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.
GP's point is this isn't the 2nd car, this is the 41st car. If they had 1 car it'd be a 2x increase. If they have 40 cars, the only way the 41st car would lead to a 2x increase is if that single car cost as much as 40 other cars.
That's what the post you're responding to was asking.
The real question is whether it's actually just the 41st car or if the demand is such that they'd have to go from 40 cars to 200 and then risk having the demand fall back off after they've already sign on to making 160 more car payments.
For the 41st car they might need to build a new parking deck. Or hire a fleet manager, because the COO can't handle it anymore. They might already run over planned capacity and every new vehicle makes it exponentially worse.
For bigger orgs the bottlenecks are a bit more vague, but they still exist.
Do we actually know that? If you're an AI company during a hardware supply crunch and Hetzner has a bunch of servers they could put GPUs in and rent out to you, what makes you ignore them?
It's just not in their DNA. Hetzner is very much a do-one-thing-well business.
They might contract to AI companies to supply servers for their AI, but the total capacity of Hetzner is less than a single AI DC so it doesn't seem very useful.
It's a business that seems to have high operational leverage. Effectively, similar to airlines. Enormous capital outlays with low marginal costs, so once the current infra has been built at cost x, the additional data centre at cost 5x (or whatever multiple) might mean that it's not profitable to keep serving at the current prices.
Businesses charge their customers as much as they can. Businesses want to raise prices almost regardless of what their input costs are.
It is wrong to believe that a product's "correct" price is simply its cost plus a reasonable markup.
There's other factors that might limit how much a business can charge their customers. But an ideal business acts more like a monopoly and charges far far more than is 'fair'.
In theory competition keeps prices in check: in practice competition doesn't work as advertised.
Your overall point is true - but I'm unsure whether you are clarifying or trying to ackshualise.
And well actually, "price/demand elasticity curve" appears to be ambiguous word-salad (although a human mistake;)
I do appreciate your response: it makes me think about what I should have written. You are correct that my comment was rather unclear. Most of us misunderstand how businesses maximize profits. I find economics hard.
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.
Then check out the Serverbörse https://www.hetzner.com/de/sb/ that's exactly what you're asking for. Used servers, some of them with Skylake (2015) processors, some even with Haswell (2014) processors.
Hetzner keeps running the old machines as long as they can find customers for them, which means they have entire buildings of 10+ year old machines still running.
It's not even the reliability which is the issue. Newer servers can put hundreds of cores in one physical machine, while taking up the same amount of rack space and using the same amount of electricity as older systems with tens of cores.
How long do you want to run something that uses 3x the electricity for the same level of performance when you're buying power by the megawatt? How about the even older ones that use 10x as much?
Yes that's a thing too, but it's a question of cost and potential revenue. So if replacement hardware is really expensive then maybe you make the inefficient stuff hang out longer, as long as it's not broken and you have paying customers.
But at some point hardware does break and if you're going to keep the datacenter open for business you'll need to address turnover of inventory on an ongoing basis, and if you haven't locked in long-term deals for hardware then you'll have to bear current market prices when you do that.
I would suspect that "hardware breaks over time" is a minor component of the cost. If you put computer hardware in a climate controlled environment with conditioned power, generally >90% of it will still be operational after a decade, with the expected replacement of wear items like fans and drives. To the point that the fraction of older hardware that fails will typically be smaller than the fraction you would replace for efficiency reasons regardless, and wouldn't be a large proportion of operating costs even if it wasn't.
Exceptions naturally if you for some reason had an abnormally high failure rate, e.g. capacitor plague.
> The price adjustment applies to new orders and cloud instance rescales starting from 15 June 2026; 8 AM CEST.
> For orders placed before 15 June 2026, but delivered after 15 June 2026, the previous prices will apply.
I am surely/definitely happy that the price doesn't increase for me. If it increased for everyone, I'd expect a much smaller jump. Also I'm not sure how much flexibility they have to increase the price for everyone without notice, given that they are in EU.
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?
> 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.
If the rest of the market stays at current prices sure. It's very likely that Hetzner is just popular and ran out of hw quicker than others. I doubt they changed their market strategy, they are probably just being realistic. We will see if/when the hikes come to others.
What sucks is that it might just be that hyperscalers have long term hw contracts and completely starve any competition.
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.
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.
I almost pulled the trigger on a 2x64GB 6400MHz CL36 kit for about $650 last August. It's been insane since the spike happened, but went down somewhat last month. I just checked and it has gone up $1500 since then to $5000.
So that's over 7.5x what it was when I (sadly) did not buy it. Totally not kicking myself for dragging my heels on that purchase...
Cheapest 2x48GB kit I can find here now is $1500, yay. That's 5200MHz CL48 stuff.
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.
All the money in the world won't spin up a fab in a year, much like all the women in the world couldn't take a child from conception to birth in a month. Some things do just have a latency to them.
>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.
It seems to be 2-tier pricing. You can pay 4x for guaranteed delivery... they may have excess capacity at some point (-LTD SKUs) and then you can pay 1.1x while stocks last.
