Some questions about building a data center

I think you are probably right Dylan.
Shame, as I’ve been talking to CEO of a UK green energy supply company ( with an empty suitable building- 250sqMtrs), with the intention of securing some kind of partnership with them.
I’m thinking perhaps 25 racks 14x nodes to rack.
I can’t work out what minimum investment would be required for this, which is frustrating.
To set up data room inside building with the right power/ fibre/ fire/security I can work out. The rest,…???
I’m sure this information will be available to us soon,… I appreciate that Genesis is the main focus for now.
Still,… for now i’ll just keep on dreaming of my independent data center, dedicated to the Internet Computer.


Thank’s for your messages and help Dylan :cowboy_hat_face:.

1 Like

Hey John!

My colleague Mark is handling that topic and he informed me that we’ve currently got a queue of interested parties but are always looking for more. I’ve DM’d you his email address, so please feel free to reach out to him directly and he’ll help you with next steps!

We’ll be aiming to open this process to the broader public within a few months of launch, but for now we’re still handling each one on a case-by-case basis after direct inquiry.

Sorry for the late response on the topic and for not having much concrete info to share, but thank you for staying active in the thread. This isn’t my area of expertise (app engineer, not ops) but I’m also interested in the topic and happy to track down more info to share.

Looking forward to hearing about a new eco-friendlier DC in the UK soon!


When forum category, Node Providers?


Thank you Taylor…
Thank you Dylan…

1 Like

So the Internet Computer Foundation maintains a list of all the personal contact information of node providers and the data centers the nodes are in? Will this be phased out over time? I can see this being a big issue for adoption and just in general privacy/saftey concerns.


Great question! I want to find out about it more too

me too, I want to find out about it more

Wait what? I thought ICP is a permissionless protocol, even for node prividers, is it not?


It is! Data Centers apply to the NNS to be part of the network. This application turns into a proposal that people can vote on. If people decide the data center is a good fit, the application is granted. Everyone can apply to become a data-center.


Are there any prerequisites for applying for a data center?

Are there any prerequisites for applying for a data center?

If data centers need to apply, and ICP stakers need to vote on whether or not to accept it (aka grant permission), how is this “permissionless”? I would think that the only pre-requisite for a data center being integrated into the IC is a technical requirement: stability/uptime, hardware, etc., which can be determined in an algorithmic trustless way, and not something that would require staker consensus.

Seeing as the majority of ICP will be controlled by Dfinity / Dfinity’s investors, and thus the voting power index for Dfinity/Investors would be much higher than the average user, having a staker vote to integrate a new data center is essentially centralizing this entire process at launch.

Why does integrating a data center require a staker vote to pass?


That is the definition of a permissioned protocol. I was extremely disappointed by the NNS documentation Dfinity released today. For months we have been hearing that the NNS would approve nodes algorithmically only to learn that it will actually be done with a token holder vote. Not only that but it’s not possible to host a node anonymously, node operators must provide a legal name. I don’t have a problem with permissioned or centralized systems existing (although I am now questioning whether I want to deploy my own projects on the IC), but Dfinity’s marketing is proving to be extremely misleading and they should be upfront and honest about the IC being a permissioned system. There is no reason why node operators should need token holder approval. Token holder voting should be a system of last resort, it should pretty much be reserved approving changes to the protocol at most.


I don’t see the point, what’s the value proposition of requiring approval from stake holders to join? Can only see this would potentially open up an arena for money/politics under the name of democracy. I hope you guy has strong reasons to justify this design decision.


The tokens the foundation holds are not participating in governance afaik.

How would you algorithmically determine wether data centers applying really are independent? How would you determine they’re not AWS machines? (Asking out of curiosity, not saying you’re wrong)

I can help clarify some points by actually going broad: what is the goal?

To us, the goal is building a tamperproof system where no one player or group of players can collude to mess with the system (e.g. make 2 + 2 = 5). This includes making sure that each player can not stop or slow the system either. Part of this goal is to also have this tamperproof platform be “web speed” and scalable. We want a platform that grows with more machines (throwing more machines in most protocols do not increase scale).

In studying many cases, it became clear we had to do a few things:

  1. Provide a great experience for developers and their users -
  • most blockchain protocols, machines “mine” and if they get lucky they get a reward. This incentivizes miners from having branches that eventually converge. They are analogous to “bounty hunters” who get paid by the job and can turn off their machines if a better protocol comes along… even hourly. In the IC, all infrastructure providers get paid a steady stream. Nodes do not compete with each other. All nodes have to do is behave well and NOT get removed out of the network. If there is low demand on a Saturday, the node still gets paid to be part of the subnet (so it’s available for a developer). Remember, the IC is meant to have beefy, world-class professional setups, not people in a garage. The NNS detects bad or faulty ones. If a node is off a few hours per day or week or is routinely slow, the NNS will remove it. This helps maintain a great experience for the developers who now have a platform that is fast, reliable, stable.
  • Because every node gets paid by minting ICP, every node creates inflation in the network. The NNS tries to maintain a healthy balance of nodes being paid so that it has what the network needs, not too much, not too little. If it needs more nodes, it can raise rewards.
  • The IC is unbounded so it can add as many nodes as NNS approves
  • Better developer experience like this will lead to a better consumer experience (better apps built)… which leads to more adoption. Remember, there are only a few thousand apps in production using blockchain as backend… and none of them using the ENTIRE app on the blockchain.
  1. Practical vs Theoretical independence - it is important that the nodes be independent. we found that permissionless platforms pretty much led to the following cases: a few big players ruled blockchain infrastructure. This is why transparent (not anonymous) identities are important. We don’t want Jeff Bezos controlling 51% of the network. The NNS tries to maintain a healthy balance of independent parties, locations, etc…

  2. Channeling The community - In our experience, there are 100x-1000x more community members and token holders than there are infrastructure providers. We want to empower these people to have a say in the direction and keep the infrastructure side focused on high-performance, reliability, and following the protocol. The community gets paid in rewards for their votes. This is a much much lower barrier to entry for getting rewards for being a good community member. Also remember… if the community wants to change the current rules, they can do that too.

  3. Focusing on tamperproof - we found quite a number of attacks that users can do in a network without an NNS or identities.

  4. Teasing apart decentralization - Notice we teased apart what people usually mean when they say they want decentralization. They usually mean two things:

  • Their apps are living in an environment that no one can tamper with (no 2 + 2 = 5)
  • There is a way for people to take part in the growth of the network (getting rewards by voting not doing puzzles)
  1. "Does this mean I will never be able to connect my machine as infrastructure?" - no. We have been playing and studying some ideas where we could use lower-end machines. But we want to get a good core of data centers and professional-grade setup first.

  2. Much more energy efficient - by having machines do the actual computation for apps, that means they can spend energy on computation that matters. This is something that played a role in designing the system as well


The problem is Dfinity consistently published information saying NNS would approve data centers algorithmically and that canisters could be set as their own owner, becoming “unstoppable”. Today we very suddenly learned both of these are false. Data center and node approval are decided by ICP holders and canisters can be frozen (i.e. stopped) by ICP holders. Some people want to use Binance Chain and Ripple, others want to use Ethereum and Solana based on what they consider important in a network. I don’t think anyone is upset with the decisions you made, I am personally upset that we were mislead.

As to some of the specifics:

Jeff Bezos could just buy a bunch of ICP.

You need peoples legal national identities to tamperproof the system? Is your method of tamperproofing calling the police on a malicious hosts?

ICP holders != the community. As it stands now Dfinity, Dfinity team, and big VC firms hold almost all the ICP. It will probably be years before any coalition of others can accumulate enough ICP to vote to do something against the will of Dfinity and the big VCs and by that time the precedent of a heavily permissioned model will be set and changing it will be extremely difficult. I think it makes sense to relinquish this control slowly but let’s just be realistic about what could potentially be done and in what timeline.


by the way… I don’t think it’s an easy answer: it’s perfectly fine to accept the risk of the “Jeff bezos” or “three mining companies control 51%.” For some protocol designs, it made sense for them. It did not for our vision of IC, that’s all

1 Like

Again, let’s use practical vs theoretical.

Ethereum has 8000 full nodes.

What do you think is cheaper: buying 8000 servers and running them or buying 51% of a network? We are talking orders of magnitude.

These are the practical trade offs to consider.