Hacker News Reader

Comment by lukeqsee | Hacker News Reader

While the pricing is interesting, I think the fact it is driven by the input costs of electricity is even more telling.

The fact that electricity costs is the dominating (edit, correction: marginal) cost of cloud computing basically means that:

1) We're up against physical barriers to lowering the cost (and ecological impact) of our clouds.

2) Optimizing for less compute will pay multi-fold dividends over the next few years where we can expect energy prices to continue rising.

3) Building more power-efficient chips is a big deal, and I'm glad to see more general ARM deployment, given their efficiency gains.

View on HN

To be fair, Hetzner is notorious for pushing down prices for everything as far down as they can by any means necessary. So this most likely makes their price point more sensitive than many other providers to any cost increase. Here is a great walk-through of one of their data centres [1] and you will see how streamlined everything is down to the hardware components. Nearly nothing goes to waste.

[1]: https://yewtu.be/watch?v=5eo8nz_niiM

Why "notorious" though? Isn't this a good thing for them and for customers as well?

I would expect "famous", since this seems lika a feat/achievement to me.

"Notorious" isn't always negative—the connotations of the word depend on context (and probably regional use?) in sometimes confusing ways.

Indeed! It's a really interesting usage. Maybe it developed in a similar way as "badass" to convey a sense of (positive, approved) vaguely dangerous wildness and coolness.

It kind of makes sense here -- Hetzner's competitors probably hate this "notorious" behavior of being so good at ruthlessly[1] slashing prices

[1] See, another example of this sort of thing. Cool!

Interesting, didn't know that. Thanks for elaborating!

You still don't! I was speculating :)

For sure, and I agree, but that's underscoring my point.

The marginal cost of cloud computing isn't capex, it's opex. And of that opex, it's mostly electricity. We aren't paying for the hardware so much as we're paying for the cost of keeping it turned on and cool.

Hetzner has a data center in Finland as well. While the electricity price in Finland is going up, it is not going up as much as in Germany due to availability of nuclear power. If you want to use Hetzner services I recommend picking a location in Finland.

How is the peering to Finland nowadays? Last I heard it was all still going through their German centres.

From Telekom Germany:

Telekom -> Level3 (*.edge1.helsinki1.level3.net et al.) -> Hetzner Helsinki (core*.hel1.hetzner.com)

Wow, that's really cool. Love the old midi-towers.

I mean if your competition is Amazon I image you'd have to be like that.

Amazon's margins are astronomical. Hetzner doesn't need to give up any where near as much consumer surplus as they do in order to compete on price, but having a solid offering at very low price point is how they have chosen to differentiate themselves from other providers on the VPS to cloud spectrum. Arguably they are competing more against Linode, OVH, digital ocean etc..

Massive respect to them for linking the change to electricity prices rather than just "inflation". They know that by doing this there will be pressure to reverse the increase once electricity prices normalise.

Their competition charges XXX% markups on everything but EC2...

EC2 is definitely at a premium. If you are pricing CPU and RAM, the big three cloud providers are near the top of the pricing continuum. Hetzner is usually at or near the bottom end of the pricing continuum.

If you are including the value-add of other cloud features, then it is not a like-for-like comparison.

EC2, particularly traffic, has massive markups too.

Those consumer PCs don't look too energy-efficient, though.

> The fact that electricity costs is the dominating cost of cloud computing

In the article, they say that electricity has gone from .29 to .45 euros, a 55% increase. If electricity was the dominating cost, seems like they would have raised the price way more than 10%.

Assuming that the 10% price raise is equal to the increase of the electricity cost, that implies that after the increase the cost of the electricity represents slightly more than 25% of the price (which must also include the profit, besides the other costs).

So the electricity is not really dominating the costs, but it might still be the largest component of the cost.

There is an estimate that Amazon has a 61% profit margin for cloud services. Hetzner must have a lower profit margin, but nonetheless the electricity could be as much as half of the costs.

The article didn't seem to say how much they're raising prices, only that they're raising prices due to electricity.

Where are you pulling seeing the price change %? I may need to edit that to marginal, if it's only 10%. :)

> we are forced to increase the prices of many of our products by approximately 10%

First paragraph: "we are forced to increase the prices of many of our products by approximately 10%"

Missed that. Thanks!

The article gives the prices in a table in you scroll down a bit :

electricity/air conditioning rate old price* new price*\

Germany € 0.29/kWh € 0.45/kWh

Finland € 0.12/kWh € 0.22/kWh

That's the price for electricity/air conditioning (presumably for Hetzner), not how much their products will rise in price.

It is the electricity/air rate for their colocation service

It's been like that for a long time (not only for cloud computing, but any big IT infra). Power is everything. To a point, that you measure datacenters in megawatts, when you bring new HW to datacenters you most often talk about improvement in compute per watt, not absolute compute.

It's not only because of the cost of power, but also because power is something that's hard to increase in an existing datacenter - you can put in new servers easily, but putting new power lines and upgrading all the infrastructure is usually just too expensive.

ARM on desktop => ARM for development => ARM on servers => ARM on cloud.

Apple's multi-billion contracts with AWS are ending this or next year, Apple has their own power efficient systems and will compete with AWS, as they have to maintain/increase growth.

> Apple has their own power efficient systems and will compete with AWS

Are you saying Apple is going to roll out an AWS competitor to the public? I find that highly unlikely.

I dunno. I could see it... As long as you access the "iCloud Web Service" from an iDevice or a Mac there will be loads of useful and functional tools and features for you to take advantage of. It'll be 4x more expensive than AWS but the integration with Apple's ecosystem will be fantastic!

If you try to access it from a PC or an Android device though you'll have your bandwidth throttled and all your images will be reduced and compressed like it's 1999.

But you’ll only be allowed to deploy code that’s been reviewed by Apple.

That would be an interesting restriction/value-add. So many things block cloud IPs due to the heavy abuse they get from them. Apple could offer high deliverability IPs and block clearing as services.

This does seem to be the normal course for services companies, and that's where Apple is headed.

They used aws, gcp, azure.

The aws contract will end this or next year, I don’t recall.

They’re offering Xcode cloud build.

It’s only logical for them to also grab a piece of market share of a highly profitable / high margin market.

They’re doing the same with EV, AR, etc.

I am long apple

I wouldn't be too surprised. They already have or are rolling out cloud development environments with XCode cloud.

They don't have anywhere close to the capacity to produce the necessary amount of silicon

What other verticals does it have to seek out next, if growth is demanded by shareholders?

Apple shareholders aren’t demanding growth - they’re doing fine and given Apple’s history it’d make more sense to expand their consumer footprint, especially internationally, since that’s an area where Apple has considerable strength.

Cloud computing is also a hard field to move into: you need a significant amount of capital up front but customers will be slow to migrate. Google is struggling, I’m not sure what pitch Apple would come up with which is obviously better.

There's all kinds of consumer-oriented things they can expand into, like they're currently doing with sports on ATV+. I don't see how competing with AWS could ever work out for them, absent some anti-competitive stuff like allowing additional functionality for apps that use AppleCloud for their backend.

Or people will separate out heavy tasks that aren’t time critical to run them when power is abundant (when there’s wind or sun) and not when it’s scarce.

Some of the hyperscalers area already experimenting with this - https://blog.google/inside-google/infrastructure/data-center...

People do.

https://www.computerweekly.com/news/252481957/Google-creates...

I wouldn't talk about "clouds" as being limited by energy prices, as though they're uniform. Hetzner is much, much cheaper than AWS/Azure/Google. There's a huge amount of fat/gross margin that could be cut from the pricing of the big ones.

> Optimizing for less compute will pay multi-fold dividends over the next few years where we can expect energy prices to continue rising.

I suspect this is already happening in places like Google and paying significant dividends given the scale of code replication.

4) Cloud computing will be outsourced to low energy cost countries.

I wonder if latency allows this

It depends on your requirements. Serving user-facing content is likely to be a task best suited to data centers geographically close to the user base. There are plenty of non-user-facing tasks that could be offloaded to another location.

As an example, the product I am developing spends most of its time calling third party APIs to do bulk data extraction and processing. That can happen anywhere. After data processing and some batch ETL work, we surface that to our end users. The output of that ETL needs to go somewhat close to our users, but everything else could easily happen network-seconds away.

With cloud egress cost though, moving the data to a low-cost region could be expensive itself. I wonder when it would pay off.

I am not shipping more than 20TB/host/month: https://www.hetzner.com/cloud

Also not shipping more than unlimited/host/month: https://www.ovhcloud.com/en/vps/

Start with a cloud provider that is better in terms of cost profile, and then move only the stuff that needs wider distribution to a higher cost geo-distributed provider.

How many data centers already use waste heat for district heating?

gonna go out on a limb and say 0... made me look though and apparently it is starting to happen.

In March (2022), Fortum and Microsoft announced our joint plan for a ground-breaking data centre region in the Helsinki, Finland metropolitan area. The data centres will use 100% clean electricity and the waste heat they generate will be used to heat homes, as well as business and public premises in the area. I must say I am immensely proud of this project – but I believe it’s only the beginning.

4) Cloud computing and the electricity used is impacting housing provision:

https://www.theregister.com/2022/08/01/london_electricity_su...

Reply on news.ycombinator.com