Node Provider Inflation Spiral

Governance
21

@alejandrade , The problem with paying a fixed price in $ICP is that node providers for the IC are data centers with overhead costs that must be covered in order for them to Provide reliable compute power. Other blockchains can pay a fixed price to node providers as rewards because they have node hardware that someone can set up in their basements if necessary. The cost of operation for a data center vs that of a basement node are vastly different. Paying a data center a fixed amount of ICP would result in all of the node providers having to cease operations all at once when the price of ICP drops below a certain threshold.

I’ll give an example of what i mean:
Let’s say 1 $ICP = $10 USD and it costs $10,000 per month to operate a data center. We agree to pay the node providers a fixed rate of 1500 $ICP per month. The first month, the node providers would make enough to cover their overhead costs and take profits.

The next month, the price of $ICP drops to $3 USD. We pay the node providers the agreed upon price of 1500 $ICP. This month, they’d only earn $4500 USD. This would mean they’re $5,500 short of what they need to run operations- forcing them to shut down operations.

Now imagine all of the IC’s data centers having this issue at the same time. ICP would literally be inoperable over night.

Paying Node providers a fixed amount of ICP every month is not a solution. In fact, it’d be catastrophic.

4 Likes
22

The problem is that they are getting paid in cash value no matter if they are working or not. This is another problem caused by trying to reinvent the wheel.

This problem has been solved, proof of stake exists.

It should be a minimum amount of ICP staked to be able to host a node and payment should be either fixed return like POS or proof of useful work.

And some node providers will not be profitable while others will be. Some will make more money then others all this is ok.

3 Likes
23

Another alternative could be something similar to the neurons fund. All neurons pay maturity to a fund that distributes to providers accordingly. The current node provider rewards could be sent to governance so the current interest would be the same. To reduce inflation the NP reward aspect to governance would need to diminish over time. There would also be a need to increase the NP reward aspect each time a node provider would have normally had an increase in rewards up to the max allocation per NP. Failure to do so would lower % yield each time an NP was onboarded or added new nodes. Which may in turn promote governors to turn down new NP requests. Modulation could be applied here in such a way that during times when the network needs expansion and more nodes the reward incentive to the governance pool is higher(resulting in neurons gaining % yield) than at times when there is less of a need and thus less of a reward(which results in neurons paying to add new nodes).

3 Likes
24

Hey @Kyle_Langham (or anyone else) are there any contractual agreements with node providers that would limit making changes to their payouts?

Also, developer grants are paid in fiat equivalents as well. Hence, that funding is paid with higher amounts of ICP as ICP price goes down just like node providers. Even if we are not minting new tokens for developer grants, it seems likely that most of those payouts are adding to sell pressure. What is the scale of developer grants relative to node providers? Should it be of similar concern? I suspect most people would not advocate for reducing payouts for developer grants. Hence, it seems fair to bring it up in the context of discussions for reducing rewards to node providers. Do we have the stomach to reduce both or is there good justification for only considering changes to one?

Of all the ideas I’ve heard so far, I like the one introduced by @Cryptobaasnl the most. Specifically this part…

Using this approach, the total node provider rewards wouldn’t have to be reduced. Instead, a percentage would be “topped up” into their own 8YG neuron. This seems easy to implement into the NNS since all it requires is knowing the account ID and neuron ID. Anybody can top up any neuron with this information and it seems plausible to enable node providers to provide this information for use by the NNS. This approach would ensure there is very little sell pressure from the fraction of the rewards that get topped up into a node providers neuron. This seems like a preferable alternative to a forced reduction in the total payments that they are expecting.

I personally would like to hear from node providers on this topic. Of course nobody wins if price goes down below a certain point, so it seems reasonable to consider backup plans.

Hey @Accumulating.icp, I also want to say that you have done a fantastic job presenting this topic and leading the deliberation. Keep up the good work.

6 Likes
25

Yes! A combination of neuron rewards, a reward pool funded by donations from the those who have a vested interest in the network, multicurrency rewards (perhaps only ICP, ckBTC, and later ckETH), and direct ICP minting to fill in the gaps could provide the sustainability needed to prevent an inflationary death spiral no matter how low the price sinks.

2 Likes
26

8 years is insanity for an investment like this the risk would be high

2 Likes
27

This design would essentially operate as a cascading waterfall, each layer designed to function as a financial bulwark for the layer beneath it. Let’s delve into its architecture.

Tier 1: Donation Pools

The first echelon for rewards would consist of donation pools, aggregated from benevolent contributions in the form of ICP, ckBTC, and ckETH. This serves as the primary reservoir for node provider rewards. Its compositional diversity offers a modicum of hedging against market volatility, insulating the pool to some degree from cataclysmic fluctuations in any one asset. Payouts would start here, satisfying as many node provider obligations as possible.

Tier 2: Neuron Maturity

Should the donation pools run dry or be insufficient, we move to the next tier: neuron maturity. In this scenario, matured rewards that are accrued by neuron holders are divvied up and allocated to the reward pool. This element introduces a sort of circular financial ecology, where node providers are indirectly compensated by the very network participants who benefit from a stable and efficient network.

Rate Reset Mechanism

At the beginning of each node provider payment cycle, the rate at which rewards are doled out is reset. This dynamic recalibration ensures that the system remains nimble, capable of adapting to the vicissitudes of market conditions.

Tier 3: ICP Minting & Excess Allocation

If the first two tiers fail to meet the node provider reward threshold—perhaps due to a depreciation in the price of ICP—then the system resorts to the controversial but necessary action of minting additional ICP tokens to meet the shortfall. Conversely, if there’s an uptick in the price of ICP that results in excess ICP from neuron maturity, these surplus tokens flow back into the donation pool, replenishing it for the next payment cycle.

Reflexivity and Balancing Act

The system’s beauty lies in its inherent reflexivity; it feeds upon itself to establish equilibrium. During bullish periods, it’s self-replenishing, and during bearish periods, it’s self-sustaining, albeit at the cost of a slight inflationary bump.

However, this architecture is not without its challenges. Mechanisms must be put in place to ensure equitable allocation and to prevent any form of collusion or sybil attacks. Additionally, a constant audit of smart contracts governing this system would be imperative, ensuring both transparency and immutability.

In sum, the tiered system presents a sophisticated, self-balancing economic model that aims to reconcile the often divergent goals of incentivization and financial prudence. It’s a system replete with checks and balances, each tier functioning as both a safety net and a financial springboard for the one that follows.

edit: there are concerns that this design minimizes risk too much, so node providers have no incentive to stick with the IC network. I did like the idea of a portion of the rewards being locked in 8yr neurons, since that would provide an incentive to commit to the network long-term.

3 Likes
28

It’s fun to fantasize I doubt anything will be done and we have no power. :joy:.

If we want to fix emission we need to lower staking rewards. And if we want to lower staking rewards we should cut down everyone’s 8 year neuron to a max of 4

29

But teams take on this risk and have overhead costs as well, are node providers more important than apps and projects? What are the overhead costs of running an App?
Well you cant have ICP without nodes and you don’t need nodes without apps and projects. In truth this is a feedback loop, and there needs to be a balance. The less inflation caused by nodes the more profit for apps. The more profit for apps the more users and more ICP is burned.
There is a reason that central banks tweak interest rates, its because economics change on a regular basis due to many external factors.
Fixed rates of interest ( rewards ) don’t make sense and lead to spirals of inflation or deflation

5 Likes
30

I support this is a plan of action.

@Jan , @domwoe, @diegop, it would be nice if we could get more people from the DFINITY foundation to engage with the community on this topic. Macroeconomic conditions are likely to continue to place sustained downward pressure on the crypto market for at least another 2 years or so. The topic of the very real possibility of run-away inflation caused by node provider rewards needs to be planned for.

5 Likes
31

The lower price, the more get. If it is me, I will definitely find a way to reduce the price of ICP. This is the sword of Damocles for every ICP investor.

34

Thanks for providing such an in-depth overview of the concept!

I think this is an interesting idea, although I believe it’d come with a range of pro’s & con’s that I’d love to walk through;

Donation Pools

Donation Pools would definitely contribute to providing a diverse source & form of funds, potentially offering insulation from market volatility. With that being said, this function is heavily reliant on human psychology & behavior - during a time in which people are going to be most inclined to abandon the network. Additionally, Donations can not be viewed as a long term form of funding, as they are typically infrequent, making it challenging to fulfill Node Provider obligations consistently (~2M$ USD Monthly).

Neuron Maturity

While I see the logic in rewarding Node Providers Neurons from the Neuron Reward Pool, I believe directly front running the Neuron Reward Pool to payout Node Providers beforehand will have the unintended consequence of disincentivizing staking, potentially when it’s needed most.

For example, in the last month, if you observe disbursed & un-disbursed maturity, you’ll find that there was 2,680,437 ICP minted for Neuron Rewards. Alternatively, at present prices, ~580K ICP is minted monthly for Node Rewards. If this system were set in place today, this would reduce the current APY of Neurons by 21.6%.

As the price descends, the scenario becomes more extreme, until an overlap occurs around 1$ - at which Neurons are no longer rewarded for providing security to the network through staking & voting.

35

The tokenomics should align with responsible accounting and cashflow principles of basic corporate finance. Currently, the tokenomics comprises two token printing streams: Node provider rewards and government voting rewards.

It is advisable to start with only one token printing stream. Node provider rewards should be treated as a part of the cost of goods sold (COGS) and operational expenses, and therefore, onboarding new node providers will face constraints and limitations based on the network capacity & demand requirements. These expenses should be deducted from the voting rewards first, and any excess token inflation then will be distributed among the voters.

For example, if the current voting reward is 3 million ICP per month and Node provider rewards amount to approximately 500,000 ICP per month, it implies that the ICP voting rewards to be distributed among the governance voters will be only 2.5 million ICP.

6 Likes
36

The issue is that it doesn’t solve anything, it just shifts the burden of responsibility.

It went from a burden of the network to a burden of the staker.

You can’t just expect people to govern the network because they like it - there has to be some form of incentivization, being the staking reward itself.

When you remove the staking reward, you remove the incentivization to stake, and everyone who has already staked, gets, for lack of better words, fucked.

This doesn’t prevent a hyper inflation scenario from occurring - it just forces neurons to bear the responsibility of it until the breaking point is reached, at which point they’re indefinitely trapped.

1 Like
37

It is recommended to set up a mechanism. When the currency price drops to a certain price, part of the node’s fee will be put into neuron pledge. Avoid the death spiral of falling more and more inflation, and more inflation and more fall. While avoiding large-scale inflation, it also protects the interests of nodes. Nodes are an important part of ICP’s ecology, and nodes are also responsible for maintaining a healthy currency price.

5 Likes
38

The stakers are the ones who bear the responsibility for hyperinflation because they are the owners of the protocol. The stakers are also the ones in charge of governance, including being responsible for deciding how many NP they want to onboard and the rewards they need to set for the NPs. So, if we want to avoid hyperinflation, we cannot make everybody happy. We should find compromises.

39

I agree we need to find compromise. However, this is not a compromise that resolves the problem.

It is one that temporarily delays it, until Neurons can no longer bear the burden - at which point we’re back where we started, yet stakers lost a few months of rewards along the way.

We need a solution that makes this a non-issue in its entirety.

40

Just to confirm from my side as well: I am actively following the various discussions and ideas that are being presented in this thread. As @Kyle_Langham also pointed out, node provider rewards are currently not a significant factor in the overall inflation of ICP or in the amount of ICP moving to exchanges. However, I believe it is prudent to gather ideas and formulate a plan for various exchange rate scenarios.

11 Likes
41

I kept seeing you & @Kyle_Langham reference this as a non-issue in comparison to other forms of inflation, so I decided to aggregate current data to depict the reality of ICP inflation & an extrapolated future.

Monthly N.P Rewards | 583,577

Monthly N.P Rewards Sent to Exchanges | 275,287.49

Monthly Un-Disbursed Maturity | 1,582,659

Monthly Disbursed ICP | 1,091,778

Monthly Staked ICP | -1,644,659

From here, the total potential inflationary sell pressure on a monthly basis can be determined by adding the monthly disbursed ICP, Monthly staked ICP, and monthly rewards sent to exchanges - equating to a total of 3,011,724.49 ICP. It should be noted that this equation does not include Un-Disbursed Maturity, however this could also add to inflationary selling pressure.

This shows that as of present, Node Provider Reward sell-off equates to ~9.1% of the monthly inflationary sell pressure at current prices.

However, given Node Provider Rewards are paid in a static fiat amount, yet a variable ICP quantity, this data can be extrapolated to determine the contribution to inflationary sell pressure at the pre-determined points referenced within this article:

Price - ICP to Exchanges - % of Inflation Sold

2.5$ - 385,627.6816 ICP - 12.3%

2$ - 482,034.602 ICP - 14.9%

1.5$ - 642,712.8026666668 ICP - 19%

1$ - 964,069.204 ICP - 26%

0.5$ - 1,928,138.408 ICP - 41.3%

0.25$ - 3,856,276.816 ICP - 58.4%

In conclusion, I’d agree it’s a non-issue in comparison to other inflation rates within the immediate future… Until it isn’t…

6 Likes
42

@bjoernek

This sentence you wrote “ As @Kyle_Langham also pointed out, node provider rewards are currently not a significant factor in the overall inflation of ICP or in the amount of ICP moving to exchanges. ” is absolutely wrong .

A year ago, it was less than 100k node rewards per month. Now it’s 600k node rewards per month. A simple back of the envelope calculation — if icp goes to $1, it’ll be close to 2 million rewards per month. When that happens, I guarantee you icp price goes straight to 0 within 1 day, just like what happened to other tokens that had supply-demand design fault. When that happens, the whole icp project becomes worthless.

2 Likes