia gaming card.
That's such a tiny segment of the market, and the tools are all horrible to use anyway. It's like we learned nothing from "The Year of Linux on Desktop 1999". Only when we realized the data center was our friend did we frame our open source effort appropriately.
We have this class of models already, Kimi 2.5 and GLM-5 are proper SOTA models. Nemotron might also release a larger-sized model at some time in the future. With the new NVMe-based offload being worked on as of late you can even experiment with these models on your own hardware, but of course there's plenty of cheap third-party inference platforms for these too.
The fact that I didn't know any of this is what is significant here. At some point I stopped caring about this sort of thing. It really doesn't matter any more. Don't get my wrong, I am as nerdy as they come. My first computer was a wire wrapped 8080 based system. That was followed by an also wire wrapped 8086 based system of my own design I used for day to day computing tasks (it ran Forth). If someone like me can get to the point of not caring there is no real reason for anyone else to care.
(A large factor here is, obviously, the cloud. With photos, documents, e-mail, IMs, etc. all hosted for cheap or free on "other people's computers", the total hardware demands on the end-user computing device is much less. Think storage, not just RAM.)
It's true even in tech; half a year ago I switched my phone to a Galaxy Z Fold7, and I haven't used my personal laptop since then, not once. I have a separate company laptop for work, and I occasionally turned on my PC, but it turns out that a foldable phone is good enough to do everything on personal side I'd normally use a laptop for. So here I am, with my primary compute device I don't have full control over - and yes, I'm surprised by this development myself, and haven't fully processed it yet.
There have been memory chip panics before, the US funded RAM production back into the 80s/90s in competition with Japan at the time.
The AI boom/"hyperscale" currently is almost exactly like the dotcom boom.
It's already starting to shake down. Anthropic is occupying the developer space, OpenAI has just exited the video/media production space. More focused and vertical market AI is emerging.
The current vortice of money between OpenAI <-> Microsoft <-> Oracle <-> NVidea <-> Google <-> etc etc is going to break.
Yes but these Chinese firms are a tiny share of the overall RAM/SSD market, and they'll have the same problems with expanding production as everyone else. So it doesn't actually help all that much.
Maybe... just maybe, a TODO list app shouldn't run 4 processes, and consume hundreds of megabytes of RAM?
More seriously and more ironically, at the same time, we've now reached a strange time where even non-programmers can vibe-code better software than they can buy/subscribe to - not because models are that good, or programming isn't hard, but because enshittification that has this industry rotten to the core and unable to deliver useful tools anymore.
Tongue in cheek: we urgently need fusion power plants. For the AI and the helium.
Non-helium hard drives are basically limited by their bearing spin hours. If one only spins a few hours a week, it'll probably run for decades. Not so with helium.
Is that likely? History says it's inevitable, but timeframe is an open question.
If this does occur, unfortunately it isn’t like any of the production capacity is going to immediately shift or be repurposed. A lot of the hardware isn’t usable outside of datacenter deployments. I would guess a more realistic recalibration is 2-3 years of immense pain followed by gradual availability of components again.
The capital from the gulf is already disrupted. It isn't anymore a matter of if or when.
Stuff like that already exists for flash memory; I can harvest eMMC chips from ewaste and solder them to cheaply-available boards to make USB flash drives. But there the protocols are the same, there's no firmware work needed...
The tipping point for MCUs was WiFi - which not only allows you to speak multiple protocols (UDP/Zigbee/HTTP/etc) and have audio IO, but also P2P communication and novel new form factors. There's been incredible progress with the miniaturisation of sensors and how we're able to understand and perceive our environment.
So yes, whilst traditional hardware is getting more expensive and locked down, there's a strong counter movement towards computing for everyone - and by that I also mean that there's going to be less abstraction in the entire stack. Good times ahead!
Does all this not apply to businesses buying computers for their employees?
For gaming, I have a dedicated device - a Nintendo Switch, but I also play indie PC games like Slay the Spire, Forge MTG, some puzzle games e.g. TIS-100.
Linux with i3 is fast and responsive. I write code in the terminal, no fancy debuggers, no million plugins, no Electron mess.
It’s enough for everything I need, and I don’t see a reason to ever upgrade. Unless my hardware starts failing, of course.
The experience is quite immersive and well worth the upgrade that happened very progressively (WiFi 5 1080p then WiFi 6/7 4K).
That said.... hopefully at least on Android side you can get a free (as in unchastified) OS to run on it.
Until they come for the HW.
There could be a swing in the future where people will demand local AI instead and resources could shift back to affordable local AI devices.
Lastly, this thesis implies that we will be supply constrained forever such that prices for personal devices will always be elevated as a percentage of one's income. I don't believe that.
Consumer hardware will always be a market worth serving for companies who don't see their stock price as their product.
If the existing companies are unwilling to make a sale, I am sure new players will arise picking up their slack.
https://www.youtube.com/watch?v=SrX0jPAdSxU
They will let the hyperscalers buy their supply at a premium and wait for the bust. Then they will shift back to the consumer space.
Hardware is going to be expensive for awhile but its not as dire as the article makes it out to be.
At the same time, the article’s argument that the value of personal computer ownership is only going to rise, in terms of the value of speech, not strictly in terms of the value of lunch, is important to call out.
I’m glad I held on to my 2009 MacBook, for example, as it still functions today as an active part of my homelab, at an amortized yearly cost of practically the price of taking a nice steak dinner once a year.
The US is headed for a cataclysmic crash at this point and it's not clear what will trigger it, but all those companies pushing underpriced tokens and Rust ports of existing tools by agents aren't going to survive it.
But also consider that PCs have been an anomaly for very long. I don't think there's an equivalent market where you, as a consumer, can buy off-the-shelf cutting-edge technical pieces in your local mall and piece them together into a working device. It's a fun model, for sure, but I'm not sure it's an efficient model. It was just profitable enough to keep the lights on, thanks primarily to a bunch of Taiwanese companies in that space but it wasn't growing anywhere and the state of software is a mess.
Apple the PCs collective lunch before DCs did. So have gaming consoles. So I weep for consumer choice but as things become more advanced maybe PCs and their entire value chain don't make a lot of sense any more.
Obviously at the end there will still be consumer devices, because someone needs to consume all of this AI (at least people are thrown entirely out of the loop, but then all those redundant meat sacks will need entertainment to keep them content). We have the consumer device hyperscaler Apple doing rather OK even with these supply crunches although I'm not sure for how long.
whatever happens it's crazy and hope AI madness is worth it
For example, my current Thinkpad T14-gen5, was bought with 8GB ram and 256GB NVME, and then upgraded to 64GB ram and 2TB NVME, for the same price as 16G/512G would have cost at Lenovo. And then I still have the 8GB/256GB to re-use/re-sell.
Can dang/a moderator please ban the domain from HN? Even if its not exactly malware, it's pretending to be malware to grab your attention and it's obviously intending to fill your browser history with inappropriate content, which didn't work on my browser because I opened the blog in a private browser session. The operator clearly doesn't run his blog in good faith.
I opened the tab on my work laptop and having NSFW title and icon in the office is unacceptable, I understand the intent but the implementation and this way of forcing people to do something is ridiculous. I do not own or control this machine, I trust the links of the frontage of HN to be somewhat safe and not put me in an uncomfortable position. Yes, the site not necessarily malware but a dark pattern and that’s not how you teach your average day-to-day user.
I thought it was clever. But it also seems ham-fisted, and in poor taste.
You are also completely speculating on the intent. Less drama please.
You might have a DVD collection, ten external drives, three laptops, and a workstration. You may still for all intents and purposes be wholly dependent on cloud computing, say, because that it is the only practical way to run whatever AI-driven software three years from now.
Edit: That’s an example. It goes beyond AI. and...:
Liberty goes beyond that.
https://www.reddit.com/r/LocalLLaMA/comments/1s0czc4/round_2...
Those who are best able to use a resource are willing to pay the most for it thus pricing out unproductive usages of it.
This is pure Capitalism.
If one is in general against Capitalism, yes, one can complain.
But saying "I want free markets" and "I want capitalism", but then complaining when the free markets increase the price of your RAM is utterly deranged.
Some will say "but Altman is hoarding the RAM, he's not using it productively". It's irrelevant, he is willing to pay more than you to hoard that RAM. In his view he's extracting more value from that than you do, so he's willing to pay more. The markets will work. If this is unproductive use of Capital, OpenAI will go bankrupt.
And the RAM sellers make more money, which is good in Capitalism. It would be irresponsible for them to sell to price sensitive customers (retail), when they have buyers (AI companies) willing to pay much more. And if this is a bad decision, because that AI market will vanish and they will have burned the retail market, Capitalism and Free Markets will work again and bankrupt them.
Survival of the fittest. That is Capitalism. And right now AI companies are the fittest by a large margin.
AI and Capitalism are the exact same thing, as famously put. We are in the first stages of turning Earth into Computronium, you either become Compute or you will fade away.