Livepeer Streamflow Transcoding Financial Model

Gleb Dudka and I have been working on a Livepeer Streamflow Transcoding financial model. It models Orchestrator and Transcoder economics. Today we’re making it available to the community.

You can download a copy of it here.

(As the mandatory disclaimer, we’re providing this to the community for informational purposes only. It’s not intended to be investment advice, in any way shape or form.)

Note that there are many assumptions built into the model. Some are Orchestrator-specific assumptions. Others are network-wide assumptions. Pavel Tuchkov from Flussonic supported us in validating some of the latter.

Rather than hold up the release waiting to provide better documentation, we thought it made more sense to release the model as-is. Please review it and ask any questions in the comments :point_down:

We’ll update the model over time, to address the questions that arise. We’ll also do our best to keep the model updated, as network and other parameters may change.

For now, please take a look and let us know if you have any questions. Also, if there’s anything we missed or any assumptions that don’t look right, particularly on the network side, please let us know.

We hope the Livepeer community finds this of benefit :slightly_smiling_face:

Chris Remus, Chainflow

Gleb Dudka, Deutsche Telekom / StakingRewards.com

P.S. - If you’d like to support this work, please delegate to the Chainflow Livepeer Orchestrator.


First published on the Chainflow blog here.

Wow, this is pretty awesome.

In general I think you do a good job of making space to capture all the different assumptions and inputs that someone may want to make. The “Cost of Effort” is certainly one interesting tradeoff…as yes, it is significant incremental effort to add this capability to an operation that doesn’t currently mine, but perhaps it is far less incremental effort to add it for someone who already has many mining rigs deployed. Similarly, the $200 cost to mine, if I’m reading correctly, is assuming mining operations at a loss…where again…someone who is mining profitably may have an advantage here.

One of the biggest variables is obviously the assumptions around how much streaming volume there is in the network. Hopefully operators would invest expecting this to grow over time (and maybe even play a role in creating that growth). One of the other variables that impacts this is pricing. The lower the pricing, the more potential demand there may be - but of course this all requires some scale and elasticity in the market.

In general though I think it’s a very helpful start for someone considering doing this from scratch, and can certainly be iterated upon over time as you learn based off the assumptions.

1 Like

Thanks for the feedback. Yes, the intention here was to provide a starting point, which can then be refined as we learn along the way. We plan to continue iterating on this, viewing it as a living and evolving model.

Agreed, then in this case if a miner doesn’t have an Orchestrator running, they’d have to factor in their “Cost of Effort” to run the Orchestrator.

Yes, the assumptions here are very conservative placeholders, based on current streaming volume. An increase here will change the model considerably, not to mention how attractive the business case is.

Hi Chris, thanks a lot for sharing this. This is great.
A couple of questions:

  1. It looks like the major portion of revenue is because of the inflationary tokens. So what happens when inflation goes to zero by the end of this year? Will transcoders be operating in losses?
  2. I thought the cost of transcoding through Livepeer is 90% cheaper than the Centralized alternatives. But as you mentioned in the cost of pixels, it looks like the cost of centralized services (60,000 wei) is only 20% higher than Livepeer transcoding (50,000) wei?? Could you please explain?
  3. If a transcoder with 2% of overall network stake is earning 1100$ per month from transcoding that means we are expecting a demand of 55,000$ worth of transcoding for one year right? It looks so small. Is Livepeer team seeing any early signs that could increase this demand? How long does Livepeer think it might take to reach 1mn$ transcoding revenue?

Thanks once again for sharing this. Very helpful!

1 Like

Hi Doug, thanks for the feedback, great to hear that the general approach was correct.

With regards to mining and the effort for it, this very personalized assumption and will as you pointed out significantly fluctuate. Mining calculations would heavily depend on the mining operation doing the mining.

With regards to the demand side, the place holder number is pretty low, but one can see significant growth as the demand scales up. Excited to see how this develops!

We will certainly continue to adjust and update the assumptions.

1 Like

Hey, i think i can answer your questions :slight_smile:

  1. Inflationary revenue is higher than the transcoding one as we assumed low demand side for now as the network is still young. Inflation will not go to zero as the target participation rate is 50% and inflation will self-adjust to incentivize that participation.

See primer: “…so in order to hit this target, the protocol incentivizes participation by increasing the inflation rate by 0.0003% for every round the participation rate is below 50% and decreasing it 0.0003% for every round the participation rate is above 50%.”

  1. Indeed the cost of pixel as of now is not that much lower than a centralized alternative. This has to do with a) low demand as as the network is new as well as b) we include the scenario where we account for the efforts of operating the Orchestrator (if you toggle to the scenario without effort, the difference to a centralized alternative will significantly increase). This does however make sense to account for time/effort, however as the network and tools mature, automation will increase and Orchestrators will need less time to be attended to, decreasing the effort.

  2. As we already mentioned we currently assume rather low demand side of an average of 25 simultaneous streams at all times throughout the year. We picked this KPI/measure as the one we can derive further calculations from (as assuming the number of billions of pixels per year, or MB streamed is hard to ballpark and get a feeling of). As demand will go up, the cost per pixel will decrease, making Livepeer more cost-efficient compared to centralized providers. I cannot answer the questions about the Livepeer team’s plans of increasing demand… i personally would have pursued aggressive partnering.

Hope this was helpful :slight_smile:

1 Like

Thanks @glebdudka! Thanks a lot for those answers. I agree with almost everything except your statement that says “Inflation will not go to zero.” I think that is wrong because if the participation rate stays above 50% (which it already is and I think most probably it will remain above 50% for the foreseeable future as people keep staking for either inflationary tokens or for ETH fees generated through transcoding), inflation can indeed go to zero (according to my calculations, most probably by end of this year).

1 Like

Ok, let me be a bit more precise.

Staking LPT entitles you to two revenue streams, ETH- and LPT-denominated. Assuming less than 50% of LPT holders are super long-term believers of the project and thus tend to act rationally/opportunistically when it comes to capturing staking yield, we have to look at their opportunity cost. By this I mean yields captured elsewhere like e.g. staking, lending, even your bank account. Here I take the benchmark of staking rewards on Cosmos and Tezos (pretty established projects, yields in which can be considered relatively low risk) at roughly 8% annualized. So as soon as Rewards(LPT) + Rewards(ETH) < 8%, we will see an outflow of stakers, decreasing participation. When participation goes below 50%, inflation will start to increase, gradually attracting more stakers.

The only case where LPT rewards can go to zero or close to it, is when the Rewards (ETH) >8% in themselves, and this would have to warrant an expremely high demand for transcoding which we are unlikley to see this year for sure. Thus I cannot agree with 0% inflation prediction, definitely not on a time scale of 1-2 years.

1 Like

Got it! Thanks a lot for the explanation @glebdudka. I didn’t think about it in those lines. Now it all makes sense. Thank you!

1 Like

That is an interesting theory - but I don’t agree with your main assumption :slight_smile:

You have to be a long-term believer in every project that you stake with, since currently none of them produces a “real” yield (as far as I know). Every staking yield is based on inflation, so it’s just a matter of “don’t get diluted”. So unless you believe that a project will produce real yield in the future, you would be better of just staking/lending DAI (or USDC/TUSD/USDT) for a ~10% $ yield.

Yes, there will be some outflow of stakers (mainly from staking service providers) but I don’t think that this outflow will bring us below 50%. Due to the price decay and the low liquidity (hard to acquire LPT in the first place), I don’t think we have a lot of non-believers left. Also, many stakers are probably betting on an upward price movement if the inflation is 0 and the Livepeer team continues to deliver.

So I guess we’ll see how it plays out - but I fully expect 0% inflation this year, so way before we will see meaningful ETH rewards from transcoding.

1 Like

Agree with everything you wrote. With current state of affairs (low price, horrible liquidity) you can hardly sell any lpt, let alone the quantity necessary to bring the stake below 50%. This is a long term play, rewards could be great if things work out. If not, I hope nobody invested anything that they can’t live without.

1 Like