I need help with cycle burn rate math - ICP deflation

General
1

So there’s some discussions about the cycle burn rate and making ICP deflationary. It was brought to my attention that BOB doing a 10x against ICP would make the entirety of ICP deflationary. However someone else said that even a 5x would do it. I don’t like this at all though, I need numbers and math. As of yesterday evening BOB is burning roughly 90% of all cycles on the IC. I can’t find anything about a direct math coeficient for cycles to icp. Also has anyone approximated the exact amount of ICP created per day, so that we can determine what the price of BOB would actually need to be in order for it to be fiscally logical for people to burn ICP at a rate higher than the inflation rate and thus make us deflationary?

If anything I simply need clarity on the relationship of ICP to cycles, but if you could solve that math, it would help in the marketing of BOB, and the inevitable deflation of ICP which would obviously make ICP itself moon and we all want that.

2

I can help with that, the rest I’ll leave to people that like to talk about tokenomics :slightly_smiling_face:

Here is a high-level page about cycles. Long story short: 1T cycles cost you 1 XDR, which is ~1.3 USD at the moment. Depending on the ICP price you get a varying amount of cycles for 1 ICP. The current conversion rate is visible here if you scroll down to the ‘Conversion Rate’ panel.

3

This article gives an overview on the overall mechanics of supply & demand; it also provides a projection analysis on the development of total supply of ICP.

1 Like
4

So does this mean that if ICP does a 5x up in value that the burn rate would go down 5x since it’s that much cheaper to get 1XDR?

5

In ICP terms yes but not dollar.

6

That’s really bad, that means that the higher the price of ICP, the slower the burn of ICP. It’s an interesting paradox!

7

Yes it’s a tradeoff. In return for that you get predicable prices for running an application, which has quite some value IMO

2 Likes
8

Hi bjoernek - thanks for the clear, insightful posts on ICP tokenomics and forecast modelling. Since your May 2023 article (you referred to above), the number of cycles on the IC has increased dramatically. Looking at the node remuneration scheme (Node Provider Remuneration - Internet Computer Wiki) and estimating cycles burned per month at an average of ~2000T/month per node, it seems to me that we are rapidly approaching the point at which the ICP burned equals the ICP minted to remunerate node operators (both costs set in XDR). It would be great if you could comment on this or perhaps provide an update on the article to benefit ICP holders. I also note that the ICP price recently tracked relatively close to your forecast (based on alpha of 0.2 relevant to the increase in cycles burned). From some other Dfinity articles and replies to posts I understand that the cycle burning capacity of the IC with the existing nodes is still much greater than that currently achieved, which suggests that the IC should soon make progress in catching the inflation associated with voting rewards. It’d be great if you could confirm the above and provide a general update on the state of inflation. Many thanks in advance!

2 Likes
9

Yes - but note that a higher ICP price reduces the cost of rewarding node providers as well. An interesting ‘rule of thumb’ for interests sake is the following: should ICP hit US
$80 then compute on the IC (number of cycles burned) requires to be ~100x more than it is today to become fully deflationary. That’s feasible but not in the short term.

10

Hi Arthur2 - I’ve done a quick and dirty assessment of the state of ICP inflation/deflation. Below the key findings, but there’s a bit more detail in my Nuance post here: https://nuance.xyz/fractaleer/12095-434go-diaaa-aaaaf-qakwq-cai/icp-tokenomics-part-i and in a follow up post next week.

Like you, I’m interested in figuring out how large the IC needs to be (in terms of numbers of nodes and cycles burned) for revenue (ICP burned) to exceed ICP minted for node providers + staking/voting rewards. To make it easier to understand I’m breaking it down in two questions.

The first of these is relatively easy to answer, because the value of the ICP token has very little influence on the outcome. Question 1: When does revenue (ICP burned) exceed operational cost (defined as ICP minted to reward node providers)? This question is relatively easy to answer because the price for cycles and the reward for node providers is defined in XDR. The ICP dashboard provides the data from which you can calculate ICP burned for cycles/month (revenue) and the ICP minted for node providers/month (operational cost). If you do that for 2024, you end up with the following figures:

ICP table
ICP table408×255 6.64 KB

To cancel out the noise created by the fluctuating ICP/XDR exchange rate, I calculated the ratio of ‘ICP burned/ICP minted for node providers’. Once the ratio is equal to 1 then we have reached the point at which the compute on the IC (and the ICP burned) ‘covers’ the operational cost of the network (the ICP minted to reward node providers). I’ve done this only for 2024 as there was no point of going further back (as you’ll see in a minute).

Up to the end of August there was so little compute on the IC that the ratio rarely exceeded 0.01, which means the revenue (ICP burned) was up to two orders of magnitude too low to offset a significant part of the cost of the system (ICP minted to reward node providers). But the situation changed in September 2024. The compute associated with Bob.fun and presumably also some other dapps released since, has kickstarted the compute on the IC. In December 2024, the ratio was 0.63 meaning that revenue covered 63% of the operational costs of running the network of nodes! The following graph visualises the trend.

ICP Deflation Step 1
ICP Deflation Step 11920×964 94.4 KB

Clearly we’re constrained by a small sample size, but I wouldn’t be surprised to see ICP burned exceeding ICP minted for node providers within the next 3 months. That answers Question 1.

The second question is much trickier to answer, because the value of the ICP token has a massive impact on the outcome. Question 2: When can we expect compute to outweigh the minting for all rewards (node provider and staking/voting rewards)? This is the point at which deflation is achieved and the total ICP supply will start to shrink. I did a little preliminary calculation how much compute (i.e. how many cycles) it would take to offset the 6.48% voting rewards associated with a total current supply of 530M ICP tokens and with an ICP token value of $US80. The voting rewards result in the minting of ~2,862,000 new ICP tokens/month. To offset that when the ICP token price is US$80 would take a compute equal to ~170,600,000 trillion (T) cycles/month! To put that number in perspective, I estimate the IC currently runs ~2,000,000 T cycles/month, so that’s two orders of magnitude less than required. In other words, compute on the IC would have to increase by 100X to make a serious dent in offsetting voting rewards. This is entirely feasible, but not in the short term and the picture is of course a lot more complicated because we’d have to calculate how many nodes it would take to expand compute to that degree.

I’d love to do some more work on this, but we need to hear from Dfinity with an estimate of how much compute capacity the nodes have on average in order to make more accurate predictions.

Hope this is helpful

11

1 XDR = 1T cycles
USD/XDR exchange rates fluctuates around 0.75
ICP/USD exchange rate fluctuates pretty wildly as you know

I tend to use an ICP value of US$80 for medium term forecasting purposes (which equates to a market cap for ICP of ~40B)