Their scale is simply too small to affect the market outside China, majority of their chips will be eaten up by HBM3 production with yet unknown yield rate.

They are forbidden to buy foreign equipment beyond their current process node, which is already obsolete, die size is 40% bigger than Samsung, not to mention lithography, the big 3 are using EUV while they are stuck with lobotomized DUV.

They can start making some decent money now, but vastly expanding capacity as is means enormous losses if the cycle went downward a few years later, that's how all previous makers went bankrupt.

They can squeeze out a bit more performance if they are ready to go beyond their current node using only domestic equipment and be blacklisted by the US government.

But the cap is there, unless they can make a working EUV machine in 5 years, they are doomed to be a minor player, if the current cycle even lasts that long.

They will first fill the local demand for all their electronics manufacturing. Then their massive computer infra and AI. And if any is left, it will be bundled to local PC exporters like Lenovo.

I am very hopeful of CXMT. But then then it could take a while for them to ramp up production. Maybe by then, the AI bubble would've burst.

One problem with US sanctions is it could hurt US companies too, like in the case of cutting-edge EUV and CXMT. This is when China is actually a hero and not a villain.

China also needs RAM for AIs, especially since they have plenty of electrical power and building speed to pump out data-centers.

That's probably what is going to happen, it's a strategic opportunity for the Chinese government here, there's a big market demand that can fuel their domestic production capabilities that nobody wants to take.

I previously ran another VPS host who did the same exact thing (before OVH did it?).

Unlike Fourplex.net which uses modern ASRock Ryzen 9000 servers, Qeru.net used older HPE DL360 Gen9 servers.

I gave 3GB of RAM for $3-4/mo then. But these servers weren't very fast. I ended up selling the business, and am happy I did.

For providers like us, we have to lease IPv4. We came long after IPv4 was already depleted. IPv4 prices did go down. Despite that, the $15/year 128MB BuyVM plan is long-gone.

But for a new provider like us, we'd have to spend more than an established player like BuyVM or RackNerd who bought most of their servers pre-AI-boom.

Exactly this. The old hardware from a year ago is fine as the typical use-case for a VPS didn't change.

So with 8 customers per box (AMD EPYC™ 9645 CPU has 96 cores) if they have single-cpu boxes, that would need 256 GiB RAM.

CPU launch price 11000 USD. RAM will likely be another 10000 USD

20000 / 8 customers / 40 USD/mo = 62 months just to recoup CPU and RAM let alone other components.

Weird, whenever I napkin math offers of any HW for renting, I get that I could buy it myself in 1-2 years of rent. Sometimes faster.

Do they not intend to recoup the costs of HW? :)

Until recently I had a 4gb ram 80gb ssd+2tb hd VPS running debian in a Montreal data centre with a real use 700 mbit pipe to my city with a budget provider for the equivalent of $80USD/year. When fio speeds were slow they moved me to a less crowded server. I gave it up as don't need it and moved my personal sites back to NFS for peanuts a year and services to my NAS. The pricing, offsite storage for my backups, Canadian sovereignty, lack of perceived complexity with a big provider was all attractive. I'm a physician with a tech hobby and last serious tech work was in the LAMP days with perl and php. Trying to think of learning about AWS and screwing up usage based billing was daunting!

Much better prices, and simplicity. The power you get from Hetzner or Kimsufi is crazy compared to AWS.

If I need to host something small, I don’t want to mess around with the many permissions and quirks that are required to deal with AWS. It is often much easier to just setup the server on a standalone service.

Most of my customers (small VPS host here) don't like the companies behind AWS, Azure, and Google Cloud, especially the amount of influence they have in the world and how they wield it. And the pricing often isn't that much different between a small VPS host and either a cloud provider or one of the larger VPS providers (Akamai/Linode, Digital Ocean, etc.) - larger providers have eoncomies of scale, but smaller providers don't have as much overhead for paying sales and C-suite.

There's also the human touch in terms of who you talk to: a lot of the smallest VPS hosts are 1-2 people, both technical, so customer support = sysadmin = contact for everything.

I could pay like 30 bucks a month for an absolutely overspecc'd VPS (64GB/16c) that would cost around 20X on AWS (According to ChatGPT; which sounds about right based on the last time I cared to even look into it)

Does it have a billion 9's of reliability? No, but I don't care, it has literally never not worked when I've used it

Customer Service so far has been human, but that will vary greatly for the provider

I also use a different provider for work related hosting, and the reduced latency of being within 20 ms of the DC has been probably the single biggest (perceived) perf improvement my users have ever seen, specially on the legacy webforms platform we recently decomissioned (We're a bit too geographically far for most Datacenters of most large providers)

I'd use digital ocean over AWS for any SMB or lean startup (so... anyone not attached to an infinite money hose that has to either scale to NEED AWS, or die trying) just because of 1) their UI not being broken glass you have to crawl over and 2) not having eight trillion features that make doing simple things hard and 3) pricing

If the only thing you need is "server, accessible via the internet, always online", and you're not interested in all the vendor lock-in masquerading as useful services offered by the big cloud providers, then small VPS hosts are 100% the way to go. For mid-sized servers they're cheaper (i.e., stuff that wouldn't be free on the big clouds, but not "I want a petaflop"), with more transparent pricing (I pay $12, every month. If I get inundated with traffic, I'll get cut off until I choose to pay more).

I used a vps service hosted in a country with strong digital privacy laws to host a personal wireguard+pihole vpn. I could probably think up a decent argument why that privacy with the smaller guy was only nominal but I could absolutely think up a good argument why doing that on a big name would have no privacy guarantees at all, especially as someone who would be in the bottom rung payment-wise.

Never had problems with downtime and I payed, like, 40 bucks a year over 3 years. I think I had to restart the thing once because of something dumb I did on my end.

The only advantage is cheapness, for personal use.

If you’re a government agency or a company you don’t care about saving $14/month, you want a secure provider. And these hosts are not secure, you’re basically just on your own.

Price. 1 vCore, 2GB RAM, 20GB SSD, unlimited traffic (though throttled to 200mbit/s after transferring 2TB within 24 hours) = 1.85€

That is a nice way to have a static IP on the internet and enough resources to do small things like host a nameserver and/or OpenVPN/Wireguard.

I may have had 4 hours of downtime in one year, always announced days in advance.

Spending caps is the biggest reason for me. Granted, some VPS don't offer this (vital!) feature, but none of the big 3 or similar services do.

Low cost, simplicity and customer service.

AWS does offer Lightsail which is similar pricing.

Small VPS hosts oversell like crazy and they offer much lower prices. Also their reliability might be worse, because they don't migrate VM between hosts.

10-100x cheaper

Others have mentioned the general pricing, simplicity etc.

Outbound data pricing is a potentially huge saving.

AWS is as much as $90/TB outbound with 1GB free. Hetzner is $1.20/TB (in EU and US) with 1TB/20TB (US/EU) free.

(Good) Smaller places are more likely to have actual technical staff you can talk to.

Predictable and extremely low costs for less critical stuff. My 2 main ones are respectively around 4 and 8 EUR per _year_.

I use them to run wireguard to evade geoblocks when I'm travelling, a few redundant monitoring scripts alerting me of reachability issues of more critical stuff I care about, they serve as contingency access channels to my home (and home assistant) if my primary channels are down.

I get no support, no updates, it's all on me - which is fine, it allows me to stay current and not lose hands-on practice on skills which I anyway need for my job (and which are anyway my passion). I don't even get an entire IPv4 - I get.... 1/3000th of it? (21 ports, the rest are forwarded to other customers). Suits me fine.

Where small VPS hosts can make a difference: require no KYC, accept crypto payments.

Simplicity, price, stability.

Not being US owned can be an important one in this geopolitical climate.

I like Vultr for the simplicity of my own projects. I really hate spending my time on provisioning and similar labyrinths.

Simplicity, and low price.

VPS services are usually really, really simple and fairly cheap.

I'd say that actually VPS prices is where we actually see computing prices going down rather than on the big 3.

AWS used to optimize further and pass down the savings to the customers back in the day, now they don't do it anymore.

Pricing. They overprovision aggressively but most people actually just need a 0.1 CPU available remotely for the majority of their use cases.

I replaced with a home server and it costs way more just in power hahaha.

Pricing (both cheaper and more predictable), and reduced complexity.

Not being Amazon, Microsoft and Google.

I moved from AWS to Hetzner, because: 1. lower prices, 2. not American

Yes, even if prices would triple, VPSs would still be an attractive offering.

At a certain price people choose not to buy

Ideally all cloud applications hostable on any platform would just provide the following services to clients:

A. rendezvous services so clients can connect to one another,

B. storage/retrieval of encrypted data where the host does not have the key to decrypt,

C. transport of encrypted data which cannot be known by the host due to B above.

> How can you trust him to have redundant networks

You can't, so abstract that away at the application layer. Make it not dependent on a single host or network.

You definitely can use Linux with few simple servers with 128 MB RAM.

Install can be tricky indeed, but if you have installed system, it's easier.

My NixOS SSH jump host server here idles at 234 MB of which 64 MB is systemd-journald (which I assume can be reduced with some settings of how much to keep in RAM).