An idea from @Nicola and @Alex North.
A storage provider must pledge Filecoin tokens for the right to commit a sector and earn consensus power and block rewards. The pledge is proportional to sector power (including FIL+ multipliers). This pledge comprises two components:
- a storage pledge: held to pay penalties in case of faults or termination, approx 20 days of expected rewards at time of commitment
- a consensus pledge: held to align long term incentives with network, pro-rata of approx 30% of circulating supply at time of commitment
This pledge is released back to the storage provider when a sector expires or (less applicable fees) is terminated.
The circulating supply used to calculate consensus pledges excludes currently all pledged tokens. Committing a new sector thus causes circulating supply to decrease, while letting one expire causes it to increase. Pledge for positive net onboarding is the main (potential) outflow from the circulating supply.
There is a circular dependency between pledge amounts and circulating supply which causes a destabilising dynamic. When circulating supply is increasing (from block rewards, vesting, or net sector expirations), the pledge requirements for committing a new sector increases too, and thus the rate of return on that pledge decreases. An increasing circulating supply means decreasing the aggregate incentives to commit storage.
Conversely, when the circulating supply is decreasing (from high sector onboarding, or possibly gas fees), the pledge requirements for onboarding decrease too, increasing incentives for yet more onboarding. This cannot be sustained: eventually the per-sector returns are competed to subsistence, the onboarding stops, and the supply turns back to inflation.
So the supply:pledge relationship is a destabilising force on sector onboarding and hence the token supply itself. When the rate of change of token supply is far from zero, changing incentives for onboarding push it yet further from zero. In particular, when net onboarding is low or negative, the return on pledge of committing/extending sectors is decreasing, further discouraging onboarding.
For a stable circulating supply, Filecoin needs the pledge from positive net onboarding to match the emissions from vesting and block rewards.
This is a very surface outline of the issue, without considering storage providers’ share of power (which eventually counters this dynamic), the baseline minting function, and other complications.
Idea - the pledge token
Instead of pledging FIL, storage providers can be required to pledge a new token, PFIL, at the same rate as the current pledge formula. The network permits a 1:1 conversion of FIL to PFIL, but not the other way round. The FIL for sector pledging is thus permanently allocated to pledge. When a sector expires or is terminated, PFIL tokens are returned to the provider. These PFIL tokens can be used to pledge or extend a new sector, but cannot be converted back to FIL. Unlocked PFIL can be transferred freely, traded, etc.
The PFIL tokens do not form part of the circulating supply, just as locked tokens don’t today. The difference is that they remain outside the circulating supply after being unlocked. Sector rewards continue to be paid in FIL.
In times of positive net onboarding, the incremental pledge requirements would be a constant outflow from the FIL circulating supply, just like today. However, in times of negative net onboarding, the released PFIL tokens would not impact the FIL circulating supply or pledge requirements. Instead, the supply of unlocked PFIL tokens would increase. A market exchange rate of PFIL for FIL would drop below 1, meaning that sector onboarding would become cheaper, and returns on FIL would increase.
The pledge token thus introduces a damping force on the cycle of decreased onboarding → higher supply → decreased returns for onboarding, in two ways:
- Released PFIL tokens do not increase the circulating supply of FIL or pledge requirements
- Abundant PFIL can be priced below 1 FIL, reduce the cost of onboarding and increasing returns to new commitments
Converting FIL to PFIL is a long term commitment to the network. A provider that intends to pledge indefinitely can do so directly, but a provider expecting to exit after a short period will bear risk that PFIL is valued below 1 FIL at that time. When PFIL is cheap, providers are incentivised to either hold for a time when the network is growing, or even better, commit storage in order to earn a return on their PFIL.
We might maintain the storage pledge (for penalties) in FIL, while the consensus pledge would be PFIL.
If a pledge token mechanism were introduced only for new commitments, in the short term we might expect a slight decrease in net onboarding. This is because at least some storage providers could not assume full return of their pledge in FIL after completion of their commitments, but would need to factor in some price risk on their PFIL. PFIL is in essence a less valuable token. The conversion would amount to a decrease in total expected returns.
There are good arguments that FIL-on-FIL returns have been unsustainably high in the past, and should be reduced to give greater network stability in the future. Mechanisms like a duration multiplier or higher pledge target pursue this goal.
This decrease in expected return rate would buy stability of storage and FIL supply in the future:
- When onboarding is high and circulating supply decreasing, this mechanism would damp the cycle of higher returns by requiring SPs to convert FIL to PFIL, a less valuable token.
- When onboarding is low and circulating supply increasing, this mechanism would hold expired sector pledge out from supply inflows and, when PFIL abundant, increase returns to new commitments.
The PFIL token would effectively buffer the FIL supply from downside volatility in onboarding rates, and transfer onboarding-related volatility in FIL price to the PFIL price instead.
Extension - redemptions
As an extension to the core pledge token idea, the network could offer a redemption mechanism to convert PFIL back to FIL at a rate below 1.0. One reason to do so is that PFIL would form part of the core network infrastructure, and so it would seem risky to rely solely on third party marketplaces (user-programmed DEXs etc) for PFIL/FIL liquidity. A redemption mechanism could form the buyer of last resort when other PFIL markets fail.
One mechanism might resemble a bonding curve based on the scarcity of unlocked PFIL. When almost all PFIL is locked in pledge, the network might offer buy PFIL at, say, 0.95 FIL. The 0.05FIL difference between the original issuance price and redemption would be burnt.
As the fraction of unlocked PFIL increases, the redemption amount decreases towards some floor of, say, 0.3 FIL.
A network redemption function would make taking profits easier for an SP at times when the network can absorb that with least impact, and possible but costly when the network is less healthy. We’d expect SPs to usually sell their tokens to another SP instead, for a better price. The discount on redemption represents a fee paid to the network for acting as a market maker.
The redemption curve represents the key parameter of the pledge token mechanism which would govern its power and effectiveness.
- If redemptions were always 1:1, the mechanism would have no effect.
- If redemptions were impossible, the mechanism would be maximally powerful, but the PFIL token most risky. This would achieve the greatest tradeoff between aggregate commitments and the long-term stability of those commitments.
- A positive redemption price decreases the risk and increases the expected value of PFIL, but allows some leakage of unlocked pledge tokens into the FIL circulating supply (and hence pledge calculations, etc).
This would impact lending markets. SPs need to borrow PFIL to pledge sectors, but lending PFIL involves taking on the PFIL:FIL price risk. Lending would no longer be neutral to network health in FIL terms.
This is, unfortunately, a difficult mechanism to introduce to the network except during a period of strong growth. The near term effect would likely be a reduction in onboarding, but the stability gains from that would not be enjoyed until commitments expire, 6-18 months later (or, maybe 1-5y later if we increase commitments). It’s a mechanism to wish we had from the start, or to introduce in the next bull market.