So the AI boom is resulting in a) fewer jobs, b) massive increases in hardware, and c) exponential acceleration in wealth inequality (world's first trillionaire anyone?).
With some work on crypto many people could be a trillionaire on paper. Whether it translate to actual wealth and liquidity is another matter. We are talking about P/E of 300 / P/S of close to 100 for Tesla and SpaceX.
>When exactly are the upsides going to hit?
A lot of people take Moore's law, or technology improvement as granted. It will always come. It will always become cheaper. But none of that is true. Massive R&D is required along with ROI.
The AI Boom pushed a lot of technology forward by at least 2 - 3 years or 1 cycle. What normally would have taken 10 years to happen is now getting close to 5 years. We were suppose to stagnate or slow down with 3nm and 2nm, we are now rushing to push through everything from interconnect, smaller transistor and massive increase in Foundry capacity. PCI-Express 8.0, Nvidia Photonics, DRAM Improvement, HBM, HBF, even capacitor, immersive cooling. I don't even record the last time we had such a massive shift and changes in hardware technology. Even the start of smartphone era wasn't like this as majority of its start was picking on lower end PC components. Instead the AI is pushing the frontier hardware technology. With multiple trillion companies, insane appetite from market. We are basically saying we have Trillions to spend over the next 5 years. Give me everything you have got.
Upsides will be seen into societies that are not capital-above-govt, but govt-above-capital. China for instance: they will show advantages of AI (amongst other technologies). Sure they've got surveillance there, but there's also surveillance in the west. In China you have clean streets and low crime, while in the west it's surveillance without tangible benefit for the common people.
Because the alternative is admitting the main upside so far is: a few VCs and early employees get yacht money while everyone else gets a gate-kept chatbot and constant fear mongering. But hey, I guess we all have our own opinion of "upside".
Isn’t this the same thing every other industry and market has been experiencing with technology, automation, etc. for decades, while tech workers basked in their joy at being safe from being replaced because we were the ones powering the replacing?
I mean, that very well could be the case. It often is the case with any kind of automation. There's a plausible claim that things will turn out fine, and another plausible claim that it could be a disaster for the profession. But whether or not it's a disaster for professional developers is separable from whether (a) it's a disaster for, like, society, and (b) whether or not it fundamentally is an important new fact about the world --- like, whether we're OK with it or not, it does appear likely that a huge swathe of professional software development work is fully automatable.
What's your source? The Bureau of labor statistics that is marking most of unemployed people as "not seeking employment" because they aren't walking into workforce centers?
Payrolls are up about 214k in March, 180k in April, and 140k in May. The BLS data is from broad monthly surveys, not people showing up in workforce centers.
Personally, I can do more than I could before as the result of AI.
Honestly I could "retire" to a senior level role and have AI do 90% of my work and nobody would know the difference.
The benefits COULD hit by employers reducing everybody's hours.
The blocker to this is the middle management disease where there's a class of people who spend 40+ hours a week in some kind of update meeting or another and that much talking can't be replaced by AI. (much of it could be replaced by just not doing it any more but that's a different story)
You can't shift productivity by 10X and expect the rest of the supply/demand equilibrium to stay the same, with you working 10% of the time and sipping drinks on the beach while retaining the same job opportunities and expected salaries.
There will be increased competition for job openings, reductions in real wages, or increased expectations of productivity. Probably some combination of all three.
No, but you could cut everybody to 30 hour workweeks and hire more people.
Once it becomes the norm even for a small section of the economy it will spread.
People are more productive in an absolute sense working fewer hours anyway.
It just takes a union, an ambitious company, or a state to force that 30 hour workweek to show some success with better talent attraction and retention and better corporate results to start a trend.
It is possible for everybody to get a piece of the pie.
Yes, in theory but if working more confers an advantage then work expectations will stay the same and will show up as higher productivity. So the real question is how will that productivity be distributed. Right now it's being concentrated into shareholders value rather than wages.
Strongly agree. Reduce the work week today to 4 days 32 hours, the US already generates $5T in profits per year. That is time taken from workers from the one life they get. If corporations want to be more productive, take your best shot with LLMs. If it works, great, we keep reducing the work week. If it doesn't work, well, take it up with who sold you the magic beans.
We are already productive enough to have a shorter work week and more leisure, anyone saying no has specific incentives to not support it (either via financial gain from the capital accumulation funnel or work bound to their identity).
So tax them as much as needed, or require state ownership of some amount. Startups, and company entities in general, exist because a jurisdiction allows them to, or allows them to accept payments, control accounts with financial resources, pay vendors and workers, and operate within a commercial framework. That is a privilege, not a right. The rules are a shared agreement and delusion, the rules can change at any time.
It would be one thing if those increases in productivity were improving my QOL as a consumer of produced goods, through cost reductions or increases in quality, but I'm not seeing any of that.
All that seems to be happening is that these productivity gains roll up as profits for the owner class.
It's easier to enter the owner class. It's easier for you to do a startup where you're not the expert in everything. It's easier for you to rely less on buying things when you can make them yourself.
The moat of the owner class is lower because now information is everywhere and it's less possible to hide behind trade secrets and implementation effort.
Based on what? The macroeconomics don't work out that way. IF productivity goes up, but consumption does not, that means that it's harder to enter the owner class, because fewer productive enterprises (owned by non-working people) are supplying a larger share of customer demand.
This may make a difference on the margins for people in the software bubble. But for the other 8 billion people on the planet, they aren't all going to become owners in your brave new world, unless consumer demand goes to the moon to soak up all that productivity. It's not doing that. Prices aren't dropping. Quality isn't increasing.
If you think I'm wrong - is there a cross-economy explosion of small one-person businesses that I'm somehow not seeing? Are gigacorps across the board all losing market share? Because on the macro scale I see nothing but further consolidation.
Which is only relevant if one of the following two things happens:
1. Consumption goes up. (It's not going up. 40 hours at your job buys you less shit today than it did 3 years ago.)
2. Mega-corps start losing marketshare and revenue to this avalanche of new one-to-two-person businesses. (They aren't. Their revenues are climbing, which implies that consolidation is what's happening, not diversification.)
Yup. Everything after the first decade of a career is either jockeying for a promotion, jumping ship-to-ship as needed, or sitting around in meetings.
It's not that people get lazy or are underemployed, but that expertise is just that valuable. You'd think this would be proof enough to break the AI echo chamber already.
Doesn't that mean your org could find someone 10% as productive as you and fire you? We seem to ignore that side of things.
Concretely: if you can do 90% of your work with AI, someone else can also do that same work, making you interchangeable unless that 10% is really important.
I think this is partly why it's so hard for people to find jobs right now. Everyone is interchangeable thanks to AI, so skill gives you less of an edge than it did in the past.
> Concretely: if you can do 90% of your work with AI, someone else can also do that same work
That's the part that is not true. Prompting and guard-rails and generally harness engineering do matter a lot lately. Seen it first-hand multiple times, especially after I used Fable 5 for a week.
How long does it take to learn all that then? Can't imagine it would take that long.. At least compared to, eg, becoming an experienced Rust or C programmer.
And if it does take that long, why is it so great anyway?
Making labor hyper-interchangeable is kinda like the whole pitch here. It's two steps away from b2b SaaS labor if the PR is to be believed.
Maybe you can say you're an elite prompter or whatever, but it always kinda sounds to me like "I know the secret menu at taco bell." Like the whole point of the product here is precisely to not need such pretense or complexity. You are paying hundreds (at least) a month to use something, but also you are using it in a special way? I really don't get it.
Judging by the students I have seen, it will take longer. You could become a mediocre programmer in months, and still be useful by doing grunt work. I don’t see how people will become master architects steering fleets of agents without a deep understanding of the fundamentals, which takes longer to master.
Yes, you can vibe up a demo in no time. But LLMs still need guidance to produce an architecture that will hold up to real world scenarios.
Sure, a good senior director or technical CTO could phone it in and do my job, but there aren't as many of those.
Jobs were hard to find in the drawback after the COVID hiring boom in uncertain times as the result of Trump, inflation, tariffs, war, and the constantly impending but pushed off market crash we've been expecting since before COVID started. I'm not saying AI isn't contributing, but it's hardly the only factor.
AI is far cheaper to fire than a person.
"Everyone is interchangable" isn't quite right, a tremendous amount of people don't actually add all that much value and a lot of work is just running on a hamster wheel and now instead of taking time we've got a machine for running on hamster wheels for us.
The delusion of engineers is that finally we have the means to stop compromising on quality to deliver, and fix all the garbage we’ve had to put out rationalizing about “cost-benefit”.
Trouble is management, who has the signature on the bank account which ultimately controls if and when the bailiffs are going to knock down your front door and throw you onto the street.
Management thinks we can now continue putting more, much more, of the same compromised garbage. Cheaper, faster.
> I could "retire" to a senior level role and have AI do 90% of my work and nobody would know the difference.
They would notice, and then they would fire 9/10 of the people in your role. If you are unlucky, you get laid off. If you are lucky, you get to botsit full time for the workload of 10 engineers for less pay and no career advancement.
This would last until they figure out how to remove the human from the loop entirely.
Nobody has noticed where I work. I'm thinking of getting a second job, actually.
Key factors for me:
- Company is full of old school engineers who seem to hate AI and will scrutinize every command it runs. Means that even though we're both 'using' AI, I'm still way more productive.
- Said engineers have too much inside knowledge of the horrific system they made that management can't possibly get rid of them. Helps that they're workers-rights minded too.
- Company has enough revenue to keep up payroll indefinitely.
That last part is probably the biggest risk, but we're in kind of a niche industry. Not really a big, juicy target.
Now, does the AI write good code? Often not. But the codebase is already terrible, so it's no big difference.
I’m very on the pro-ai side (check my comment history for proof), but this “ai will give us more free time” logic is seeming more and more patronizing (to be clear, I understand that you are being sarcastic haha).
I was listening to a podcast a couple days ago and Brad Gerstner was on and mentioned that with how AI is boosting productivity that perhaps one member of a household would be able to start staying home from work if they wanted. I shut off the podcast after that (to be fair, the podcast just seemed to be one massive SpaceX IPO pump).
It’s just so divorced from reality and every new advancement is just making *higher expectations for doing more work*.
The unfortunate reality is:
Companies that are selling ai will sell that ai will make life easier.
Companies that are buying ai will demand more from employees using ai (why else would they buy it?).
Yep. With ai tooling I can keep prompting at 3am while sleep deprived. As a result I have mountains of slop plans / code to review. Hours of work which can't be matched by anyone who thinks they can poke a prompt for the day & go
Just as happened with the horse, with the car, with the steam machine, with the industrialization in general, ... oh wait, we still have to work 8-10 hours 5 days a week, times two, to make enough for a living.
So when exactly is this productivity going to hit that doubles my income?
I guess the argument would go that your income is significantly higher in the sense that the quantity and complexity of stuff that you can afford now is vastly greater than 100 years ago (e.g. washing machines, cars, clothes, computers). I’m not that saying it’s making anyone happier, mind you
Don't forget most people are stuck renting a small apartment at a significant percentage of income for eternity. Then if you hit the layoff jackpot and become homeless, then I've got good news for you: homelessness is illegal now.
It's not forced on you. If you do a minimal amount of research (which LLMs are very helpful with!), you can still find durable stuff. A Speed Queen washer is still built like a tank. It's just that the less durable stuff is absurdly cheap now. /r/BuyItForLife/ is a decent place to hang out if you care.
The typical middle class family 50 years ago lived in a house you’d consider small and dingy, ate food you’d consider poverty meals, and drove a car you’d consider a poorly assembled death trap. Ask your parents or grandparents how often they got to have real butter growing up.
I guess it's "only" 40 years, but my 1985 Civic was an amazing car. Definitely not a death trap, but after I did end selling it for a 99 Acura with airbags. Still kind of regret that one. My house was built in 1970, it's enough for the two of use, but would admittedly be cramped for more. That said at 850 square feet, it's quite a bit smaller than the 1,400 average for 1970.
This is a truth very few people are willing to confront. My grandmother lived in a village, on a farm, growing her own food and slaughtering her own animals, with no working plumbing, using a well for water. Of course a lot of that changed even just moving up to the 70s, but at that point there still wasn't quite the consumerist "buy whatever you want from wherever and whomever you want and have it almost immediately" environment. I can go to a grocery store here in Canada, buy tropical fruits year round that grow nowhere near me. I never have to concern myself with "this ingredient won't be here because it is seasonal", it'll be there, it'll just be more expensive out of season, worst case I just have to go a bit farther out to a different grocer than I usually go to.
This point is valid. However, lifestyle improvement rate is something that's slowing over time because of physical constraints.
For example, the vehicle mortality rate is 1.44 per 100 million miles driven. That's down 17% from 2000 (so 25 years ago). However, the change from 1975 to 2000 was 53%. That's because as we get closer to 0, it gets harder and harder to improve those rates. On this metric at least, I don't think another 25 years will result in a noticeable amount of improvement?
In the other direction, some things will become scarcer (and therefore cost more). Real estate is the obvious one; we can't create more land, and we keep having more people. Easily accessible drinking water is another; desalination is getting cheaper, but it's still way more expensive than pumping aquifer water.
And some improvements are necessarily 1 time things. You can get tropical fruits year round, but that's been widely available since the 80-90's from what I can tell. So come 20 years from now, what will people be able to buy in a grocery store that I can't buy right now?
Being able to eat pork without cooking it to death for fear of trichinosis is a recent development.
Also, the old movies where someone tries to commit suicide by sticking their head in an oven. That was coal gas and we don't heat homes with it anymore.
That's total bullshit. Middle class families in 1976 did NOT live in smaller houses than today, and certainly did NOT eat "poverty meals"... What on earth are you even talking about.
Especially silly that you mention housing because if there's one thing that is absolutely fucked for the middle class of the 2020s is housing.
Again, I would encourage you to research historical statistics or talk to people who lived in 1976 about their practical living conditions, rather than going off of your intuition about what is "bullshit" or which things are "absolutely fucked". Our intuitions about these things are heavily warped by social media, where stories that feel true without being true are easy to tell and often more viral.
In every developed country whose numbers I've seen, the size of the average living space is up 30-50% since 50 years ago.
Capitalism, by nature, will want to force everything to the limit.
It's the same thing that happens to housing. People complain housing gets expensive because of "tech workers". No. The reality is greedy landlords can charge 20% more so they charge 20% more. They could be happy with what they make now, but no, they'd rather have the extra 20%. And so housing prices go up 20%.
The thing is, it's not just landlords that are greedy. Everyone is greedy. Companies are greedy. Yeah, you can get the same amount of work done in 1/5 the hours per day. But why not do 5x the work instead?
Well, it's a mania, it's stupid now and it will correct itself. There will be pain. There's always pain after stupidity. Then we will see if LLMs are as useful as we were told.
The only correction will be anyone not in the concentrated wealth part being left out to rot. There is no upside for pretty much anyone posting on hacker news
Never. At this point I think the only way out is a Sea Peoples[1] level of collapse. Maybe they'll call it the Late Chip Age collapse. People will not put with with being obsoleted. Americans at least have the means to resist. The rest of us will probably need 3D printers.
Your complaints could have been made in the 80s as well. When the personal computer was introduced it removed tons of jobs, Bill Gates became the first 100 billionaire.
But we ended up with lots more jobs than the ones that disappeared. The industry has kept millions of people employed for almost 50 years.
The billionaire founders are usually "worth" far less than the economic activities in the companies they own. Would the same economic activities have happened in a system where billionaires can't appear as a result of the same activity?
I don't see how, and it certainly hasn't happened yet.
Anwyay! Talking to my computer and have it do the things I tell it, like I'm in freggin' Star Trek, feels like a pretty huge upside already! :)
This is the real risk after the slow death of personal computing. Even internet resources like servers will be hoarded by the hyperscalers that are the only ones who can afford to order years of compute hardware in advance.
I’ve already started buying cheap old business PCs just in case I’ll ever need to have simple barebones machines to run things on.
> This is the real risk after the slow death of personal computing.
By the time you can have a slow death of personal computing, capacity will improve and prices will improve.
In the shorter term sitting on an old computer or regressing a couple years on specs or paying an extra $100/$200 for 8GB/16GB works.
> Even internet resources like servers will be hoarded by the hyperscalers that are the only ones who can afford to order years of compute hardware in advance.
I don't see why hyperscalers would be so much better at handling price increases.
For some average business paying a week's wages for the computer you use, they can afford that doubling to two weeks just fine.
For all the normal server rental companies, okay the guy on the $10 plan either pays $16 now or cuts their resource allocation and keeps paying $10. That's not going to cause a sea change. And higher end hosting isn't that much different.
Hyperscalers buy so many CPUs and so much RAM they can dictate prices (to a point) or at least make an agreement to buy at a much lower price but buy X amount for the next 5 years.
When the market's so hot I don't see why anyone would give them a particularly big discount. And they would have been getting a discount before, so they probably end up seeing a larger percentage spike in what they have to pay.
Maybe if they're locking in long enough to fund new fab construction? But in that case after a few years a ton of capacity will come online so they're actually helping solve the problem.
I work on a desktop, but as a backup I have bought a refurbished Dell Latitude, there are a lot of decent ones for €250 on eBay. Put Linux on it, it’s good enough for most workloads.
I just hope my top shelf 2020-era desktop doesn’t die on me because it would get very expensive to get a new build these days.
> I just hope my top shelf 2020-era desktop doesn’t die on me because it would get very expensive to get a new build these days.
I could probably sell my gaming rig (12900K, 64GB of DDR5, 4TB NVME, RTX 3090) for more today than what I built it for about 4 years ago, it's absurd. I won't, of course, because it's still glorious for 4K gaming even today. In retrospect, $5000 very well spent.
I have a similar rig, but a 10th gen i7, 128GB of DDR4, 2TB NVMe, and dual RTX 3090s. Built it in 2021 for crypto mining, it ended up paying itself off just before Eth went PoW. I kept it mining even after profitability, because it ran warm enough to heat my apartment leaving my HVAC on fan-only to circulate the heat around.
After winter, I started playing with various other GPU loads until LLMs and SD became easy enough to use. Now it's my experimentation machine.
It's already paid for itself, so anything I sell it for would be profit, but it is still super nice for running local LLMs that power various projects "for free".
I did exactly this and suggest it. Used Dell Latitude (the one I got is one from 2019 - model 5300). I put Linux Mint Debian Edition (https://linuxmint.com/download_lmde.php) on it. Works absolutely great.
Small Dell Optiplexes are good for desktop computers.
> I’ve already started buying cheap old business PCs just in case I’ll ever need to have simple barebones machines to run things on.
Hmmmm...
That's a bit dramatic. I did buy a few SSDs when RAM prices started going up for I didn't want to be facing a shortage of SSDs but China is already very busy at producing cheap server motherboards: gone are the days where he only option for a server motherboard was a $1000 one. There are ultra cheap, and fully functional, chinese motherboards now.
I'm totally convinced China is not going to sit doing nothing: if RAM prices goes up and there's a business opportunity, China is going to seize it and start producing lots of RAM.
We did see "RAM price quadruple in no time so nobody had time to adapt", that's for sure.
But I'm really not sure it's "RAM prices goes 4x and then the world stays as it is and nobody adapts for decades to come".
Also as to used servers: people ordering years of compute hardware in advance aren't hoarding five years old servers.
Heck, a ten years old Xeon machine is plenty capable and the usual people are buying stocks of those (from companies updating their fleet), refurbishing them and reselling them one by one on the usual marketplace, at the same prices (OK maybe instead of 150 EUR it's now 200 EUR if there's 64 GB of ECC RAM).
My usual seller is doing its business as usual although we're well into the memory craze.
Just to put things into perspective: the "PC sales crash" from 2021 to 2024 saw, what, a 20% drop, 10% of which already recovered. In 4 years more than a billion PCs were produced.
A PC is not a rare thing. We won't run out of PCs, just like we won't run out of cheap used ten years old Xeon servers.
But OK you got me: I may contact my seller and hoard two or three more just in case.
Sheesh is HN full of paranoid people and, darn, is it contagious.
If the hyperscalers demand is persistent, more hardware players will notice and will want a piece of the pie. You already see non traditional players entering.
But what you see is a cautious strategy from the existing players. They are hedging against a bubble. They don’t want to pour today tens of billions of dollars in capacity that they will have to sell it to a deflated market
This is a market shock. The reason it doesn't alleviate sooner is the chip manufacturers are very wary of investing billions on capacity that won't come online until after the shock is over and subsequently going bankrupt. Because that's happened before.
IF IT LASTS, capacity will increase.
But it won't last. The AI boom is in exponential growth but it's based on heavy speculation about future value and the bubble will absolutely pop, how agressively depends on how dumb people are about now. The current growth may or may not be entirely justified but it's not sustainable, the free investor money does run out. These back and forth self-dealing deals where companies that own big pieces of each other announce "partnerships" where companies are selling resources essentially to themselves and counting the revenue several times... those are a sign of the approaching peak.
I can't wait for a scrappy startup to spend a couple billion dollars and several years setting up a new fab for chips so we can see the prices go down!
That's the rub. There is no such thing as "anyone that can supply hardware cheaply"
Either it's an established vendor with designs and fabs or it's a newcomer that needs to invest a massive pile of cash in designs and fabs. Neither are cheap.
It is supply and demand but chip supply isn't very elastic and the producers are conservative about adding capacity that won't be needed by the time it's online.
I mean, I want to double the the number of billions of dollars I have by 2031, doesn't mean I will. (actually I have 0 billion and double 0 billion is still 0 billion, so unlike them I'll accomplish that goal).
No. It doesn't have to be a bubble at all. It just has to be a cycle of rapid capital equipment build-out that returns to more normal levels in a few years.
Hard to believe that an industry making money hand over fist are reluctant to spend tens of billions expanding capacity that won't come online for years?
There have been SEVERAL crashes that have wiped out the market and it's the reason there are so few players, the rest of them went bankrupt after periods of over-expansion. (in the 80s caused by Japan, in 1997, in 2001ish after the dotcom bust)
You're even calling it a bubble so it's not exactly "hard to believe" it will pop.
Idealistically, actual consumer driven capitalism would be much closer to supply/ & demand than this (imo) almost completely artificial, government sponsored bubble. I think pricing before this current bubble reflects that.
Because, oddly enough, there are a lot more things going on. Government incentives, companies trying to stranglehold markets, FOMO-fueled massive investiment into the current fad, etc. Just to name a few.
I was also told this. Why isn't it happening? Can some capitalists explain why capitalism isn't happening the way it's supposed to? Is it because of government regulations, do we need to deregulate?
“Is it because of government regulations, do we need to deregulate?”
Insufficient law enforcement. The same memory manufacturers already broke antimonopoly laws in the past, pleaded guilty. Apparently the fines were too small for these companies to care, and the people responsible were promoted instead of being punished. More information: https://en.wikipedia.org/wiki/DRAM_price_fixing_scandal
In this context, it takes spending enormous piles of money over the course of at least several years to spin up new semiconductor production.
We do need more capacity to keep up with AI datacenters' usage, yes.
But adding long-term capacity years down the road for a thing that some folks seem to confidently think is a bubble that can pop at any time is risky. And (because capitalism), we have to manage carefully balance our risks and rewards in order to maximize our odds of success.
If there is no bubble and demand stays high long-term, then the payoff for that risk is potentially enormous.
If there is a bubble and it bursts, then the cost of that risk is potentially devastating.
(Capitalism works most-predictably when cheating is possible, such as with Biff's use of the time machine in Back to the Future II. But without cheats, it's always a gamble.)
Sadly this is not a joke nor some doomsday thinking. Google recently repurposed thousands of phones for server needs. Why? Well guess why… to teach everyone tis coming.
To clarify: UC San Diego is planning to build a cluster of 2000 Pixel phones for computer science class use and Google, in support, helped with a test with 20 phones.
To show they’re working on reducing the impact of data centres on the environment, and that they’re taking action on e-waste, all while saying their pixel phones are so powerful they can be clustered into servers.
And their announced test with 2000 phones, where one server is 25-50 phones, is only 40-80 servers. Interesting, but hardly hyper scale.
It’s not just the hyperscalers, it’s also the boom in personal projects being launched. AI has massively decreased the bar to entry, and a lot more people need servers compared to before
True but this is probably because now they have much more demand as other competitors got to expensive and now people are going for the smaller ones even with low service levels
Right, instead of using a managed serverless platform where you pay a premium for proprietary cloud provided services to take away the overhead of managing and patching servers for you and let you just deploy containers, you can… pay a premium for proprietary cloud-hosted AI engines to take away the overhead of managing and patching servers for you and let you just deploy containers.
I don't think any VPS provider has come close to raising their prices this much. There's absolutely not enough people flocking to Hetzner to justify this.
That's all true, but this seems disproportionate even to the hardware market. They already raised their prices recently as hardware got more expensive, and now they are quadrupled. Maybe I am just misreading the hardware market right now.
If you see photos/videos of Hetzner datacenters, their servers are essentially plain, low-end motherboards (1G network, few slots) sitting on shelves (don't mean that as derogatory, it's an efficient design). What it does mean is that their per-server costs are absolutely dominated by the very components that are exploding in cost right now: RAM, SSDs and (to a lesser extent) CPUs.
OVH still seems cheaper here for smallest VPS (what's needed to host a personal website), especially when you consider that you have to pay your IPv4 1.7 € with Hetzner.
For the exact same specs (CX23 vs VPS-1 2027), price is 6 vs 4.5 € and you can get a 15% discount on OVH if you order a full year.
A ton of cloud workloads are still running on old Haswell-era CPUs with RAM that was bought a decade+ ago. Probably the costs will be made up with new VM shapes.
I doubt this has to do with the hardware discussion. This is just them increasing their lock-in and trying to curb businesses running to other CDNs (whole point of the peering).
Inching closer to Vultr prices. There are some Rails projects I might have later this year, and I had already been thinking of putting them onto Vultr via Hatchbox since Vultr offered a managed db. Maybe for some stuff that I can run a Rails 8 Solid Stack app with just sqlite, I'd use Hetzner. I tested both with Hatchbox but have nothing in production on either yet and generally use Heroku and Render still.
Has anyone here used Vultr much? I'm curious how they felt about bang for buck. At least with Hatchbox it's easy to run multiple domains on one box.
We use Vultr, DigitalOcean and Hetzner for global coverage. Vultr is by far the worst - some DC like Australia are pretty bad, lots of connectivity issues, some are OK. Their forte is that they offer a lot of DCs. We are migrating some workloads back to DO, where things are usually way smoother. Hetzner is our core, but does not offer DCs in Asia, Africa or Latam.
Oh, didn't realize that. Any recommendation for mitigating that if deployed there? Had thought about putting something on Vultr in a few months. Also open to any other good recommendations for providers that have a managed Postgres.
I use it as a hosting target for automated deployment tooling we wrote. Tools were originally Digitalocean-only and I wondered if LLMs could add support for a second provider, so asked Claude to add Vultr which it did very nicely. But other than run the automated tests (which create VMs, DNS entries, check validity and tear down) and pay the monthly bill, we haven't yet used them for production deployments.
The hyperscalers are the ones buying up all the memory and storage hardware in the first place. They aren't affected by the problem, they are the source of the problem.
Just throwing it out there I've got an AX162-R (48c EPYC) with 128GB additional memory (384GB DDR5 total) that's been running a side project of mine for a year or so... I could be talked into moving that project if someone had a need for the hardware.
They might have had delivery contracts from before the prices increased, so they didn't have to pass them to customers. Maybe the last servers from those contracts got delivered already and any new orders need to be bought at much higher prices.
Another possibility: They were growing too fast and need to slow down. At some point additional growth might become too risky, or even exponentially more expensive. It might require fundamental organizational changes.
I’m not a business person, but they’re already at the “hundreds of thousands of servers” scale, what about the 41st data center be organizationally far more expensive than the first 40?
Simple example: Pizza delivery service. The company runs very well, customers are happy, demand increases. At some point the demand gets so high, that they need to buy a second car for deliveries and a second pizza oven.
They look at the numbers and see the risk of making less profit than before, if they expand. Especially if demand decreases at some point, instead of growing further. So they decide to just raise the prices, lower demand and make even more money without additional risk.
Hetzner and OVH and other bare metal but low cost providers use commodity hardware. When that commodity hardware increases there is simply no other option. The secret to the success of these providers is using common off-the-shelf hardware instead of specialized server hardware, which is now being cannibalized.
doesn't this imply that they buy massive hardware and thus replace their hardware constantly? At what rate? It seems the rate is massive. I'd gladly use a 4year old server for the old price.
The AI bubble has increased the prices of nand and ram by a factor of 4, so a 3x increase seems reasonable. Companies that are not big enough to have long term contracts with ram/nand vendors have been hit really, really hard by this.
These prices have absolute nothing to do anymore with memory prices. Do not forget that Hetzner already increased the setup fees by a factor of 4x before to compensate for the price. And also servers getting price increases.
It seems they have shifted by reducing the setup fees, and increasing the monthly costs. As this generates more revenue. And its easy to prove this...
AX42 ... Its 8700GE that has gone from 65 Euro to 225 Euro. With the setup fee now being 112 Euro instead of 225 Euro. It has 64GB memory, and 1TB storage. The storage even in todays market is 100 Euro. The memory is 644 Euro.
Do the math ... Hetzner servers had a hardware payback periode of between 9 to 11 month if you took the market value. This calculation has always been very stable over the 20 years i used Hetzner.
This new price, reduced the hardware payback periode to ~4 month. It seems to be that Hetzer is trying to use the memory price issues, as a excuse. The revenue of those same servers now increased to a insane level. More revenue with less hardware.
The real issue is that a lot of companies are moving from US hosting to EU hosting because of the problems with the US. Hetzner sees this as the perfect time to cash in on Enterprise customers.
They have been trying to replace the "cheap" normal consumers with enterprise. This trend has been going on for a while already.
Every customer that now leaves, is a server they can rent out to business customers.
If you want to see the same thing, look up what happened to Microsoft/Github Copilot where they turn around has been sudden and very strong, with a clear goal of moving everything to enterprise.
The big increases here are for their cloud product, which is hourly billing with no setup. In that context it seems more reasonable. I guess we need to remember that hourly billing and flexible prices cut both ways, eh?
Also, a price increase like this can be used to address over-subscription/under-utilization .. there will be a lot of dormant chaff blown off by this, or in other words the provisioning demand will also be adjusted by this aggressive price change, imho.
> It seems they have shifted by reducing the setup fees, and increasing the monthly costs.
Monthly costs have gone up as well. Payroll has seen significant increases in Germany, construction has exploded far beyond inflation and, most importantly, electricity prices are still ridiculous due to merit-order and the refusal of splitting up Germany into multiple power pricing regions.
I remember the price increase that Hetzner did during 2022 because of the invasion in Ukraine. The said they will adjust the prices down when the electricity price reduced.
Guess what? I am paying as a consumer about the same price as before 2022. Did Hetzner change their price down? Remember, the industrial price also dropped (and they also build out a large solar plant). No ...
Ok, inflation? But those price increases already covered part of that... Just saying, its not been the first price increase that happened. There have been multiple ones that Hetzner did over the years. Some flew under people radars.
> Payroll has seen significant increases in Germany,
Yea, we have seen nothing of that increase... O, wait, they reduce our income because the social security increase their costs. Yay ..
Since I started using Hetzner in 2020 they have increased prices 4 or 5 times and I am now paying 50% more than I started, but my grocery bill also went up 50% and my rent went up 50%, so that's just matching inflation (even though the government said inflation was 5%). Now they're doing a 300% increase (not for me) all at once.
The reasons really doesn't matter. As long as they are in top of price/performance/quality nobody will switch. Once they stop to be then people will think about it.
Its not only RAM. I have seen people who are vibe coding their app and instead of choosing Vercel as default they are learning about dedicated server hosting, docker etc and moving to providers like Hetzner. This is why whenever someone says - with AI everyone is going to write their own SaaS, I am always like - and what happens to hosting? Even Vercel might increase their costs if that comes to pass.
>billion dollar companies can't ramp up production?
you're a semiconductor manufacturer who wants to take advantage of the current boom. your options are:
A) invest a hundred cubic meters of money into doubling your manufacturing capacity
B) raise prices by 100%
I can't really blame them for going with B. the blame lies entirely with America's ability to invest billions of its infinite money into companies that make no profit now and have no plausible path to profitability in the future.
As I understand it ramping up a new fab takes a couple of years and several billion dollars. The last time they ramped up production prices had crashed back down by the the time the new fab was fully up and running, so this time they're betting that the scarcity will resolve itself like it did last time.
They are scaling up, but most will only come online in end 2027-2028 time frame. And Memory, as in what we use in PCs is easier to manufacture then HBM memory. But all the money is in HBM ...
So for every ~4GB of memory that you can produce in normal DDR5, you can only make 1GB of HBM. But you make multiple times the revenue.
The demand for HBM memory is not going to go away. LLMs are memory bandwidth hungry, and we are going to see production going to AI. But also to "lower end" like B200's.
That means, they are producing multiple times less memory (if we look for the normal market demand), but still need to produce more for the memory bandwidth hungry market.
We are seeing more products entering the "prosumer/business" market that are also memory bandwidth hungry. This demand will not go away. It will actually increase as companies move to more localized workloads. There is is a issue with data privacy that a lot of companies legally deal with.
The lacking ramp up is not a sign of them being scared of over production, its a realization that 3 companies hold the market in a strangle hold, and "slow" scale. If everybody plays friendly, they can milk this for years.
China is a solution but China does not have the HBM production levels, and will take years to scale and put a dent in the market. And China is ... allocating a lot to domestic production of AI > HBM ...
The reality is, that unless competition ( as in China ) does not start scaling beyond the expected levels, the big 3 have no reason to scale too fast.
And money is not the issue ... have you seen their revenue (and net profit!! ) numbers. A few billions is peanuts for them at this point. They simply do not want to scale too fast because that means less milking ... Memory demand is not going to away. When people talk about the AI bubble popping, its more in terms of the stock market. The product is here and not going away.
This does suggest a path to improvement, though. A significant factor in the demand for HBM is how expensive the actual GPU chips are, making you want to use the absolute best memory to support them. When there's more competition in GPUs and the memory is actually a lot of the total price, you see things like Apple silicon with LPDDR5 being very popular. You can get a lot of bandwidth out of normal memory if you put in 256 or 512 bit bus. If we can get more midrange competition, we can focus more manufacturing capacity back on some form of DDR, and lessen the squeeze.
Because they think it may be a bubble. If it's not, no harm done to the hardware manufacturers, they just make more money per unit, but if it is a bubble, they don't want to be stranded with excess capacity.
There is potential harm if it’s not a bubble and their competitors scale and capture more of the market and they don’t. That’s why CXMT is a real wildcard - they could use this situation to become a big player.
I use to be a huge Hetzner fan but it doesn't appear they have launch any new hardware in the past 3-years.
One of the reasons why I loved Hetzner so much is that you could always get the latest generation hardware ... but unless I have missed it - it seems like their hardware hasn't been refreshed in awhile.
(Still really like them, just wish they had dedicated servers in the US as well)
EDIT: maybe what I use in hardware is uncommon, but have been wanting an update to their AX102 line
Is there a lot of new enterprise server hardware coming out lately? Consumer stuff has been stagnant with all the ram cost issues so I could see servers running into similar issues.
it's not that bad, EX63 dedicated servers were a great deal until this week (especially with 192 GB ECC RAM upgrade) and it has a 2025 CPU, for example
It seems like what's missing here is lower cost plans, because the existing plans had been fairly affordable, but now they're basically triple.
The least expensive one seems to be CPX11, old price $6.99, new price $20.49. That's 2GB RAM, 40GB SSD. RAM and SSD are now much more expensive, fair enough, but maybe I don't need all that for my mostly-idle VM, so then where's the plan with ~0.67GB RAM and ~13GB SSD for the old price?
Problem with Scaleway (for me) is physics. I get 200ms (at minimum) roundtrip latency to any of their EU servers (and for their dedicated offerings that can boot custom Linux distros that's pretty much all I have access to).
Hetzner just achieved their pricing by using commodity consumer hardware.
This is now making them the canary, as they don't have the multi year business contacts the others have - so they're uniquely vulnerable to the current consumer hardware price increase.
But the rest will follow, unless the bubble burst, which is unlikely to happen before the others increase their costs, too
It was never the same quality though. A VM at AWS or GCP is not even remotely comparable to a VPS at some low-end provider. Things like RAID, network/powersupply redundancy, internal privacy controls (nothing stopping a VPS provider from snooping in your filesystem), software patch quality, noisy neighbor mitigation, bandwidth quality, SLAs etc. Then all the value-added stuff like Terraform support, logging, monitoring, IAM roles, instant scalability etc. Not to mention the benefits if you use the same cloud's managed services like databases.
If I can get a same-spec EC2 instance at AWS for just 2x the price of a Hetzner box, I would never consider Hetzner. It had to be a "10x savings".
The price of server auctions seem pretty much the same as before: https://www.hetzner.com/sb so if you don't need the latest hardware, Hetzner is still and outstanding deal.
They price hiked my existing CCX13 from 16€ to 43€ - that is a 2.6x price increase (!). Technically they didn't increase it yet, for as long as I don't order a new one or rescale (up/down) they let me use the old price. This is insane (it is a cloud server, with 2 dedicated cores and 8 GB RAM).
Yeah, I host three small-scale game servers for an indie-game, all out of pocket. 50-80 players hop on every day and the community is super nice. Its been possible because the price has been €16 per instance, but with these new prices, it will absolutely not be possible anymore. I imagine a lot of small game server communities will meet the same fate now, since Hetzner has been the go-to for this kind of stuff.
"Fun" times for services selling micro-VMs running on top of Hetzner bare metal machines for a small margin. 3x increase in input pricing with essentially no notice (there was a 2 weeks notice of "we will change pricing" but no details. People assumed a max of 50% increase, I guess, not 3x)
As a long term metal customer, I understood the need to raise prices for energy usage.... but for disk/ram I'm struggling to be sympathetic. The hardware I am using is already procured by them, and until such time there is a hardware failure I cannot support a price rise, because were is their justification for existing hardware?
An ongoing business that intends to remain ongoing has to charge current customers based on the replacement cost of consumable items used to service those customers.
That's why for example gasoline prices react almost immediately after something affects (or even threatens to affect) the price of crude oil, even at gas stations that have just filled their storage tanks and will be selling that already purchases gas for quite a while.
Most of us don't usually thing of computer hardware as a consumable but to a hosting business it effectively is.
Because in a free market, making rational choices about pricing in line with the industry allows you to build capital to further expand, which coincidentally also lets you buy more RAM.
Hetzner doesn't offer dedicated servers in the US. They operate two of their own data centers in Germany and one in Finland. In the US and Singapore they offer only cloud servers, running on hardware in others' datacenters.
Time to optimize your software stack. Stop using slow interpreted languages & bloated "scaling" setups in favour of tightly integrated natively compiled systems.
Getting close to losing any point of them existing. They were the cheap one. You take that away and their unique offering vanishes and makes them pretty pointless to even consider as a provider. Really hope they can sort out something for their hardware sourcing as this isn't sustainable.
Seems like it doesn't apply to older machines, I have AX41-NVMe, it's not on the list, I also didn't get any notification from them (and they usually send some) - no need to panic if you're longterm customer.
That'd be nice. The only server I can't host at home is basically just ferrying bytes to and fro, to have the right IP address for email deliverability. I'd buy DDR0.5 and a Pentium 2 if they'd offer it
Hetzner has long been affordable due to its subpar latency and protection, making this price increase very questionable. No one has real reason to choose hetzner if it is not affordable. OVH ends it, one Romanian provider as well.
I just used Claude to convert my app to a serverless architecture and migrated to Cloudflare and their generous tiers. Not every app fits that model, but it's more than you'd think. Now I only pay when the app is used, not a penny more.
1. Factory limits basically. There's a limit to the amount of fabrication lines that can create ram. Combined with the market incentives right now to make high bandwidth memory (HBM) over server memory (DRAM)... HBM starts as DRAM dies, so it competes with normal DRAM for wafer starts / cleanroom fab capacity.
2. Eventually more plants will come on line. Most of the main manufacturers have announced expansions but these can take O(years) to come online.
The entire capacity of RAM production is basically booked out, for at least the next year. All fabs have sold their allocations already, and it takes years to build a new one. As a result, no it will likely not ease up if demand continues like this, again for at least a year.
They sold their allocations to people who don't have a clear path to profitability, and were paid with massive amounts of money that don't exist in reality.
> 1. What's the bottleneck in ramping up RAM production? Is it the availability of silicon itself? Or the factories are at capacity?
For a RAM manufacturer, the incentive is to ramp up production AND prices. I doubt any of the names in the business is doing any work at all to lower their unit prices.
The current pricing isn't sustainable, and if you try to wring out the maximum amount of profit, you're gonna have competition spring up. The Chinese are probably salivating at this opportunity. Previously they would've had to sell memory at discount rates to get anyone to switch over to them, but now they'd have customers lined up for years if they were selling at the prices Samsung/Micron/SK Hynix were selling a year ago.
Contabo, an even cheaper alternative to Hetz, raised prices this month as well by 30-35% as well. Just noting it here because i couldn’t any info elsewhere.
Contabo is still notably cheaper than it was 2 years ago. I just shut down a 2 year old VPS and made a new one because it was ~20% cheaper than it was two years ago, and I also get more CPU and more RAM
Interesting, but still very good pricing compared to the enormous increases we now see at Hetzner (plus the increase before that one). I'm wondering if they have more hardware on stock and/or better means of hardware supply.
I have a few more or less idle VMs running at different providers, and keeping an eye on the european VPS market it seems that many struggled, while contabo is relatively stable. E.g. ovh at times limited the buying process to 1 instance (now it is at 5, not sure what was the default before) and also available locations were limited at one point a few weeks back.
Damn. I set up a 2vCPU 4GB RAM VPS on Linode for my web app (Linode 4GB). Was ready to move there. Then I found Hetzner has servers in the US and I could get more for my budget (~30$/month). Not anymore!
CPX22 (2vCPU 4GB) seems to be the exact same price as Linode 4GB (24$/month).
For the record, those providers already charged a significant premium for the equivalent compute you could get at Hetzner and OVH. The margins have decreased but they've had a profit margin on compute for a long time.
Yes, and they will to preserve margins at some point. They're just doing a huge Mexican standoff, waiting for others to move. All smaller clouds have raised some prices already.
They'll probably wait for summer, the world cup finals, or whatever's last big US government thing is so it flies under the radar.
Step change makes sense for a smaller provider. You bait people to build on top of you with the best prices, then rug pull and hope the switching costs and economic frictions delay the attrition.
idk if you're comparing like-for-like here. I think on EC2, other customers can balloon up to fill your idle time because they're running on the same physical hardware. What this is talking about is renting hardware that is exclusively earmarked for your usage, so your compute is not fungible with other customers. It's more correctly priced against buying your own hardware and maintaining it yourself.
Wow. This is crazy. And not only much more expensive but the standardization means you have to buy more disk if you e.g. just need more RAM or similar.
And does the standardization mean that I can no longer buy extra hardware?
They are also becoming greedy. I rent their 20GB VRAM instance GEX44, for which they now ask a 500 euro one-time setup fee. Whereas it was something like 60 euros a year ago.
We (company) do the same. Though in our case the setup cost was 80 Euros (I think), less than a year ago. As the GPU proved not really suitable for any serious server workloads (it's a workstation class card), we'll soon be ditching that machine anyways (now for sure). Maybe our other inventory at Hetzner too. Not even because of the price increases themselves, but rather because the way in which they've communicated those. Personally, I've been a loyal customer and avid advocate for Hetzner, for well over a decade. They sure knew how to nuke that in record time. They can spin their story any way they like, but I'd say their board better consider sending most of their management packing, without bonuses or severance pay.
Interesting over the price increase rollover now the setup fee is around 110 euros.
The machine itself is basically useless for any type of realtime inference, no matter what the marketing page states, but I still use it for prototyping LLM integrations and running comparisons across MoE models.
If only the alternatives to framework desktop wouldn't be so poorly built, I might swap it out for a local machine which has more ram but comparable performance for stuff like gpt-oss-20b (around 70tok/s)
> I rent their 20GB VRAM instance GEX44, for which they now ask a 500 euro one-time setup fee
Along with the increase in monthly prices they've dropped setup fees back to more approachable levels, though not as low as they were a year ago. For the GEX44 it was €79 a year ago, now €114. Monthly price was €184 a year ago, now €234.
You are saying that like it's some kind of rag pull. This is situation of the market. The people who moved still saved the money and are still saving money because Hetzner is still cheaper and they are also paying the old prices.
It's pretty silent happiness over at the Hetzner camp.
Hardware prices, especially RAM, have skyrocketed. Priced out new baremetal servers recently and prices were 3-6x what we paid 4 years ago for the newer equivalents.
This continuing trend is going to do a fantastic job of ensuring fewer and fewer individuals can launch casual projects and gating (non-VC) startups to those who already have the means.
For EU-based ops I moved to UpCloud, they have top customer service and their offerings are far more complete than Hetzner... Plus, they have more zones.
I was shopping for this past weekend. I found a 16core 128gb-ram(1!) unit for 110€/month with a competitor. Crazy deal. I suspected it to be a mistake, but after 24 hours it actually came online. Runs very smooth.
There are actually two separate pricing increases. One was introduced about 1-2 months ago. The one from today was announced at the end of May, without actually revealing new pricing then. The new levels were made public today or yesterday, I believe. And they are much bigger than before, some hikes are well above 200%.
I'll happily pay the new prices, if they actually have the servers available. Cheap pricing is nice, but not that useful when in practice you can't actually buy most of the time.
Short: Yes, but of course it depends.
Long: I am dealing with a huge amount of fitness related health data which gets aggregated into metrics. If those metrics are wrong, for whatever reason, this is not great.
If you just run some blogs, of course, this is not important.
The risk of bit flips is a percentage. The more RAM you have, the greater the likelihood. I have experienced heisenbugs on my 64 GB desktop that I bet were because of random bit flips
Just to throw it out there, surely we've ruled out ECC RAM gets away with worse modules because the checking bit will catch it? Because I actually tried to generate bit flips for a project once and we got nothing. Also a relatively high traffic site, registering a few bit flip variants (so like example.com -> fxample.com) got no hits whatsoever across a year. I keep seeing reports from people with ECC or people who found an available bit flip variant of a top X website, but it never happens to me on normal RAM. I can't imagine that ECC vendors are selling worse RAM with a check bit, but the more I read the more I wonder if maybe we should check just to be sure
AI seems to be ruining every single major thing that drove economic growth for the past 4 decades. PCs, the Web, software in general, high-capacity servers, Raspberry Pis and so on. The next thing to be affected will probably be smartphones. All of these things are foundations of profitable businesses right now and we are destroying them on the mere promise to get to some idiotic utopia in the future.
Correct me if I am wrong, but AI made software development and operations more expensive than before. Yes, it is faster too, but the question: is it worth the price? Can the users consume new features in that pace?
I start to think that AI might be a net negative on the economy. It consumes an enormous amount of resources, just to keep doing what we did before.
Laying off people also doesn't reduce cost as much as it might look like. There is a lot of hidden cost shared by everyone (also the companies that did the lay offs are hit by them). Unemployed people still have to eat and pay rent, and someone is paying for that. They spend less money on services and goods, which affects every company in the end.
AI is great, but I think it got too big.
Just my thoughts, not backed by any data. I'm not even sure I'm right.
> I start to think that AI might be a net negative on the economy. It consumes an enormous amount of resources, just to keep doing what we did before
Outside of HN, this is all people are seeing. Gamers in particular aren't seeing a benefit. They are being priced out of their hobby. The recent DDLS 5 meme is what people think of when they hear AI.
Even inside the software bubble it's debatable if there is a significant benefit on productivity (judged by total cost).
But even pure software companies are hit by higher hardware prices. Their customers need to buy expensive hardware and have less budget left to spend on software.
It will clearer soon after the api pricing model enforced on enterprise accounts.
I suspect this will soon follow and no fixed subscription model, which will enforce companies/developers to be moderate and thoughtful when using AI. Also I think Microsoft will do the same for copilot
I think we are also going to see a lot of new software that is less hungry on compute and RAM than before. Probably first with games not requiring new hardware anymore.
Because the consumers won't upgrade to new hardware as fast as before. People who buy their first gaming PC in 2027 might even get a lower spec in average than people who bought in 2025. So new games might require even lower hardware specs than before, to sell enough copies.
This is a wishful fantasy. Vibe coded applications like ChatGPT desktop and Claude Code routinely take 1~2GB of RAM to display <1kb of text. There is zero initiative to make better software, the software industry is just saying "fuck you" to the rest of the world until the bubble pops.
Sure, that's why I think it will start with games. You don't buy a game if you can't (properly) play it. Either you get a new gaming PC, or stop buying new games and keep playing old ones.
AI investors bet on the high return in the next 10-15 years, they have no reasons to care if everything which require semiconductors will become much more expensive in the process.
Yes but paying massively inflated prices for compute are not anymore. Even for consumer gear I am being quoted prices that are more than 2x what they were only a year ago (and availability is hard).
But its impact on new hardware costs and replacement costs is not.
A company like Hetzner probably replaces hardware on a 5 year cycle. Maybe shorter. Maybe they could try to stretch that out but they can't avoid the cost of new hardware for very long.
I think it is worse for Hetzner. The secret is out. Their prices have attracted a lot of demand, and they need to buy equipment to satisfy that demand.
The last Hetzner box I leased I had to poll for availability as if I was Ebay auction sniping. It took me 2 days to acquire it.
> AIslop is of higher quality than devslop on average.
Is it? If by higher quality, you mean commenting properly, sticking to naming conventions etc. I can agree. But to me, AIslop looks like it lacks "intentionality" of code written by devs, no matter how bad they are at naming things and sticking to conventions.
i.e. people who are adequately good at their jobs usually do things for a reason, and they can explain it. Even if you don't find it agreeable, it usually is consistent.
Is it AI that lacks intentionality or your prompts?
Just remember we are comparing slops. If you care about your code it really doesn't matter if you write it manually or with the help of a glorified typewriter.
> AIslop is of higher quality than devslop on average.
How did you come to that conclusion? That goes against everything I've heard from people who understand development. Every resource I can find about AI vs non-AI development comes to the exact opposite conclusion you did.
Devslop is spaghetti code that grows organically. AIslop is overly complicated code that someone neglected to read. I just distrust the former more than the latter.
AI could theoretically also be used to optimize existing code instead of producing new features, allowing existing tech to run on lighter hardware. Rent me an rpi if the software is fast.
Not really. You can still not use AI. The solid data is that RAM is more expensive than before… so there’s a small case to be made there because of you need a new computer you can’t avoid that.
> Can the users consume new features in that pace?
It's pretty clear by now that coding productivity increases by 10-15% with AI. Given coding is only a small part of the developer's job, there's just nothing new to consume.
The only change I have noticed in software since LLMs have hit the mass market is degradation of software quality, not increase in feature releases.
> It's pretty clear by now that coding productivity increases by 10-15% with AI.
Completely offtopic for this thread but I can't be the only one that would find this hilarious if it wasn't being said in earnest in every thread.
The only thing that is clear is that measuring programming is just as impossible as it has always been. In all my years of projects they've either been resounding successes or gone down in flames. The difference between good and bad is a difference in kind. Most of the bad ones didn't even know what the hell they were building and built the wrong thing.
Like, the entire idea that some omniscient manager is looking at a thousand timelines and pondering over whether to pick the $11.5M successful one or the $9.5M successful one is literally laughable. Half of them are going to make the Hindenburg look like a bit of a whoopsie and the other half you would lock in sight unseen without a second thought.
This is such disappointing news. I was planning on migrating some of our workloads to hetzner specifically to take advantage of the AX162 pricing which was incredibly competitive.
Does anyone else have any suggestions for competitive pricing for this kind of thing (e.g. batch jobs)? Was this applied retrospectively to existing customers?
And tells you in an FAQ to go to a competitor if you want to send emails (personal or commercial does not matter). Various bugs in the ordering systems too. Domain names run smoothly (fast and cheap) once you got the payment systems dealt with but for servers it doesn't seem great
>Was this applied retrospectively to existing customers
They are quite good at costs remaining predictable. However, a few years back they cut the low-end hosts 1Gbps unlimited data transfer down to a 20GiB/month cap, and wanted everyone to go full cloud/retard to fully leverage the hardware infrastructure.
If you serve large files, a CDN may have a very narrow use case where the budgets make sense. If you are already pushing 23 TiB/month, than cloud providers are usually not worth the effort. Some rent colocation rack space. =3
> If you serve large files, a CDN may have a very narrow use case where the budgets make sense. If you are already pushing 23 TiB/month, than cloud providers are usually not work the effort. Some rent colocation rack space. =3
Unfortunately I'm needing to run a lot of batch compute jobs (for which the hyperscalers are just insanely expensive - even to have a machine that outclasses a nice laptop becomes silly very rapidly)
I'm considering buying some machines and racking them in a colo but it feels like buying right now is also insane because of current pricing.
I have but it seems to me that the stuff that's actually particularly well priced is really old (2016-2018ish). Even a single DIMM of used 32G DDR4 from ETB is now > 250GBP. The used market is also blown to pieces because of the price pressure.
Yeah it doesn't look good. Idk your requirements but this is one of the better deals I could find https://www.ebay.co.uk/itm/126986603191 . For used eBay servers it's always worth making an offer rather than Buy it Now.
Most people IN TECH still don't realize how much hardware prices are going to continue to increase, much less the 95% of the population that's completely unplugged from what's going on with AI-driven hardware demand.
All of these price increases are going to get passed down to consumers eventually via increased prices.
Little bit of number crunching shows that the median increase is 112%, so a 2.12x increase. Median for GER/FIN region is 120%, USA region is 157%, and Singapore is 71%.
Worst of all increases is the Euro pricing for USA plan CPX41 with a 209% increase, or 3.09x.
Raw number comparison for those interested:
Tried to format it to be readable but its
[Plan] [Old Hourly] [New Hourly] [Percent Change] [Old Monthly] [New Monthly] [Percent Change]
The solution I think is to tell all of them starting with clouds to go fuck themselves (where possible of course) and start self hosting. I do this for years and am totally happy. I was also hosting on Hetzner and was thinking to do it again for some (not all) services but not with this new wonderful deal.
AWS does not give "bare metal" and the equivalent is still way more expensive. However Hetzner was much preferrable to let's say OVH who also "deals" bare metal but not anymore. So yes Hetzner has lost what I consider was their only advantage.
It's kinda awesome, the DDR prices will go extremely down (lower than before the bubble), and it will be spring time for self-hosted opensource models, cause people will be able to afford the hardware. We just need to wait for the peak.
2005 - 2015'ish the DDR prices were artificially inflated by factory fires and such in asia. hope it's not going to repeat, need to have some excuse to drive the prices up..
Just like the energy industry is killing itself until this idiotic bubble bursts again. Can't wait till oil / solar prices implode and there's a decade of drought where no one users energy for a long time.
It is sucking away electricity from cities, making hardware costly for common user, pumping heat into global atmosphere, taking away jobs, creating massive amount of slop, killing human motivation for creativity, making academics and exams harder, creating fakes that are undetectable, filling internet with plastic content, pumping tons of unmaintainable code, wrecking websites ...
Things do tend to end up being a net positive or negative, though. Even if both aspects are present, it doesn't seem far-fetched that someone with hindsight in the future would conclude that it was overall a bad thing in the middle of an energy/heating crisis. That would imo preclude the label 'great achievement', but time will have to tell (or we adjust how we use it before history decides on a judgement)
I was signing up for Hetzner years ago and it asked me to upload my passport to use their service.
At that same time, I was reading about this story about WireCard. It was like Stripe for Europe and worth billions. Turns out it was run by a Russian spy network and was all a sham. That video alleged Germany’s bureaucracy is filled with Russian agents and this can be traced back to the East/West Berlin days.
To save a few bucks a month over DO didn’t seem worth it to me to send my passport to a foreign country.
I actually sent them a picture of my passport, and they still denied my account.
Hetzner was widely recommended and I was more than happy to pay a premium for their supposedly-excellent service, but I guess they didn't want my money.
Oh well. Went with OVH instead, and haven't had any issues since.
Same exact story here, they denied my account despite me sending them everything they requested, no explanation given. Went with OVH and had zero issues.
I mean it's not great that Hetzner require this but it's a bit of a jump to assume that means they have links to Russian intelligence. This kind of thing is pretty common in Germany; not every private company is captured by Russian intelligence
Yes, no connection to WireCard but reading about that situation made me pause about giving my passport. At that point, I was like I’m trying to save a few bucks a month and risk is not worth it. Now if you’re buying their huge servers and are saving thousands, I can see why someone would do it.
They also don’t ask every person for the passport picture so maybe me using a custom DNS and VPN might’ve triggered something on their end.
> I was signing up for Hetzner years ago and it asked me to upload my passport to use their service.
I don't really understand what bothers you so much about providing a photo of a "passport" (if you are an European citizen they require a ID card) but credit card info didn't registered as a concern worth noting. Can you explain what is the difference?
Credit card is a largely fixed risk of financial loss, with some legal safeguards for recovery, and the ability to get a replacement card with a different number. Passport carries an open long-term risk of impersonation and you can't just get a new passport because some company has a copy. Just the financial side of that risk can have much greater impact. Unless a company has a legal requirement to "know your customer", e.g. a financial institution, this is a red flag.
Couldn’t have put it better myself. Even with payment processors, most they ask for is SSN and business EIN.
When I read about the WireCard scandal, the KYC stuff sent to them over the years is probably in the hands of foreign intelligence already. That’s what gave me pause.
When many sites are collecting these photos, it increases possibility of them leaking. Since these are also used for KYC process in crypto sites etc, this in turn increases risk of identity theft.
I'm a Hetzner user in the US, but I pay for it with PayPal and was never asked to give my passport or identity. Americans are very rarely asked for these documents online, and even then it's typically only for government or financial services. It's also drilled into us that this info can be used for identity theft, so it's only natural to be wary of any non-government entity asking for them.
FWIW, if Hetzner had asked for my passport when I signed up, I would not have given it either.
If they are spies, they are taking their time to use the data for sure. I run servers with them since 2009 or so. That's 17 years.
I feel like the whole password thing was meant as a protection against SPAM or using servers for nefarious purposes as they know who's really behind every server.
Although, I can also see how real criminals would work around that easily by supplying fake identities. Sounds like one of those "why we can't have nice things". Well, at least the password I gave them 17 years ago has expired since.
I was doing the same here, trying to set up a Hertzner account. Getting away from US companies and buying European and all that. But after I had made the account (and wasting a lot of time on back and forth with their buggy sign-up flow), I got told that I needed to upload a picture of my passport to do anything.
Nice work from DO marketing team! Prices are completely not comparable and Hetzner was fighting scammers and kiddies, because low prices worked like a magnet for those.
Russian spies? WOW, the earth got really flat these days. Seeing what US is doing with citizens and private companies I would love some Russian spy to be interested in exactly mine, boring passport.
Look I'm just a dude who happened to land on DO a decade ago due to podcast ads and now wants to move away from it. Not due to prices but because I would prefer a European alternative. I didn't bring up Russian spies and I don't know if there's any validity to that story, I just don't want to upload passport pictures to random services. Their competitors don't require it.
I'll probably find the time and energy to move to OVH or something some time.
I encourage you to read about WireCard. It wasn’t just a normal sham company, it was able to fool auditing firms (one of the big 4) and the executives got away with it and are in Russia hiding. I’m trying to dig up the video also I can link it. There is no connection between WireCard and Hetzner outside of both being German companies.
Time to dust off the Whataboutism Is Bad speech again: but don't glaze over yet, because I might say something new. Not only is it still as rhetorically fallacious as it has ever been to treat complaints about third parties like they are responsive to first order concerns, but also (wake up! here comes the new part!) the deeper problem with whataboutism is that it assumes people can't consistently object to both.
I don't expect this to persuade, to be clear. I don't believe that people engaging in whataboutism are unable to understand why it's wrong so much as they have a different approach to language that detaches it from accountability to any sort of conceptual coherence that people are normally searching for when testing integrity of arguments; commenting on it is more about revealing a difference in which background values inform the way you choose to communicate.
"And at that time, that's what everybody did. If a hotel wanted electricity, they had their own electric generator. And I looked at this, and I thought, this is what computation is like today. Everybody has their own data center. And that's not gonna last. It makes no sense. You're gonna buy compute off the grid. That's AWS."
- Jeff Bezos
There are just 3 or 4 DDRAM manufacturers (SK Hynix, Samsung, Micron). They fully intend to make it impractical to purchase a server outside of the hyperscalers.
When exactly are the upsides going to hit?
With some work on crypto many people could be a trillionaire on paper. Whether it translate to actual wealth and liquidity is another matter. We are talking about P/E of 300 / P/S of close to 100 for Tesla and SpaceX.
>When exactly are the upsides going to hit?
A lot of people take Moore's law, or technology improvement as granted. It will always come. It will always become cheaper. But none of that is true. Massive R&D is required along with ROI.
The AI Boom pushed a lot of technology forward by at least 2 - 3 years or 1 cycle. What normally would have taken 10 years to happen is now getting close to 5 years. We were suppose to stagnate or slow down with 3nm and 2nm, we are now rushing to push through everything from interconnect, smaller transistor and massive increase in Foundry capacity. PCI-Express 8.0, Nvidia Photonics, DRAM Improvement, HBM, HBF, even capacitor, immersive cooling. I don't even record the last time we had such a massive shift and changes in hardware technology. Even the start of smartphone era wasn't like this as majority of its start was picking on lower end PC components. Instead the AI is pushing the frontier hardware technology. With multiple trillion companies, insane appetite from market. We are basically saying we have Trillions to spend over the next 5 years. Give me everything you have got.
I mean for God's sake, people have been saying this shit for 3 years now, but nobody can fathom what those jobs may be.
;)
The next great American novels.
Honestly I could "retire" to a senior level role and have AI do 90% of my work and nobody would know the difference.
The benefits COULD hit by employers reducing everybody's hours.
The blocker to this is the middle management disease where there's a class of people who spend 40+ hours a week in some kind of update meeting or another and that much talking can't be replaced by AI. (much of it could be replaced by just not doing it any more but that's a different story)
There will be increased competition for job openings, reductions in real wages, or increased expectations of productivity. Probably some combination of all three.
Once it becomes the norm even for a small section of the economy it will spread.
People are more productive in an absolute sense working fewer hours anyway.
It just takes a union, an ambitious company, or a state to force that 30 hour workweek to show some success with better talent attraction and retention and better corporate results to start a trend.
It is possible for everybody to get a piece of the pie.
We are already productive enough to have a shorter work week and more leisure, anyone saying no has specific incentives to not support it (either via financial gain from the capital accumulation funnel or work bound to their identity).
https://hn.algolia.com/?q=4+day+week
(we get there eventually with structural demographics, it’ll just take longer)
Eventually a new startup will replace your large inefficient employer with people working 10% of their time.
All that seems to be happening is that these productivity gains roll up as profits for the owner class.
What's the point of all that, to me?
The moat of the owner class is lower because now information is everywhere and it's less possible to hide behind trade secrets and implementation effort.
Based on what? The macroeconomics don't work out that way. IF productivity goes up, but consumption does not, that means that it's harder to enter the owner class, because fewer productive enterprises (owned by non-working people) are supplying a larger share of customer demand.
This may make a difference on the margins for people in the software bubble. But for the other 8 billion people on the planet, they aren't all going to become owners in your brave new world, unless consumer demand goes to the moon to soak up all that productivity. It's not doing that. Prices aren't dropping. Quality isn't increasing.
If you think I'm wrong - is there a cross-economy explosion of small one-person businesses that I'm somehow not seeing? Are gigacorps across the board all losing market share? Because on the macro scale I see nothing but further consolidation.
Based on the far lower bar to get a product out the door.
1. Consumption goes up. (It's not going up. 40 hours at your job buys you less shit today than it did 3 years ago.)
2. Mega-corps start losing marketshare and revenue to this avalanche of new one-to-two-person businesses. (They aren't. Their revenues are climbing, which implies that consolidation is what's happening, not diversification.)
Your theory does not match reality.
It's not that people get lazy or are underemployed, but that expertise is just that valuable. You'd think this would be proof enough to break the AI echo chamber already.
Concretely: if you can do 90% of your work with AI, someone else can also do that same work, making you interchangeable unless that 10% is really important.
I think this is partly why it's so hard for people to find jobs right now. Everyone is interchangeable thanks to AI, so skill gives you less of an edge than it did in the past.
That's the part that is not true. Prompting and guard-rails and generally harness engineering do matter a lot lately. Seen it first-hand multiple times, especially after I used Fable 5 for a week.
And if it does take that long, why is it so great anyway?
Making labor hyper-interchangeable is kinda like the whole pitch here. It's two steps away from b2b SaaS labor if the PR is to be believed.
Maybe you can say you're an elite prompter or whatever, but it always kinda sounds to me like "I know the secret menu at taco bell." Like the whole point of the product here is precisely to not need such pretense or complexity. You are paying hundreds (at least) a month to use something, but also you are using it in a special way? I really don't get it.
Yes, you can vibe up a demo in no time. But LLMs still need guidance to produce an architecture that will hold up to real world scenarios.
Jobs were hard to find in the drawback after the COVID hiring boom in uncertain times as the result of Trump, inflation, tariffs, war, and the constantly impending but pushed off market crash we've been expecting since before COVID started. I'm not saying AI isn't contributing, but it's hardly the only factor.
AI is far cheaper to fire than a person.
"Everyone is interchangable" isn't quite right, a tremendous amount of people don't actually add all that much value and a lot of work is just running on a hamster wheel and now instead of taking time we've got a machine for running on hamster wheels for us.
This will never happen in the USA. Together with UBI.
Trouble is management, who has the signature on the bank account which ultimately controls if and when the bailiffs are going to knock down your front door and throw you onto the street.
Management thinks we can now continue putting more, much more, of the same compromised garbage. Cheaper, faster.
Are you getting paid more or are you just doing more for fun.
And then you’re fired.
They would notice, and then they would fire 9/10 of the people in your role. If you are unlucky, you get laid off. If you are lucky, you get to botsit full time for the workload of 10 engineers for less pay and no career advancement.
This would last until they figure out how to remove the human from the loop entirely.
Key factors for me:
That last part is probably the biggest risk, but we're in kind of a niche industry. Not really a big, juicy target.Now, does the AI write good code? Often not. But the codebase is already terrible, so it's no big difference.
I was listening to a podcast a couple days ago and Brad Gerstner was on and mentioned that with how AI is boosting productivity that perhaps one member of a household would be able to start staying home from work if they wanted. I shut off the podcast after that (to be fair, the podcast just seemed to be one massive SpaceX IPO pump).
It’s just so divorced from reality and every new advancement is just making *higher expectations for doing more work*.
The unfortunate reality is: Companies that are selling ai will sell that ai will make life easier. Companies that are buying ai will demand more from employees using ai (why else would they buy it?).
So when exactly is this productivity going to hit that doubles my income?
Still a win but not as big as many are selling it.
Actually good quality stuff is more affordable than ever. People just don't want to pay for quality and things that last.
It's hard to know whether moving up in pricing just buys unnecessary features in a checklist, higher quality veneer, brand name, or actual quality.
Compared to 50 years ago, the middle class is getting poorer.
What’s your data source?
Keep in mind that the modern, mass middle class was created in the mid-20th century through government policies and post-WWII economic growth.
https://www.ebsco.com/research-starters/history/real-estate-...
For example, the vehicle mortality rate is 1.44 per 100 million miles driven. That's down 17% from 2000 (so 25 years ago). However, the change from 1975 to 2000 was 53%. That's because as we get closer to 0, it gets harder and harder to improve those rates. On this metric at least, I don't think another 25 years will result in a noticeable amount of improvement?
In the other direction, some things will become scarcer (and therefore cost more). Real estate is the obvious one; we can't create more land, and we keep having more people. Easily accessible drinking water is another; desalination is getting cheaper, but it's still way more expensive than pumping aquifer water.
And some improvements are necessarily 1 time things. You can get tropical fruits year round, but that's been widely available since the 80-90's from what I can tell. So come 20 years from now, what will people be able to buy in a grocery store that I can't buy right now?
Being able to eat pork without cooking it to death for fear of trichinosis is a recent development.
Also, the old movies where someone tries to commit suicide by sticking their head in an oven. That was coal gas and we don't heat homes with it anymore.
Especially silly that you mention housing because if there's one thing that is absolutely fucked for the middle class of the 2020s is housing.
In every developed country whose numbers I've seen, the size of the average living space is up 30-50% since 50 years ago.
It's the same thing that happens to housing. People complain housing gets expensive because of "tech workers". No. The reality is greedy landlords can charge 20% more so they charge 20% more. They could be happy with what they make now, but no, they'd rather have the extra 20%. And so housing prices go up 20%.
The thing is, it's not just landlords that are greedy. Everyone is greedy. Companies are greedy. Yeah, you can get the same amount of work done in 1/5 the hours per day. But why not do 5x the work instead?
Not that I disagree otherwise though.
Never. At this point I think the only way out is a Sea Peoples[1] level of collapse. Maybe they'll call it the Late Chip Age collapse. People will not put with with being obsoleted. Americans at least have the means to resist. The rest of us will probably need 3D printers.
https://en.wikipedia.org/wiki/Late_Bronze_Age_collapse
But we ended up with lots more jobs than the ones that disappeared. The industry has kept millions of people employed for almost 50 years.
The billionaire founders are usually "worth" far less than the economic activities in the companies they own. Would the same economic activities have happened in a system where billionaires can't appear as a result of the same activity?
I don't see how, and it certainly hasn't happened yet.
Anwyay! Talking to my computer and have it do the things I tell it, like I'm in freggin' Star Trek, feels like a pretty huge upside already! :)
I wonder how much leverage the hyperscalers like AWS/GCP/Azure have on their own supply chain to keep costs level in their clouds.
I’ve already started buying cheap old business PCs just in case I’ll ever need to have simple barebones machines to run things on.
By the time you can have a slow death of personal computing, capacity will improve and prices will improve.
In the shorter term sitting on an old computer or regressing a couple years on specs or paying an extra $100/$200 for 8GB/16GB works.
> Even internet resources like servers will be hoarded by the hyperscalers that are the only ones who can afford to order years of compute hardware in advance.
I don't see why hyperscalers would be so much better at handling price increases.
For some average business paying a week's wages for the computer you use, they can afford that doubling to two weeks just fine.
For all the normal server rental companies, okay the guy on the $10 plan either pays $16 now or cuts their resource allocation and keeps paying $10. That's not going to cause a sea change. And higher end hosting isn't that much different.
Maybe if they're locking in long enough to fund new fab construction? But in that case after a few years a ton of capacity will come online so they're actually helping solve the problem.
I just hope my top shelf 2020-era desktop doesn’t die on me because it would get very expensive to get a new build these days.
I could probably sell my gaming rig (12900K, 64GB of DDR5, 4TB NVME, RTX 3090) for more today than what I built it for about 4 years ago, it's absurd. I won't, of course, because it's still glorious for 4K gaming even today. In retrospect, $5000 very well spent.
After winter, I started playing with various other GPU loads until LLMs and SD became easy enough to use. Now it's my experimentation machine.
It's already paid for itself, so anything I sell it for would be profit, but it is still super nice for running local LLMs that power various projects "for free".
Small Dell Optiplexes are good for desktop computers.
Hmmmm...
That's a bit dramatic. I did buy a few SSDs when RAM prices started going up for I didn't want to be facing a shortage of SSDs but China is already very busy at producing cheap server motherboards: gone are the days where he only option for a server motherboard was a $1000 one. There are ultra cheap, and fully functional, chinese motherboards now.
I'm totally convinced China is not going to sit doing nothing: if RAM prices goes up and there's a business opportunity, China is going to seize it and start producing lots of RAM.
We did see "RAM price quadruple in no time so nobody had time to adapt", that's for sure.
But I'm really not sure it's "RAM prices goes 4x and then the world stays as it is and nobody adapts for decades to come".
Also as to used servers: people ordering years of compute hardware in advance aren't hoarding five years old servers.
Heck, a ten years old Xeon machine is plenty capable and the usual people are buying stocks of those (from companies updating their fleet), refurbishing them and reselling them one by one on the usual marketplace, at the same prices (OK maybe instead of 150 EUR it's now 200 EUR if there's 64 GB of ECC RAM).
My usual seller is doing its business as usual although we're well into the memory craze.
Just to put things into perspective: the "PC sales crash" from 2021 to 2024 saw, what, a 20% drop, 10% of which already recovered. In 4 years more than a billion PCs were produced.
A PC is not a rare thing. We won't run out of PCs, just like we won't run out of cheap used ten years old Xeon servers.
But OK you got me: I may contact my seller and hoard two or three more just in case.
Sheesh is HN full of paranoid people and, darn, is it contagious.
But what you see is a cautious strategy from the existing players. They are hedging against a bubble. They don’t want to pour today tens of billions of dollars in capacity that they will have to sell it to a deflated market
IF IT LASTS, capacity will increase.
But it won't last. The AI boom is in exponential growth but it's based on heavy speculation about future value and the bubble will absolutely pop, how agressively depends on how dumb people are about now. The current growth may or may not be entirely justified but it's not sustainable, the free investor money does run out. These back and forth self-dealing deals where companies that own big pieces of each other announce "partnerships" where companies are selling resources essentially to themselves and counting the revenue several times... those are a sign of the approaching peak.
I've been saying the same thing, but that's why they made the move to IPO, no?
There are only so many trillion dollar IPOs out there. And then what next?
I wish I could say I am disappointed.
Either it's an established vendor with designs and fabs or it's a newcomer that needs to invest a massive pile of cash in designs and fabs. Neither are cheap.
Seems kinda hard to believe at this point, no?
There have been SEVERAL crashes that have wiped out the market and it's the reason there are so few players, the rest of them went bankrupt after periods of over-expansion. (in the 80s caused by Japan, in 1997, in 2001ish after the dotcom bust)
You're even calling it a bubble so it's not exactly "hard to believe" it will pop.
Insufficient law enforcement. The same memory manufacturers already broke antimonopoly laws in the past, pleaded guilty. Apparently the fines were too small for these companies to care, and the people responsible were promoted instead of being punished. More information: https://en.wikipedia.org/wiki/DRAM_price_fixing_scandal
In this context, it takes spending enormous piles of money over the course of at least several years to spin up new semiconductor production.
We do need more capacity to keep up with AI datacenters' usage, yes.
But adding long-term capacity years down the road for a thing that some folks seem to confidently think is a bubble that can pop at any time is risky. And (because capitalism), we have to manage carefully balance our risks and rewards in order to maximize our odds of success.
If there is no bubble and demand stays high long-term, then the payoff for that risk is potentially enormous.
If there is a bubble and it bursts, then the cost of that risk is potentially devastating.
(Capitalism works most-predictably when cheating is possible, such as with Biff's use of the time machine in Back to the Future II. But without cheats, it's always a gamble.)
To show they’re working on reducing the impact of data centres on the environment, and that they’re taking action on e-waste, all while saying their pixel phones are so powerful they can be clustered into servers.
And their announced test with 2000 phones, where one server is 25-50 phones, is only 40-80 servers. Interesting, but hardly hyper scale.
If you buy a dedicated server at Hetzner, you actually need immediate hardware.
Many VPS providers also just resell Hetzner, OVH or other dedicated servers so they won’t increase the price until their own provider does.
Hetzner has a "cloud" offering. The price increases aren't small either.
They shifted right (VPS-1 2026 is now VPS-2 2027) and increased prices.
Crazy stuff
For the exact same specs (CX23 vs VPS-1 2027), price is 6 vs 4.5 € and you can get a 15% discount on OVH if you order a full year.
So far, haven't seen any other notable cloud price increases. Thought for sure they'd be reevaluating by now, I'm surprised to see the stability.
The increase was 25% and was, of course, mainly due to hard drive prices.
The peering announcement or did I miss something?
I doubt this has to do with the hardware discussion. This is just them increasing their lock-in and trying to curb businesses running to other CDNs (whole point of the peering).
Has anyone here used Vultr much? I'm curious how they felt about bang for buck. At least with Hatchbox it's easy to run multiple domains on one box.
1. https://sive.rs/ti
Another possibility: They were growing too fast and need to slow down. At some point additional growth might become too risky, or even exponentially more expensive. It might require fundamental organizational changes.
They look at the numbers and see the risk of making less profit than before, if they expand. Especially if demand decreases at some point, instead of growing further. So they decide to just raise the prices, lower demand and make even more money without additional risk.
Hetzner and OVH and other bare metal but low cost providers use commodity hardware. When that commodity hardware increases there is simply no other option. The secret to the success of these providers is using common off-the-shelf hardware instead of specialized server hardware, which is now being cannibalized.
It seems they have shifted by reducing the setup fees, and increasing the monthly costs. As this generates more revenue. And its easy to prove this...
AX42 ... Its 8700GE that has gone from 65 Euro to 225 Euro. With the setup fee now being 112 Euro instead of 225 Euro. It has 64GB memory, and 1TB storage. The storage even in todays market is 100 Euro. The memory is 644 Euro.
Do the math ... Hetzner servers had a hardware payback periode of between 9 to 11 month if you took the market value. This calculation has always been very stable over the 20 years i used Hetzner.
This new price, reduced the hardware payback periode to ~4 month. It seems to be that Hetzer is trying to use the memory price issues, as a excuse. The revenue of those same servers now increased to a insane level. More revenue with less hardware.
The real issue is that a lot of companies are moving from US hosting to EU hosting because of the problems with the US. Hetzner sees this as the perfect time to cash in on Enterprise customers.
They have been trying to replace the "cheap" normal consumers with enterprise. This trend has been going on for a while already.
Every customer that now leaves, is a server they can rent out to business customers.
If you want to see the same thing, look up what happened to Microsoft/Github Copilot where they turn around has been sudden and very strong, with a clear goal of moving everything to enterprise.
Monthly costs have gone up as well. Payroll has seen significant increases in Germany, construction has exploded far beyond inflation and, most importantly, electricity prices are still ridiculous due to merit-order and the refusal of splitting up Germany into multiple power pricing regions.
Guess what? I am paying as a consumer about the same price as before 2022. Did Hetzner change their price down? Remember, the industrial price also dropped (and they also build out a large solar plant). No ...
Ok, inflation? But those price increases already covered part of that... Just saying, its not been the first price increase that happened. There have been multiple ones that Hetzner did over the years. Some flew under people radars.
> Payroll has seen significant increases in Germany,
Yea, we have seen nothing of that increase... O, wait, they reduce our income because the social security increase their costs. Yay ..
There is an engineered scarcity, billion dollar companies can't ramp up production?
Murica is stuck depending on the good will of Korea and China for thinking rocks? le fucking mao
you're a semiconductor manufacturer who wants to take advantage of the current boom. your options are:
A) invest a hundred cubic meters of money into doubling your manufacturing capacity
B) raise prices by 100%
I can't really blame them for going with B. the blame lies entirely with America's ability to invest billions of its infinite money into companies that make no profit now and have no plausible path to profitability in the future.
So for every ~4GB of memory that you can produce in normal DDR5, you can only make 1GB of HBM. But you make multiple times the revenue.
The demand for HBM memory is not going to go away. LLMs are memory bandwidth hungry, and we are going to see production going to AI. But also to "lower end" like B200's.
That means, they are producing multiple times less memory (if we look for the normal market demand), but still need to produce more for the memory bandwidth hungry market.
We are seeing more products entering the "prosumer/business" market that are also memory bandwidth hungry. This demand will not go away. It will actually increase as companies move to more localized workloads. There is is a issue with data privacy that a lot of companies legally deal with.
The lacking ramp up is not a sign of them being scared of over production, its a realization that 3 companies hold the market in a strangle hold, and "slow" scale. If everybody plays friendly, they can milk this for years.
China is a solution but China does not have the HBM production levels, and will take years to scale and put a dent in the market. And China is ... allocating a lot to domestic production of AI > HBM ...
The reality is, that unless competition ( as in China ) does not start scaling beyond the expected levels, the big 3 have no reason to scale too fast.
And money is not the issue ... have you seen their revenue (and net profit!! ) numbers. A few billions is peanuts for them at this point. They simply do not want to scale too fast because that means less milking ... Memory demand is not going to away. When people talk about the AI bubble popping, its more in terms of the stock market. The product is here and not going away.
The person you're replying to explained why they're not ramping up, and you replied "They are not ramping up", which seems awfully silly.
Old prices: https://web.archive.org/web/20260513201413/https://docs.hetz...
One of the reasons why I loved Hetzner so much is that you could always get the latest generation hardware ... but unless I have missed it - it seems like their hardware hasn't been refreshed in awhile.
(Still really like them, just wish they had dedicated servers in the US as well)
EDIT: maybe what I use in hardware is uncommon, but have been wanting an update to their AX102 line
https://www.hetzner.com/dedicated-rootserver/ax102-u/
The least expensive one seems to be CPX11, old price $6.99, new price $20.49. That's 2GB RAM, 40GB SSD. RAM and SSD are now much more expensive, fair enough, but maybe I don't need all that for my mostly-idle VM, so then where's the plan with ~0.67GB RAM and ~13GB SSD for the old price?
https://news.ycombinator.com/item?id=47944914
Hetzner just achieved their pricing by using commodity consumer hardware.
This is now making them the canary, as they don't have the multi year business contacts the others have - so they're uniquely vulnerable to the current consumer hardware price increase.
But the rest will follow, unless the bubble burst, which is unlikely to happen before the others increase their costs, too
If I can get a same-spec EC2 instance at AWS for just 2x the price of a Hetzner box, I would never consider Hetzner. It had to be a "10x savings".
(However, Hetzner did an earlier price increase 38 days ago. HN's submission logic sends posting the url to the previous discussion: https://news.ycombinator.com/item?id=48306066)
Although I did plan for OVH-level dedicated server prices, so as long as they don't jack up prices too I'll be fine...
That's why for example gasoline prices react almost immediately after something affects (or even threatens to affect) the price of crude oil, even at gas stations that have just filled their storage tanks and will be selling that already purchases gas for quite a while.
Most of us don't usually thing of computer hardware as a consumable but to a hosting business it effectively is.
https://www.hetzner.com/sb/#ram_from=256
Yeah mostly old CPUs, but considering RAM shortages gonna be much cheaper than colocation.
PS: link contains 256GB RAM filter since I guess OP need RAM.
Advertised prices for my setup are now roughly 2x what I'm currently paying.
I've been using Hetzner for many years, both personally and for business use, and I've not seen any noticeable issues regarding the latency.
Granted, my use cases are webapps/backends that are not particularly latency sensitive, and are primarily used from a few European countries.
For what's worth, I've seen cases where download speeds from Hetzner are considerably higher than from AWS eu-central-1.
But why did the the CX33 only go up by €2 whereas the CPX32 went up by €21.50?
Both have 8GB of RAM, but the CPX32 has 80GB more storage and a bigger slice of CPU time.
You can see the same trend with CX23 vs CPX22.
In fact, CPX22 is now more than twice the price of CX23, despite having the same amount of disk and half the RAM.
Is there a CPU shortage now too?
For orders placed before 15 June 2026, but delivered after 15 June 2026, the previous prices will apply.
A couple of naive questions:
1. What's the bottleneck in ramping up RAM production? Is it the availability of silicon itself? Or the factories are at capacity?
2. Is this supposed to ease up despite the AI boom? Definitely would ease up if busted.
2. Eventually more plants will come on line. Most of the main manufacturers have announced expansions but these can take O(years) to come online.
They sold their allocations to people who don't have a clear path to profitability, and were paid with massive amounts of money that don't exist in reality.
How's that possible?
For a RAM manufacturer, the incentive is to ramp up production AND prices. I doubt any of the names in the business is doing any work at all to lower their unit prices.
Safe to say they're not in it out of sheer altruism.
I have a few more or less idle VMs running at different providers, and keeping an eye on the european VPS market it seems that many struggled, while contabo is relatively stable. E.g. ovh at times limited the buying process to 1 instance (now it is at 5, not sure what was the default before) and also available locations were limited at one point a few weeks back.
CPX22 (2vCPU 4GB) seems to be the exact same price as Linode 4GB (24$/month).
Linode dashboard is really good.
They'll probably wait for summer, the world cup finals, or whatever's last big US government thing is so it flies under the radar.
Edit: just noticed this is not retroactive. Still concerning looking forward.
And does the standardization mean that I can no longer buy extra hardware?
https://docs.hetzner.com/robot/dedicated-server/dedicated-se...
The machine itself is basically useless for any type of realtime inference, no matter what the marketing page states, but I still use it for prototyping LLM integrations and running comparisons across MoE models.
If only the alternatives to framework desktop wouldn't be so poorly built, I might swap it out for a local machine which has more ram but comparable performance for stuff like gpt-oss-20b (around 70tok/s)
Along with the increase in monthly prices they've dropped setup fees back to more approachable levels, though not as low as they were a year ago. For the GEX44 it was €79 a year ago, now €114. Monthly price was €184 a year ago, now €234.
But just like a low-interest rate mortgage, I'm going to be stuck with this thing for a long time, it seems.
I told ya about silent happiness…
It's pretty silent happiness over at the Hetzner camp.
What's the next best option now?
https://www.netcup.com/en/server
If not then it is only a matter of time before other providers are forced into similar price hikes.
I mean don't get me wrong, this for sure is a factor but like others said, other services don't see such drastic price hikes.
See: https://news.ycombinator.com/item?id=48307959
For example, their 'Regular Performance' cloud server tier has seen a 173% price increase.
AX162 (256GB) went from €274 -> €844
big sigh of relief
So glad I got all I needed recently.
If you just run some blogs, of course, this is not important.
I built a homelab before the crisis started which might allow me to survive this for the next few years.
But man am I sad about folks trying to build new projects.
Laying off people also doesn't reduce cost as much as it might look like. There is a lot of hidden cost shared by everyone (also the companies that did the lay offs are hit by them). Unemployed people still have to eat and pay rent, and someone is paying for that. They spend less money on services and goods, which affects every company in the end.
AI is great, but I think it got too big.
Just my thoughts, not backed by any data. I'm not even sure I'm right.
Outside of HN, this is all people are seeing. Gamers in particular aren't seeing a benefit. They are being priced out of their hobby. The recent DDLS 5 meme is what people think of when they hear AI.
But even pure software companies are hit by higher hardware prices. Their customers need to buy expensive hardware and have less budget left to spend on software.
I suspect this will soon follow and no fixed subscription model, which will enforce companies/developers to be moderate and thoughtful when using AI. Also I think Microsoft will do the same for copilot
Because the consumers won't upgrade to new hardware as fast as before. People who buy their first gaming PC in 2027 might even get a lower spec in average than people who bought in 2025. So new games might require even lower hardware specs than before, to sell enough copies.
A company like Hetzner probably replaces hardware on a 5 year cycle. Maybe shorter. Maybe they could try to stretch that out but they can't avoid the cost of new hardware for very long.
The last Hetzner box I leased I had to poll for availability as if I was Ebay auction sniping. It took me 2 days to acquire it.
Those aren't the only metrics, quality and efficiency is also important. AIslop is of higher quality than devslop on average.
Is it? If by higher quality, you mean commenting properly, sticking to naming conventions etc. I can agree. But to me, AIslop looks like it lacks "intentionality" of code written by devs, no matter how bad they are at naming things and sticking to conventions.
i.e. people who are adequately good at their jobs usually do things for a reason, and they can explain it. Even if you don't find it agreeable, it usually is consistent.
Just remember we are comparing slops. If you care about your code it really doesn't matter if you write it manually or with the help of a glorified typewriter.
How did you come to that conclusion? That goes against everything I've heard from people who understand development. Every resource I can find about AI vs non-AI development comes to the exact opposite conclusion you did.
It's pretty clear by now that coding productivity increases by 10-15% with AI. Given coding is only a small part of the developer's job, there's just nothing new to consume.
The only change I have noticed in software since LLMs have hit the mass market is degradation of software quality, not increase in feature releases.
Prices have increased for literally nothing.
Not fully true. AI is now often used to fix a lot of bugs in old and badly maintained software.
The quality of big and popular software probably decreased a bit, but the quality of niche products probably improved.
Completely offtopic for this thread but I can't be the only one that would find this hilarious if it wasn't being said in earnest in every thread.
The only thing that is clear is that measuring programming is just as impossible as it has always been. In all my years of projects they've either been resounding successes or gone down in flames. The difference between good and bad is a difference in kind. Most of the bad ones didn't even know what the hell they were building and built the wrong thing.
Like, the entire idea that some omniscient manager is looking at a thousand timelines and pondering over whether to pick the $11.5M successful one or the $9.5M successful one is literally laughable. Half of them are going to make the Hindenburg look like a bit of a whoopsie and the other half you would lock in sight unseen without a second thought.
Sorry, I meant 10-15% at most.
If it was by more than that then we'd see the effects in an obvious way. Since we don't those 15% are already generous.
None of OVH, GCP, AWS, Azure wanted so much data about me, and I run my services in all of them successfully. Not in Hetzner.
Sorry Hetzner, you're too data-hungry. Nothing you say justifies that.
Does anyone else have any suggestions for competitive pricing for this kind of thing (e.g. batch jobs)? Was this applied retrospectively to existing customers?
They are quite good at costs remaining predictable. However, a few years back they cut the low-end hosts 1Gbps unlimited data transfer down to a 20GiB/month cap, and wanted everyone to go full cloud/retard to fully leverage the hardware infrastructure.
If you serve large files, a CDN may have a very narrow use case where the budgets make sense. If you are already pushing 23 TiB/month, than cloud providers are usually not worth the effort. Some rent colocation rack space. =3
Unfortunately I'm needing to run a lot of batch compute jobs (for which the hyperscalers are just insanely expensive - even to have a machine that outclasses a nice laptop becomes silly very rapidly)
I'm considering buying some machines and racking them in a colo but it feels like buying right now is also insane because of current pricing.
Indeed, never buy equipment unless all other choices were explored.
Note, we may be waiting till 2029 for GPU/ddr7/flash prices to fully normalize. =3
i'm currently using their hillsboro instances.
i'm not going to pay 3-4x more.
existing VPS will have the old price.
It get increased only if it is rescaled and for new VPSs.
but do you know any alternatives to Hetzner for US?
All of these price increases are going to get passed down to consumers eventually via increased prices.
Raw number comparison for those interested: Tried to format it to be readable but its [Plan] [Old Hourly] [New Hourly] [Percent Change] [Old Monthly] [New Monthly] [Percent Change]
But when Hetzner is priced like everyone else, it makes it harder to pick them over the giant AWS, GCP, Digital Ocean etc.
Tech is killing itself until this idiotic bubble bursts.
Then we'll be in a decade of drought again like 2001
Why? From what I understand, Ram production is not ramping up. Even if a bubble does pop, I don't think it's even guaranteed to drop to where it was.
Still - AI is a great achievement?
At that same time, I was reading about this story about WireCard. It was like Stripe for Europe and worth billions. Turns out it was run by a Russian spy network and was all a sham. That video alleged Germany’s bureaucracy is filled with Russian agents and this can be traced back to the East/West Berlin days.
To save a few bucks a month over DO didn’t seem worth it to me to send my passport to a foreign country.
Hetzner was widely recommended and I was more than happy to pay a premium for their supposedly-excellent service, but I guess they didn't want my money.
Oh well. Went with OVH instead, and haven't had any issues since.
They also don’t ask every person for the passport picture so maybe me using a custom DNS and VPN might’ve triggered something on their end.
I don't really understand what bothers you so much about providing a photo of a "passport" (if you are an European citizen they require a ID card) but credit card info didn't registered as a concern worth noting. Can you explain what is the difference?
When I read about the WireCard scandal, the KYC stuff sent to them over the years is probably in the hands of foreign intelligence already. That’s what gave me pause.
Germany also has legal KYC requirements for web hosting and most other things relating to telecommunications.
FWIW, if Hetzner had asked for my passport when I signed up, I would not have given it either.
I feel like the whole password thing was meant as a protection against SPAM or using servers for nefarious purposes as they know who's really behind every server.
Although, I can also see how real criminals would work around that easily by supplying fake identities. Sounds like one of those "why we can't have nice things". Well, at least the password I gave them 17 years ago has expired since.
Fuck no. I too decided to stick with DO.
Russian spies? WOW, the earth got really flat these days. Seeing what US is doing with citizens and private companies I would love some Russian spy to be interested in exactly mine, boring passport.
I'll probably find the time and energy to move to OVH or something some time.
I don't expect this to persuade, to be clear. I don't believe that people engaging in whataboutism are unable to understand why it's wrong so much as they have a different approach to language that detaches it from accountability to any sort of conceptual coherence that people are normally searching for when testing integrity of arguments; commenting on it is more about revealing a difference in which background values inform the way you choose to communicate.
- Jeff Bezos
There are just 3 or 4 DDRAM manufacturers (SK Hynix, Samsung, Micron). They fully intend to make it impractical to purchase a server outside of the hyperscalers.