CIP 207: CityDAO Compute - Build a GPU Cluster in Wyoming

I propose we start the world’s first crowdfunded compute cluster in Wyoming to take part in the AI revolution and earn income for the DAO.

TLDR

It’s time to do something big to bring CityDAO back to the forefront of the conversation and bring new life and energy to the DAO, and we are uniquely positioned to pull off building a large cluster of GPUs in Wyoming.

We have capital, land, and a Wyoming-based legal entity. And at 6 cents per kWh, Wyoming has just about the cheapest electricity in the states, making it the perfect place to do this.

Budget:

Total proposed budget: $600,000 USD

$500,000 - Purchase 300 Nvidia GeForce RTX 4090s

$70,000 - Servers, equipment, other setup costs

$30,000 - Legal and misc expenses

Business Model

I believe this could generate about $50k / mo for the DAO treasury. Here is the full breakdown of the unit economics.

Upon passage of this proposal, Scott will lead implementation of the effort, setting up a fund with CityDAO LLC as the main limited partner and Scott (and potentially others at his discretion) as a General Partner. The DAO will invest $1m in exchange for LP shares in the new fund, and be paid quarterly dividends from the operation of the GPUs. The fund will have a standard 2/20 management fee model. The GPUs will earn revenue by being put on Runpod or a similar service.

Benefits to CityDAO

  • A consistent income source for the treasury. These GPUs have an estimated payback period of 12 months and after that generate mainly profit.
  • Status and reputation - the first cluster will be called the CityDAO Compute Cluster, and as the first crowdfunded compute cluster will be a hit on Twitter and a real contribution to the AI and crypto communities.
  • Compute - having a bunch of compute is cool - we can let citizens use it at a discount and even train our own models on it.

Why should we do this?

  • Timing - This is the perfect time to start the world’s first crowdfunded GPU cluster.
  • New energy and purpose - CityDAO needs an ambitious new project to rally around. The DAO can build tooling and projects around the cluster. Maybe we can all go to Wyoming and visit!
  • Wyoming connections - We are uniquely already set up to operate in WY and I recently got connected with a Wyoming-based datacenter with just about the cheapest electricity anywhere - 6 cents per kWh that set up all the infrastructure to run GPUs.
  • Future of AI - Demand for compute is projected to increase a lot over the next decade.

Let me know your thoughts, and hit the heart if you want to see this happen!

Updates - 2/13/24

  • I’m lowering the ask to $600k - still enough to start up the cluster and run the experiment and a much smaller percent of the DAO treasury
  • At the time the cluster goes live, Citizens will each get a free CityDAO Compute NFT to commemorate the moment.
24 Likes

Bold idea. Capitalizing on the lower energy costs in Wyoming to get some passive income makes a lot of sense.

A very simple and naive question (I really don’t know much about this), but why AI? The little I know is AI is pretty geographically concentrated, and this will likely continue. Essentially, why would they want GPUs in Wyoming? I mean, I get the low energy cost, but I imagine there are other costs. I would guess that with AI, speed is everything, not so much energy costs.

I have always thought mining Bitcoin with wind makes too much sense. There must be some BTC operations in the state I would imagine. I guess you could do whatever with the GPUs, first step is just getting them.

3 Likes

I like the concept. It’s cool for CityDAO to be aligned with modern innovations.

I had three questions, since this isn’t my field and I don’t know much about it - first, will Nvidia GeForce RTX 4090’s be “good” for a long time? I don’t so much mean the depreciation, but the actual technology. How do we know that they won’t be replaced by newer tech within a year or two? Second, regarding runpod, how do we know the margins will continue to be so high? Third, you mentioned that we have land in Wyoming. Are you suggesting we put this on Parcel 0? What about the wind?

2 Likes

For the record for this to work power would need to be no higher than 8c a KW. The halving will make this worse.

There is a huge cloud compute market. It isn’t geographically constrained (In the implied sense).

Powerful internet is a pre-requisite but after that there is plenty of demand for providing/renting gpu compute, which will only continue to grow.

BTC mining also makes sense imo, and could diversify the risk.

I personally like the idea of self-sustained power like solar. Not sure what the payback period would be on this though. Not a solar expert.

I suspect GPU tech will only accelerate in the coming years with the current advances in AI/ML.

Unfortunately its almost impossible to estimate (without being deep in the industry) whether or not the will be “good for a long time”. Typically GPUs will last for a very long time, companies like Runpod have support for 3070s ( release: September 17, 2020) With there of course being use for even older models ( while sacrificing compute power and efficiency).

There is research for optimizing gpu clusters with specific models built with the highest efficiency ratios and that should be considered when choosing. I have not done the research yet and dont know if the 4090 is the most optimal but i suspect Scott will have more context on his choice. (Perhaps production and availability)

The question for me is, what is the average cost to other BTC miners? As long as you are lower than the average you will make money. The highest-cost miners will stop mining with each halving.

Anyway, I do like the idea, at least some of the GPU to BTC. As Lyons mentioned good to diversify regardless. Put our GPU eggs in a few baskets.

1 Like

I believe the upfront budget for this is insane when compared to other things CityDAO could do with its treasury and assets to generate a similar monthly return for a fraction of the cost

1 Like

From a marketing perspective, I think there’s value to it being AI + crypto, regardless of what it actually does, and could generate some press, etc. Would love to see some ideas to capitalize on this and either bring more people into the project or show value for existing citizens.

Additionally, would love to see this tie into parcel 0 somehow and breathe some life into that project and some meaning into holding parcel 0 NFTs.

Thanks for the good questions…some thoughts:

On the one about if Wyoming is too far away for AI companies predominantly in California or NY - one convenient thing for us is upload / download to the datacenter doesn’t matter too much since companies can transfer their data to a server in Wyoming datacenter easily. The important thing is having a bunch of GPUs concentrated in a single area where they can be interconnected so that you can train something large. So that means energy cost does become a really important factor.

Another great thing about Wyoming - lots of folks set up bitcoin mining farms there, and after crypto prices fell, people are looking to diversify. That means there is a lot of data center infrastructure like power and ethernet already built out.

4 Likes

Good Qs, here are my thoughts:

1. Better hardware

Nvidia will always be putting out better hardware, but from what I’ve heard, their new GPUs are sold out until 2026 or something, which means there is a lot of time for us to recoup initial costs. My ideal set up would be that we reinvest some of the profits from the GPUs into newer and better machines so that we get exposure to the newest hardware and get to diversify our holdings.

2. Runpod margins

Runpod is a great starting point that will make it easy for us to monetize the GPUs right out of the gate but in phase 2 we will sell directly to companies and have higher margins. But overall it’s important to realize there is definitely some risk to this project like in anything - demand for compute could crash like crypto did last year. But there is still a lot of intrinsic value to owning GPUs.

3. Connection with Parcel 0

Realistically, we will need to set up shop in a data center that is already equipped with heating, cooling, power, and fast ethernet. I would love to connect it to Parcel 0 somehow though, if anyone has ideas.

2 Likes

Explain? For crypto mining, maybe, but for training AI models the numbers are much better

2 Likes

Thank for the right idea.
CityDao needed a business model to create a revenue stream that Is lacked until now. With this proposal, it is the first time that a meaningful plan is presented.
I don’t have enough technical knowledge to talk about pros/cons of this plan, but I am sure even if it is not the best solution it is a big step in the right direction.
We can start with AI then use the infrastructure for other services like node hosting, cloud storage, mining, staking …

The most important thing is that with this proposal, the DAO finds its long-term strategy that fits its current situation and has the potential to generate enough revenue streams to ensure its survival or future success.

3 Likes

It has become NVIDIA’s fourth most valuable public company worldwide. It surpassed Amazon and Google.

AI chip shortages are an incredibly big problem. Sam Altman is looking to raise up to $7 trillion from investors for a new venture that will focus on chips specifically designed for artificial intelligence. An incredible number.

If we look at these developments, it is quite possible to say that artificial intelligence technologies can have an impact on cryptocurrency markets, especially AI and DePIN.

Beautiful and exciting days are waiting for us.

The idea is nice and innovative, but I think it would be better to conduct a preliminary study on its feasibility and applicability, which will last at least 6 months. You go slowly on the road you don’t know, you go fast on the road you know. $1 million seems like a lot to me. 500k should be the maximum.

1 Like
  • I like the idea.
  • I agree with ScottA in putting GPU eggs in a few baskets, we might/should think of mining.
  • Agree with Steppenwolf in using the infrastructure for other services -if we can-. There should be some positive outcomes for the CityDAO citizens that increases their participation & engaging with CityDAO. By this way they start believing us again we are in the right way to build network cities.
  • I’d like to hear more about
    ++ cons of this plan just because of I have no technical knowledge.
    ++ how we are going to/should use that income ?
1 Like

Scott,

I appreciate your effort, and I know you’re trying to profit from this project, as with any projects you’ve offered before. The idea is good, and you wrote all the details with fancy words, but what will the holders benefit from it? Why should I approve of spending so much money? Will there be any profit distribution? Will the projects using this GPU cluster give any WL to citizens, etc.?

The most important thing that CityDAO needs is to transfer the benefits and profits obtained from the “projects” to the holders/citizens rather than making “projects.”

I still can’t find answers.

Why am I a citizen?
Am I holding NFTs so a bunch of people can make money named “projects?”
Why would other people want to become citizens?

Unless these questions are answered or resolved, I “personally” will never approve such a large expenditure.

If it cannot be done due to legal problems, let’s move the company (LLC) to a place where it is possible to do this. (England, etc. and yes, it’s legal there!) We can’t tweet anything yet because we might face legal problems. It’s really funny.

Apart from all of these, I am involved in many of the NFT projects. However, I have never seen a place that does not contribute any benefits to its holders as much as this one.

7 Likes

Glad you are supportive. I think you are right that we can go more slowly down the road we don’t know, which is why I’m lowering the ask from $1m to $600k.

Two updates:

  • I’m lowering the ask to $600k - still enough to start up the cluster and run the experiment and a much smaller percent of the DAO treasury
  • At the time the cluster goes live, Citizens will each get a free CityDAO Compute NFT to commemorate the moment.
2 Likes

Just wow, you were able to reduce the budget this much just because of someone’s objection. If a few more people object, I think you’ll be able to get this done for $2K.

Your calculation and prediction skills are truly amazing! And commemorating NFT is a fantastic idea! lol

Putting all this satire aside, I don’t think CityDAO should be the place to run your “expensive experiments.” You were able to reduce almost half of the budget you originally wanted, and I don’t understand how. What kind of project preparation method is this? It seems that there is no guarantee that we will not encounter any calculation or forecasting mistakes you make somewhere in the future. So, I will reject this proposal with peace of mind. Thank you!

Thank you for taking the warnings into account, the idea is good but I feel missing some points.

Community Participation: Frankly, I couldn’t find clear information in this CIP about how the CityDAO community will participate or benefit from this project. I would like you to provide more details on how community members will be involved and benefit from the project.

1 Like