Can the IC become deflationary?

Problem Statement

Recent discussions about NNS treasury, APY changes, network expansion and storage costs had me thinking about the current tokenomics and whether the IC has potential to become deflationary at some point in the future.

Why do I think deflation is an important aspect to consider?
It’s safe to assume a majority of ICP holders and stakers invested in the project not only cause they believe in the idea of a decentralized internet but mainly for monetary gain, the IC’s security model and liveness are also based on the assumption the token has at the very least a stable or slightly bullish price action, if that weren’t the case:

  • the NNS would break cause the stakers’ locked value depreciates no matter what, so stakers are no longer incentivized to act in the network’s best long term interest but to extract as much value as possible in the short term.
  • A depreciating and inflationary token would benefit late adopters in the NNS, who will be able to buy Voting Power at a cheaper prices.
  • A sustained downward trend would lead the ICP in a death spiral due to how node rewards are minted, if such a scenario were to happen the network might not be sustainable anymore.

The ICP has 2 types of inflation: Governance rewards and Node rewards, so let’s see how they work and whether it’s possible to eventually achieve deflation.

Governance Rewards

Governance rewards are calculated by a % based on a descending curve applied on total supply.
For simplicity’s sake I’ve considered the “best case scenario” (no node rewards, low price, inflation at end of curve(5%) calculated on genesis supply, not current or future), that’d be:
20.000.000 ICP yearly inflation (400 mil at genesis @ 5% APY)
1 ICP = 4$ = 4,8 XDR
1XDR = 1T Cycles

If ICP price were 4$ that’s 96.000.000 XDR worth of cycles to be burned yearly to counter governance inflation or 96.000.000T cycles/y, in the last 7 days 30.000.000.000 cycle/s was the top according to the dashboard, so 933.120T cycles/y (30.000.000.000 x 60(m) x 60(h) x 24(d) x 30(M) x 12(y))

Even in this optimistic scenario burn rate should more than 100x to counter inflation, it gets even worse as price increases, cause number of tokens minted stays the same but amount of ICP burnt per XDR is reduced. So is deflation possible?
Well it depends on how much of the current total network capacity (or CTNC) 30 bil cycles/s is: 1%, 2%, 50%? I honestly don’t know, BUT unless it were a considerably high percentage, e.g 90%, there might still be hope cause the % is based on the current capacity, so while we can’t use more than 100% of it, the IC itself can become bigger by adding more subnets and by doing so increase the max theoretical cycle burn rate.

Node Rewards

More subnets equal more nodes, which result in more minted tokens to pay providers, it is therefore fundamental to figure out the factor by which cycle burn increases as the network grows. Deflation is only possible if:

  1. Subnet avg. cycle burn / (1T * XDR/ICP avg.) > Subnet rewards
  2. Subnet’s burn surplus (difference between avg. cycle burn and subnet rewards) * n of subnets (n being a realistic number) >= Issuance

The providers’ rewards vary based on country and in future based on specs too, for simplicity’s sake let’s assume all nodes are paid the lowest rate: 873 XDR/month, a subnet has 13 nodes in it atm, that’s 136.188 XDR/y worth of ICP per subnet or 4.766.580 XDR for the entire network (136.188XDR x 35 subnets) or 4.766.580T cycles/y.

Taking into account the registered spike of 933.120T cycles/y we find out the cycle burn rate must 5x just to offset the node rewards minting, I’m not sure whether that is doable, it depends on how much of the CTNC 933.120T cycles/y is, if the percentage is too high and a 5x isn’t possible, e.g > 20%, then increasing node count will only make things worse cause even if the network were used at 100% 24/7, it still wouldn’t be enough to burn more tokens than minted to pay for the subnets.
In such a scenario the only way to make deflation possible would be to increase cycles cost per operation without penalizing devs to harshly, as that might cause cycle burn rate to drop.

I’ll assume the 5x is possible because it’s coincidentally pretty close to the same estimate @Kyle_Langham gave a while ago and I trust in his verdict, still more analysis should be done to figure out the cycle burn to mint ratio per subnet with realistic usage metrics, keeping that ratio higher than 1 and as high as possible without impacting operational costs for devs is crucial to the network’s profitability.
I’d also like to mention a scenario related to how storage costs are currently calculated which might heavily impact cycle burn to mint ratio per subnet for the worse, for brevity’s sake here is a link: Question regarding RE EXC-1168: add non-subsidised storage cost on 20+ node subnets (behind the flag) - #17 by Zane

Mind you NNS inflation is fixed regardless of token price, but ICP burnt per XDR decreases as token price goes up, therefore even if the IC burnt more than it minted for node rewards, the network’s usage and capacity will have to increase alongside price to offset the governance rewards and keep deflation going.
I foresee even with a good cycle burn to mint ratio per subnet this is what might happen:

  • More tokens are burnt than minted → price increases
  • Price increases → less tokens are burnt
  • Network growth can’t keep up with price growth → More tokens are minted than burnt → Price decreases
  • ICP price stays at range in which total supply doesn’t significantly increase or decrease.

It’d be interesting to do forecasts and figure out where that balance could be reached with different outcomes, e.g if IC were as big as AWS and cycle burn to mint ratio per subnet were x.

Staking Deflation

Staking is considered by some as a form of deflation and while it might look like it, especially if the tokens are staked for 8 years, this conviction might prove fatal in the long term, here is why:

Governance Rewards are based on a curve, which will eventually plateau at 5% in a few years: 5% of total supply.

In a rational market token price will increase if:

  1. Cycle burn rate > Node rewards mint rate

  2. Total cycle burnt yearly / (1T * XDR/ICP avg.) >= Issuance (ICP tokens added to the circulating supply, max is 5% of total supply, min is 0)

Max Issuance: if both 1 and 2 are true, NNS driven inflation decreases as supply shrinks and scarcity increases irreversibly.
Min or low Issuance: if both 1 and 2 are true due to low issuance caused by locked tokens, the scarcity increases on paper as long as they remain in the NNS, inflation increases due to total supply growing every year and the price might still increase due to the “temporary” scarcity, but eventually investors will want to cash out that value and the newly dissolved tokens flooding the market combined with the increased baseline inflation could cause major dumps and start a chain reaction.

It would then be preferable if as many tokens tokens were burnt and total supply shrank rather than having too many tokens locked and an increasing total supply.

Conclusion

Despite calculating metrics using the best case variables for deflation’s sake, the results aren’t promising, so unless I’m missing something or have committed mistakes along the way, we might have a problem on our hands.
I truly believe this is a very important topic that needs to be discussed as it entails many aspects of the IC’s future: network’s profitability and security, investors’ and stakers’ financial prospects, operational costs for devs.
I hope to have conveyed my thoughts clearly, if there is something you haven’t understood or I got wrong, feel free to tell me, I’m looking forward to hear what the community has to say.

Thanks for reading :smiley:

21 Likes

I share three articles by Kyle, one that analyzes the long-term model and the other the current statistics that I don’t know if you used to make the calculations…

1 Like

Another thing I noticed is that you only calculated the conversion burn from ICP to Cyclos, but you didn’t calculate the transaction burn.

2 Likes

One thing to note is that when DEFI in IC is 100% developed, they will have to lock tokens to have capital backing. While this does not affect inflation, it does affect the price of the token. In addition, hundreds of thousands of transactions are generated that also contribute to burning and the latter does impact inflation.

1 Like

I’m aware of those estimates but I didn’t take them into consideration for my analysis.
I don’t quite agree with Kyle’s conclusion:

There exists a technical hardware limit to the amount of cycles a node can can burn in a given time. A while back I calculated that technical limit to be between 7500 XDR and 12500 XDR per month

If we assume that an average node on the IC burns 4000 XDR worth of ICP a month in 2030 (using 12000 XDR as the technical limit and the IC executing at 33% of capacity) and costs 750 XDR a month, then the Op would be 3250 XDR in 2030.

  1. It’s dangerous to assume XDR per node will be lower than the lowest rate being paid now, inflation, hardware cost increase (Gen2 HW is already more expesive than Gen1) might influence providers rewards for the worse.
  2. Even ignoring the optimistic assumption, profitability has to be calculated per subnet not per node, 750 XDR * 13 (avg. node count) = 9750XDR/subnet/month, even if we use the max burn rate he calculated that’s only 2250XDR/subnet/month, so how many subnets are needed to burn 5% of total supply? Say 20 mil ICPs? Depends on the price, at 4XDR, which I doubt is the price we’d like the token to settle on, that’d be: 80 mil XDR / 27000 = ~2.962 subnets. That’s a LOT!

And even then there is still the potential of entire subnets being used only for storage and burning almost no cycles as described in the post I’ve linked.

Another thing I noticed is that you only calculated the conversion burn from ICP to Cycles, but you didn’t calculate the transaction burn.

TX driven burn is quite low, 0.0001 ICP per TX, so you’d need billions of TXs to make the difference. The IC is also a decentralized cloud more than a blockchain, so many dApps might not even generate IC TXs, burn rate is a much more reliable metric, most TXs I assume will be from custom tokens and those burn cycles based on the operations used by the canister’s methods.

Even if it werethe case, I’ve described why relying on locking in the NNS for deflation might not be a good idea.

There is already a thread discussing this…

While there might be some overlap between the thread’s topics I think they are different beasts, OP wants to deliberate changes to reward distribution until the network becomes profitable, my post on the other hand is meant to discuss whether network profitability is realistically achievable under current parameters.

1 Like

I disagree with you, the growth potential is incalculable.
Transactions can be infinitely larger.

We are not only talking about transfers in terms of value, but also in terms of the action of an app or game.

I give you an example within a game:

An MMORPG game (like WOW) where every time a person kills an npc, a token transfer is made to the player. Or in a p2p every time you die or kill someone a transfer is made, that would generate incalculable transfers.

An example of a chess that every time a piece is eaten a transfer is made and also to win…

An example of a sale of a business that every time it sells or someone consumes something, a transfer is made.

An app like spotify that each song you listen to generates a transfer…

Do you understand the dimension of what we are talking about?

1 Like

Yeah and most of those TXs wouldn’t be ICP transfers in a majority of cases but ledger operations whose cycle cost is defined here and is much cheaper than 0.001 ICP: Computation and Storage Costs | Internet Computer Home.

The growth potential might be incalculable but ultimately it doesn’t matter if subnets can’t realistically burn more than they mint, TPS per subnet is finite so even if you don’t want to believe my findings, Kyle too estimates about 12500 XDR per month max theoretical burn per subnet with a 9750XDR cost to run it, the surplus which would be subtracted from governance rewards is quite small.

4 Likes

I understand where you are pointing. Just a serious question, what do you care if it is inflationary or not, what matters to you is the price. All the countries of the world are inflationary (with a small inflation to grow) and that does not generate a problem.

The problem is that you see a relationship between price and inflation. The higher the inflation, the more likely the price will go down, I know. But the price is supply and demand. That is where the amount of circulating supply comes in and the staking that exists day by day, on the other hand, if the defi advances, you also have less liquid supply, which will affect the price. If inflation is 5% and they give you 15% for staking, does that 5% really matter or do you care about the difference with what they give you?

Tell me if I’m wrong, but 99.99% of people worry about inflation, it’s about price. And the price is supply and demand, it’s another discussion.

3 Likes

An interesting tokenomic hypothesis about inflation deflation and price, and it is indeed quite contrary to most people’s mindset. Your hypothesis need to be validated for further sustainability analysis.

What I see is, ICP tokenomics has some part similarity with Modern Monetary Theory that gaining popularity nowadays. I am curious if the designer of ICP tokenomics (Anybody know who?) maybe inspired by it.

This experimental monetary policy is considered controversial due to its ability to inflate currency as much as they wish to stimulate and grow the economy, just like ICP protocol can mint token as much as it want to achieve its objectives. Some said it solves funding problem, some said it is a receipt for future disaster. And I believe not all ICP stakeholders are comfortable with MMT style tokenomics

I agree with you that ICP growth potential is incalculable right now, and maybe this is the reason of why all token inflation metrics is set by dfinity at the super-optimistic scenario, because the future growth is super huge.

But, will it be wiser if the token mechanic is set at more at realistic best case scenario ? An adaptive tokenomics that can adapt & adjust with market condition, especially in a bear market.

1 Like

When you refer to countries I assume you mean fiat money, if that’s the case it is a different topic entirely, money are a medium of exchange, they are not supposed to store value for a long amount of time, the comparison would be appropriate if the IC token were only used to be converted in cycles, but that isn’t the only use and as explained if the price isn’t at least stable the network will suffer in the long run.

Maybe I wasn’t clear enough with the Staking Deflation paragraph, it is my opinion relying on lockups to counter an otherwise impossible to overcome inflation is a slippery slope.
We agree price is based on supply and demand, if the supply is regulated by burning tokens, the price shall reflect the day to day ratio between the 2 variables and since Issuance is predictable, the main variable to observe is utility, therefore the supply/demand ratio should oscillate gracefully based on network usage.

If the supply is regulated by any form of locking, the above is only half true because Issuance is no longer procedurally predictable by looking at a formula but it is influenced by human behaviour, so if to achieve deflation the IC must rely heavily on stakers both staking tokens in the NNS and compounding their rewards, price will go up but only as long as NNS inflows outpace outflows, which as time goes on become less and less likely as:

  1. Stakers will reap their long awaited rewards by either dissolving neurons or selling maturity.
  2. Total supply increases every year as maturity gets compounded, so even if the inflation rate flattens at 5%, minted tokens increase due to total supply growing yearly and stakers have to compound an higher % of their rewards to keep deflation going.

In a matter of days the circulating supply could increase at any time by a substantial margin, imagine the impact on price, which has been inflated by an artificially lowered supply, when those tokens get in circulation, all it takes is a whale cashing out and a chain reaction might start:
supply increases → price dumps, stakers get scared and decide to sell → supply increases, rinse and repeat.
The only way this won’t happen is if staker’s never sell more than the market can take before supply/deman ratio are affected, which is wishful thinking at best.

Locking for 8 years might seem like a lifetime and akin to permanently burning the tokens but remember we are trying to build the internet of tomorrow, so the timespan to consider is decades. Plus no other chain uses DeFi and staking as deflationary mechanisms, otherwise ETH would have been considered deflationary years ago.

Rewards are based on the ratio between your own neurons’ voting power and the total VP, as more tokens are staked they get lower, eventually you might be able to counter your holdings dilution by staking for 8 years, but not everyone wants to do that nor it should be necessary.

Inflation affects supply and if a commodity can be generated faster than it can be depleted than there is no reason for its price to go up, except speculation, but while that might help price in the medium term, if crypto and the IC are here to stay eventually they will be valued based on true utility.

6 Likes

To have more accurate numbers you must find out how many cycles does an IC dapp need per year.

That isn’t the metric we should figure out, but how many cycles per month/year a subnet can consume, which I suppose is based on single subnet’s tps.
How much a dApp needs on average is almost irrelevant cause a subnet can run many dApps simultaneously, the more dApps can run on a single subnet the higher the cycle burn to mint ratio is and the less subnets are needed to scale the network, which further reduces token minting.

It’d be nice if Dfinity could provide us some numbers and clarify how much we can expect them to increase as the protocol gets more efficient.

1 Like

Exactly that I have in my mind when I asked that

While I don’t have the ability to understand most of what you are stating I fully support your questions and explain in layman terms what I mean.

The IC is built around creating a network and needs funding, ICP staking provides this, needs projects built on the IC, needs infrastructure and tries to balance to make all involved to be happy.

The balance, overtime is not in the favor of neuron holders and while we are locked in for 8 or 16 years depending how you look at it our rewards will dwindle while infrastructure and projects grow.

You can not make everyone happy and neuron holders are the sacrificial lamb. The whole design of the IC model, you call inflation and deflation, I call stagflation.

The apps that are being built on the IC, wallet, web hosting, luxury goods, social, chat, bank, market place, games, social, calculator … nothing new or impressing me and they are competing with organizations that have been around for decades with much more experience and resources.

I think the project is great but I always thought the best I can do with investing in this project as a neuron holder will be the first 4 years, by then the rewards will be less than I will receive in other investments but still locked in for many years.

I will at this time be locked without dissolving for the first 4 years and will hold onto my maturity where I will spawn to take my rewards from the IC altogether and hope the price will be above what I paid per icp which at this time is very much lower.

I don’t believe Dfinity wants the price of an ICP to increase, unlike other coins.

I am hoping that Dfinity will see that in the future many Neuron holders have done the same and take action to entice us back to investing into the IC.

Only having the current projects on the IC and future smart contracts with BTC and ETH, well in my opinion, who isn’t.

I believe like many organizations, Dfinity will find that they will have to provide other services to survive and make the only decision in my opinion and that is to use the NNS to provide the security that the Internet needs right now and will need even more on the broken system that we have today.

Why not entice big tech organizations to build on the IC, top 100 companies not related to crypto.

It’s costing billions in fines being on WEB2-3 for many of the top companies, do we not have the infrastructure to support and allow them to do business on the IC without them losing personal data of their customers to spammers.

Why is the IC just another web full of anything goes, when we could be just a RETAIL network and separated from the current public nonsense.

The username and password system will be removed on the IC and will be used to login to all your services.

What if everyone on the IC had to buy a ICP, add funds and paid their RETAIL bills and services through the NNS.

Keep the good work up @Zane

I’ve sketched this simplified version of how the IC currently works in regard to token minting/burning, hope it makes it a bit easier to visualize the concept explained in my analysis. As always I’m open to your feedback.

5 Likes

I have been following this forum discussion in detail trying to understand the reverse engineering rational of all participants. It is surprising that we, as a community, are still trying to figure out the the design principles of the Internet Computer Tokenomics strategy. Are we meant, by its design, to continuously inflate or to progressively deflate or perhaps, the design principles were meant to have a periodic inflation/deflation behavior. Whatever are these design principles and the way these principles were engineered into concrete decisions is not known. So, instead of trying to reverse engineer these latent design principles and the implementation methods, could Dfinity (@diegop) expose them clearly?

2 Likes

From what I’ve read/heard I’ve come to the conclusion Dfinity or specifically Dom considers 8 year staking as deflation and relies a lot on it to keep the supply in check. For the reasons stated in my analysis I don’t agree with that conclusion.

3 Likes

Where are the design principles, research study, simulations that lead to the conclusion of the concrete Tokenomic strategy? We should not have to speculate based on what Dom said here or there.

4 Likes