Demand-Gated Emission: Reducing the Fixed 767k/Epoch Validator Subsidy #49
Replies: 10 comments 8 replies
|
Update: the complementary lever is now running. Following the fee/usage feedback on this proposal, I built the counterpart as a working pilot: a small public gas station that sponsors onboarding transactions. Live demo: https://doggy-dex.com (IOTA testnet) - press the button, a throwaway wallet holding exactly nothing sends a real transaction, the pilot pays, and you get the gas receipt. Source, access rules and deployment: https://github.com/Zertruermmerdog/iota-gas-station-pilot Verified numbers from the pilot:
Why this belongs here: sponsored fees are still burned, so a funded public gas station converts treasury into adoption and burn at the same time. Combined with the demand-gated subsidy sketched above, the two levers close the loop - emission steps down while usage is low, sponsorship bootstraps usage, and rising burn automatically halts further reductions (open questions 5 and 6). If there is interest, I will write this up as a funding proposal (community fund / Tangle DAO) with the pilot data as the baseline. |
|
Hello, thanks for the discussion. I find it very difficult to follow along and understand exactly what you are proposing. It would help if you restructure and clean up the post with a clean idea of what you are proposing. The tl;dr that I understand is that you are simply proposing to reduce inflation because there is not much activity in the network counteracting it. While this is true, this discussion is lacking concrete alternatives on how operations of the network should be paid, how validators should cover the operating costs of the hardware and people administering them. You propose to cover these costs with a I don't think the solution to the problem you are describing is just solved by capping the subsidy without changing anything else in the tokenomics. You mention "early community feedback" a couple of times. I would welcome this feedback to be discussed here in the open so that everyone can participate. |
|
Thanks Alex - fair criticism on all points. I have restructured the post (v2, changelog at the bottom). Responses to your specific points: How operations get paid / how validators cover costs. I worked this through with live chain data (epoch 431), and it changed the proposal. At current prices the median validator's commission income (~248k IOTA/year ~ USD 10k, at 10% commission on a ~0.9% stake share) roughly breaks even with professional operating costs, and the 25th-percentile validator is already underwater on commission alone. So you are right: capping the subsidy without changing anything else would break validator economics. The proposal is now explicitly a package - every subsidy step-down is paired with a proportional increase of the protocol minimum commission (extending the IIP-8 rule to One more datapoint that sharpened this: the squeeze happened after commitment. Most of the committee onboarded in the first months after genesis, when the token traded around USD 0.20 - the fiat value of the same reward stream has since compressed ~5x - while the required hardware class (64 GB RAM, enterprise NVMe) got dramatically more expensive in the ongoing memory shortage. Today's rough break-even is drift, not design: nothing in a fixed 767k constant tracks either side. The floor plus coupling is the piece that makes cost coverage explicit instead of accidental. Viability floor. The post now contains a concrete cost model instead of a bare assertion: an absolute floor derived from where even 100% commission leaves the median validator at break-even (~90k IOTA/epoch at today's prices and committee shape), and a practical floor above it (150-200k/epoch with the commission cap at 40-50%). One limitation stated honestly: operating costs are fiat-denominated, so the floor cannot be a fixed on-chain formula - it has to be a governance-calibrated parameter revisited per protocol upgrade. Real cost figures from operators would improve the calibration a lot; that is now open question 4 - if the Foundation has cost data from its own node operations, that would be ideal calibration input. Community feedback in the open. Agreed - referencing it without attribution was sloppy. Both points came from the IOTA Discord on Jul 9: Dr.Electron suggested attacking the fee/burn side instead (now under "Alternatives considered"), and Texos suggested coupling the subsidy continuously to usage with the current 767k as a hard cap, noting stakers may be fine with minimal rewards while usage is low (now the "continuous usage-coupling" variant). Those are my paraphrases - corrections welcome, and it would be great to have both weigh in here directly. If the restructured version is still hard to follow, tell me which section and I will keep tightening it. |
|
Hello, We would like to add that the costs for running a validator go far beyond just hardware. While some operators use VPS solutions (as mentioned in your post), others run dedicated servers. More importantly, you are overlooking the operational service aspect: providing a reliable 24/7 validator involves significant ongoing expenses for monitoring, maintenance, security, updates, and general operations. Furthermore, many validators deliver additional value to the ecosystem that goes well beyond simply running a node. Examples include:
It is therefore not accurate to treat the cost side in such a simplified or generalized way. The former IIP-8 was designed to prevent larger validators from gaining even more stake through dumping (ultra-low) fees. It represents a targeted intervention at the edge of the system. However, setting the floor incorrectly could have the unintended consequence that exactly those validators who currently enrich the ecosystem with extra services — including support — may decide to discontinue these additional offerings. They would then effectively become nothing more than basic server operators. This is just an initial remark. We will examine this topic in greater depth, as a proper analysis also requires significant time and resources. Best regards, |
|
Thank you - this is exactly the operator input open question 4 is asking for, and I appreciate it coming from the authors of IIP-8 itself. Costs beyond hardware. Agreed. The cost model in the post already includes a state-sync fullnode, monitoring and operator time on top of the reference hardware (that is how it gets to USD 6k-15k/year rather than a bare VPS bill), but I take the point that a professional 24/7 operation - on-call coverage, security, updates, incident response - can sit above that range, and that generalizing across 66 very different operations is lossy. That is precisely why the floor is framed as a governance-calibrated parameter rather than a fixed on-chain formula: the calibration is only as good as the cost data behind it. If DLT.GREEN could share real figures - even coarse buckets (infrastructure / personnel / additional services) or ranges - that would improve the floor calibration more than anything else in this thread. Additional ecosystem services. Important point, and worth being precise about the status quo: the current protocol does not pay for these either. Rewards are stake-proportional and undifferentiated - a bare-bones operator with the same pool size earns exactly the same as a full-service validator running indexers, explorers, wallets and delegator support. Those services are funded out of whatever margin price drift happens to leave, and that margin has compressed roughly 5x since most of the committee onboarded (see the drift paragraph in the post). So the "reduced to a basic server operator" pressure you describe already exists today - silently, with no mechanism to respond. The coupling in this proposal at least makes the income line explicit and defensible: every subsidy step-down is paired with a commission increase that keeps validator income constant in IOTA terms. The reduction is taken from delegator yield, not from validator service budgets. Where the floor sits. Fully agreed that calibrating it at bare-bones break-even would be wrong - that is why the practical floor (150-200k/epoch) deliberately sits well above the absolute break-even floor (~90k/epoch): the gap is exactly the room for full-service operations. Your numbers would move that boundary, in whichever direction they point. IIP-8's original purpose. Fair - you know its intent best. The extension is No rush on the deeper analysis - even preliminary cost ranges would be very valuable. And I would rather have the miscalibration risk you describe managed explicitly, per protocol upgrade and with operator input, than the current version of it, where the same squeeze arrives unmanaged via price drift. |
|
We believe there may be a slight underestimation of what “additional infrastructure and services” actually entail in a professional setup. To give you a clearer picture: This easily reaches ~€ 50K per year minimum — and that is before valuing the actual service delivery, personnel on-call coverage, incident response, or any profit margin. Not every validator offers such extended services. To my knowledge, full indexers are currently operated by only a small number of teams (Ankr, IF, and DLT.GREEN). Other validators contribute different added value — such as analytics tools, ecosystem partnerships, wallets, explorers, support, etc... — all provided to the community at no additional cost. This diversity is exactly why broad generalizations are challenging. The real value a validator delivers lies not just in keeping a server running, but in the entire surrounding infrastructure and service package. This is also why validators operate with different commission rates (dependent in connection with the delegation). Even the time and effort invested in discussions like this represents unpaid service to the network — work that is often taken for granted but indirectly influences operational costs. In the end, delegators decide whether they recognize and value these additional contributions, besides the trust in the teams. If they do not, some validators will likely scale back or discontinue their extra services (or exit altogether), and the ecosystem will lose those benefits along with them. Any change on the fee/commission side indirectly affects validators. At the current IOTA price levels, many operators are already operating at their limit. From the outside, things often look straightforward, but once you go into detail, the reality is considerably more complex. We do not see a simple one-size-fits-all approach that fairly captures this variety. That said, this is not a final statement — we remain open to further dialogue and a deeper joint analysis. |
|
Thank you - a concrete number is exactly what this thread was missing. Before responding I tried to verify it against public data, and the result is more interesting than a simple yes or no. Your figure is consistent with what the chain shows - and that consistency is itself the strongest evidence for this proposal. DLT.GREEN is currently the #3 validator by stake (~132M IOTA pool, ~5.5% voting power) at 8% commission, which works out to ~1.23M IOTA/year of commission income - about EUR 39k at today's price. If your infrastructure floor is EUR 50k+, even the third-largest pool on the network is underwater on commission alone. Further down it gets hopeless: a median validator earns ~245k IOTA/year (~EUR 8k) at the usual 10%, the 25th percentile ~90k (~EUR 3k). Across all 66 validators combined, commission at effective rates totals ~20M IOTA/year (~EUR 0.65M) - 7.3% of the subsidy; the other 92.7% accrues to stakers. Whatever the right cost figure per operation is, the conclusion is the same: the status quo emits ~280M IOTA/year and still fails to fund the people running the network, because a stake-proportional subsidy is simply not aimed at them. The procurement side checks out too. Hetzner - probably the most price-visible host in this market - repriced on June 15, its second increase this year: 64GB-class dedicated machines now list at EUR 538-729/month, dedicated-vCPU cloud lines rose by up to ~176% in a single step, and coverage of the move reports DRAM procurement costs up several-fold since autumn 2025 on the back of AI demand. The remaining cheap "limited" variants are available only while discounted hardware lasts - which is exactly your daily-rates point. So both blades of the scissors are moving at once: infrastructure costs in EUR rising steeply while the fiat value of the fixed reward stream has compressed ~80% since committee onboarding. That is the drift argument from the post, with this month's numbers. A fixed 767k constant tracks neither side. Where I would still split the figure in two. The protocol-required security footprint (validator at 8 cores/64GB/2TB plus an SSFN per current docs - notably reduced from the 24-core/128GB/4TB in the original Rebased proposal) rents for roughly EUR 10-20k/year even at post-increase prices, failover included. The delta from there to EUR 50k+ is the public service layer: indexer serving, HA appliances, redundant self-hosted infrastructure. Both layers are real, but they call for different funding instruments - and the three indexer operations you list already demonstrate this: the Foundation funds its own from budget, Ankr funds theirs from a commercial API business (their effective commission is ~190k IOTA/year, ~EUR 6k - clearly not what pays for an indexer), and DLT.GREEN funds it from a top-3 commission stream that still falls short of your stated costs. None of the three is actually funded by the undifferentiated subsidy. Funding all three explicitly at EUR 50k each would cost ~4.7M IOTA/year - under 2% of current emission. The service layer is not expensive; per-stake inflation is just the one instrument that cannot reach it. The part I did not expect: under the v2 illustration, your income rises. With the example step-down (767k -> 192k/epoch, c_min rising to ~40%), your commission income at constant pool share goes from ~1.23M to ~1.53M IOTA/year - about +25%, and at today's prices almost exactly the EUR 50k you name. That is not an accident; it is what the coupling is for. By construction no validator's commission income falls: operators at 10% stay exactly constant, everyone below gains, and the reduction comes out of delegator yield, where the overshoot sits. On the dumping side: 15 of 66 validators currently set commission at 3% or below - including 9 of the 15 largest pools. IIP-8's VP-linked floor pulls only the very largest of them back up (to ~3-5%), because that floor scales with pool size. A subsidy-coupled c_min adds the uniform floor underneath, so commission dumping stops being a competitive weapon for anyone - and delegators choose on service quality and trust instead. That is competition a full-service operator wins. So: no disagreement that one-size-fits-all is wrong - the floor is a governance-calibrated parameter with operator input per upgrade, and your cost buckets (infrastructure / personnel / services, even coarse ranges) would improve that calibration more than anything else in this thread. Looking forward to the deeper analysis whenever time allows. |
|
Please note: The official validator hardware requirements listed from your side are inaccurate. Mainnet: and most validators run also a testnet validator additional with: |
|
You are right, and thank you for the correction. I re-checked the docs page: I had cited the testnet tab as the mainnet requirement. Mainnet is 24 cores / 48 vCPUs, 128 GB RAM, 4 TB NVMe; the 8-core / 64 GB / 2 TB profile I quoted is the testnet spec - and my parenthetical claiming requirements "were later reduced" was wrong for mainnet as well. The post is corrected (changelog v3.1) rather than leaving the error standing. What changes numerically:
What does not change is the direction - every corrected number strengthens the case rather than weakening it. At the correct spec, the median validator's ~246k IOTA (~USD 9k) commission does not cover even the bare security footprint: my earlier "roughly break-even" was too generous; it is underwater. Your "many operators are already operating at their limit" is more literally true than my own numbers showed. And the no-loss coupling is unaffected, since the income multiplier |
|
side note: IIP-8 was intentionally designed to intervene only in exceptional cases and in cases of extreme exaggeration — essentially acting as a last-resort measure at the very edge. This ensures that market intervention remains proportionate, which was an important concern for us. The design deliberately limits the ability of larger, more powerful validators to engage in commission dumping, which would otherwise allow them to attract ever more stake in an uncontrolled manner — creating a bottomless pit. In contrast, smaller validators were intentionally given limited flexibility to conduct introductory promotions and marketing activities. Especially when the price of IOTA rises, the dynamics introduced by IIP-8 become particularly relevant. If validators then lower their fees, the decision-making process for delegators shifts away from APY alone toward additional services and utility. Otherwise, stake will tend to distribute more evenly across validators on average. This scenario has not yet materialized to date, as the token price has been too low. In our view, no matter how one looks at it, without genuine utility that drives the token price upward, it remains a challenging period for all validators — regardless of their ranking. Increased delegations mostly occur when validators are actively contributing to the network, which in turn incurs higher operational costs. Therefore, this issue affects nearly all validators across the board. In our opinion, this can only be sustainably resolved by bringing real utility into the network, rather than through other measures. |
|
Point taken - this was a framing error on my side, and I have reworded the post accordingly (v3.1). Calling the coupling an "IIP-8 extension" conflated two instruments that merely share an enforcement point. IIP-8, as you describe its intent, is a proportionate last-resort guard at the very edge: it limits commission dumping by the largest pools while deliberately leaving smaller validators room for introductory promotions. The coupling proposed here has a different job - preserving validator income while the subsidy steps down - so the post now presents Two substantive consequences of your point, both now explicit in the post:
On "only real utility resolves this": full agreement that utility is the endgame, and nothing here competes with it - the demand gate is built to self-deactivate exactly when usage arrives (a rising burn ratio automatically halts further reductions), and under the usage-coupling variant every reward recipient becomes directly invested in adoption. But waiting for utility at ~5.7% dilution is not free: ~280M IOTA/year of new supply meeting thin demand is itself a headwind to the very price recovery your analysis says operators need. The two levers point the same way. |
|
In general, it’s quite difficult to follow everything — the proposal isn’t really well presented or easy to digest. A few clear graphs or diagrams would definitely help to better understand the implications. We also noticed some errors in the validator basics like already mentioned (now corrected on your side), which doesn’t give us a good feeling at the moment that the rest of the proposal is well thought through. Additionally, this suggestion appears to come mainly from just two people in the community (correct us if we misunderstood that point). We couldn’t find any public poll or broader survey on the topic. (that it came from the community - that could be misinterpreted) At the moment we see no viable option to reduce the validator rewards (based on demand of the network) as proposed — especially since you have already calculated this yourselves. Therefore we do not support this approach right now. This might change once we hear more input from other validators. When the price rises to a good level, validators will surely lower their fees. Once the losses from bad times are covered. Most Validators bring real value to the network in many different ways, and they set their fees based on their own calculations and business models. Almost every validator now has an active community, and these have grown significantly in recent months. So it makes sense to wait and see what the broader validator community thinks. |
|
Thanks for taking the time to write this up. Fair points on presentation - let me answer them in order. Presentation. You are right, and @alexsporn said the same: the body has grown too dense to digest. I have posted a condensed summary with diagrams as a separate comment below. The full body stays as the reference document; the summary is the entry point. The spec error. It was real, it is fixed, and the changelog documents it openly. Every income and stake figure in the body is reproducible from public RPC fields (methodology note in the body), so I would ask that the numbers be judged on whether they check out - and if anything else is off, I want to know. "No public poll." A public poll exists: on July 9 I submitted a Phase-1 governance proposal with a token-holder poll to govern.iota.org, per the official governance framework. It is still sitting in the moderation queue - the forum has been largely dormant - which is exactly why this technical discussion runs here in parallel. I have asked the moderators to unblock it and will link it here once it is public. On "two people": the text has one author (me), with credited input from @alexsporn, DLT.GREEN, and two Discord community members (Texos, Dr.Electron). That is the same path IIP-8 took: a single community author building consensus in the open. The IIP process defines this discussion as the consensus-building step; it does not require a prior survey. The core point, as plainly as I can put it. The proposal does not reduce validator income. The commission-floor coupling is calibrated so that no validator earns less than today (the no-loss property in the body); the reduction comes out of delegator yield and net inflation. Under the worked example, DLT.GREEN's own commission income rises about 25 percent, because your 8 percent commission sits below the coupled floor of ~31 percent. So "no viable option to reduce the validator rewards" argues against a design that is not on the table - and if the write-up made you read it that way, that is on me, which is what the summary below is for. On "commissions will come down when the price recovers": agreed, that is the healthy dynamic. c_min(S) is that same dynamic made explicit and symmetric, instead of left to hope during the downside. More validators. Fully agreed - operator input has been Open Question 4 since day one. If you can point other operators to this thread, please do: their cost data and their views are exactly what is missing. |
The proposal in 90 seconds - diagrams included@alexsporn and @dlt-green both asked for a more digestible presentation - fair. Here is the entire proposal in two diagrams, one table and five numbers. The body above remains the reference; nothing below is new. 1. What is proposed, in one sentenceStep the fixed 767k/epoch subsidy down only while on-chain demand grows, and couple the validator commission floor upward at the same time - so validator income stays at least constant by construction, and the reduction comes out of delegator yield and net inflation, not out of operator pay. 2. Where the subsidy goes todaypie showData
title Annual subsidy split (~280M IOTA/year, epoch 435 on-chain data)
"Delegator staking yield" : 92.7
"Validator commissions (all 66)" : 7.3
The emission is, in practice, a delegator yield program: validator operations receive 7.3 percent of it. This is why naively capping the subsidy (v1 of this proposal) was wrong, and why the coupling below exists. 3. The no-loss mechanic, in numbersWorked example from the body: subsidy 767k -> 250k per epoch (-67 percent), commission floor c_min rises 10 -> ~31 percent. All incomes from epoch-435 on-chain state (votingPower x effectiveCommissionRate; method in the body).
No validator earns less than today: the floor is calibrated so that today's highest effective commission (10 percent) maps to exactly unchanged income, and everyone below it gains. c_min activates with the first step-down - at today's 767k it is not binding, so nothing changes for anyone until demand actually moves the gate (this also keeps IIP-8's promo flexibility for small pools intact until then). What shrinks instead: delegator APY ~10.4 -> ~2.6 percent, gross emission ~5.7 -> ~1.9 percent of supply per year. 4. How the gate worksflowchart TD
A[Epoch boundary] --> B{Demand gate:<br/>trailing burn/subsidy ratio<br/>above threshold?}
B -- yes --> C[Subsidy steps down one notch]
B -- no --> D[Subsidy holds]
C --> E[c_min rises by the same factor:<br/>validator income preserved]
E --> F{Staked share below<br/>security floor?}
D --> F
F -- yes --> G[Subsidy re-expands]
F -- no --> H[Continue]
C -. never below .-> I[(Absolute floor S_abs:<br/>full operator break-even,<br/>governance parameter)]
5. Five numbers to keep
6. What this is not
Full derivations, assumptions and the changelog of corrections are in the body above. If a number looks wrong, say so - everything is reproducible from public RPC fields. |
|
Your comparison isn't quite accurate. Ankr only needs to adjust their commission to 10% — they themselves chose such a low commission. Additionally, DLT.GREEN has already calculated the commission, so it's our decision to set it at 8% instead of 10%. |
Primary-source update: the July 15 Foundation AMA speaks to this proposal's premisesYesterday's IOTA community AMA (X Space, recorded: https://x.com/i/spaces/1kJzDDAbVYRKv) contained three statements by Dominik Schiener that bear directly on this thread. Quotes from the replay, timestamps in brackets. 1. The price-security coupling is now on the record from the Foundation itself.
That is the mechanism this proposal starts from: the subsidy is denominated in IOTA while operator costs are fiat, so when the price compresses, the emission stops covering what it was calibrated to cover (the two-tier cost analysis in the body), while dilution continues at the full fixed rate. 2. The Foundation itself does not expect usage to close the burn/subsidy gap any time soon.
For this proposal that cuts both ways, and both halves are worth stating plainly: (a) the demand gate would remain inactive for now -- by design, nothing changes for anyone until the trailing burn/subsidy ratio actually moves; and (b) the status quo keeps emitting ~280M IOTA per year into that same low-demand environment for however long commercialization takes. The asymmetry between (a) and (b) is the case for gating. 3. The fee/burn side is explicitly named as a future lever.
The "raise the fee/burn side instead?" section in the body treats fees as a complement to a demand-gated subsidy, not a substitute. The Foundation now names that lever itself. The two compose cleanly: any future fee increase or TWIN-IOTA fee sharing raises the burn/subsidy ratio -- which is exactly the gate metric of this proposal. Adoption-side progress would translate into emission-side progress automatically, with no further governance action needed. For completeness, the AMA also confirmed: roughly 25 percent of the Foundation team was let go (about 85-90 people remain), more than two years of runway, and first TWIN Cloud clients expected in about two to three months. None of this changes the numbers in the body; the epoch-435 income analysis stands as posted. To be clear about scope: nothing in the AMA endorses this proposal. But premise 1 (price-security link), premise 2 (burn lagging subsidy for the foreseeable future) and the fee-side lever are now stated by the Foundation in a recorded public session, which makes them a firmer basis for this discussion than my own estimates were. |
|
I think the issue could be largely resolved simply by automatically distributing 10% of the staked delegated tokens to all validators. |
|
Withdrawing this proposal and closing the discussion. The related work is shut down as well: the doggy-dex.com gas station pilot is offline and its repository is deleted, so the links in my earlier comments no longer resolve. The DeepBook port I had been preparing is cancelled too. The reason, plainly and with no hard feelings toward anyone who took part here: there is currently no venue in which a question like this can actually be settled. The parallel Phase-1 proposal on govern.iota.org has been sitting in the moderation queue since July 9 and was never published - that forum has been effectively dormant for months with a single moderator. Here, the Foundation engaged once and then went quiet. The July 15 AMA described an organisation in survival mode: roughly 25% of staff let go, no public roadmap, attention elsewhere. Whatever one thinks of the emission question on the merits, a proposal process needs someone on the other end of it, and I no longer believe there is one. Thanks to @dlt-green for the operator cost figures and the pushback - that was the most useful input this thread produced - and to @YUICHI-00, whose alternative deserved a better answer than it is now going to get. The full argument, chain data and worked examples stay in the post above for anyone who wants to pick it up later. |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
What is proposed
Three coupled parameter changes, delivered together behind a protocol version bump:
This is deliberately a package: cutting the subsidy alone would break validator economics - the numbers below show why.
Current state (live chain data, epoch 435, July 14, 2026)
767_000 * 10^9nanos per 24h epoch - verifiable live viaiota_getProtocolConfig(protocol version 29:validator_target_reward = 767000000000000; source:iota-protocol-config), no decay schedule. Annualized ~280M IOTA = ~5.7% of the 4.93B total supply (supply live viaiotax_getTotalSupply).subsidy - computation_charge_burned(match_computation_charge_burned_to_validator_subsidy). The computation-fee burn is the only structural supply sink - storage deposits are locked and refundable, and optional tips (a pass-through to validators, not a sink) are ~0 today. The burn currently offsets only a tiny fraction of the subsidy.effectiveCommissionRate = max(set_commission, VP%), bounded at 10% by the voting-power cap. 15 of 66 validators currently set 3% or less (9 of them among the 15 largest pools); the VP-linked floor lifts only the very largest of those back to ~3-5%, because it scales with pool size.effectiveCommissionRate, total validator commission is ~20M IOTA/year (7.3% of the subsidy) - 92.7% of all rewards accrue to stakers, i.e. ~10.7% net APY.votingPowerandeffectiveCommissionRatefields (epoch rewards are distributed proportional to voting power; the validator's cut iseffectiveCommissionRate x pool rewards- seevalidator_set.move). Fiat conversions use ~USD 0.037 / EUR 0.032 (July 14, 2026); fiat figures move with the token price, the mechanism itself does not depend on the price level.Why change anything
Validator economics today: who is actually paid what
(This section responds directly to the cost-coverage question raised in the comments, and folds in the first real operator cost figure shared there.)
Rough revenue model: a validator's service income is the commission on its pool's rewards. Yield on self-stake is return on the validator's own capital and is treated separately.
Cost side - two tiers, because operations differ widely (operator feedback in the comments made clear a single range is too coarse):
Conclusion: the status quo funds neither tier by design. The median validator's commission (~USD 9k) does not cover even Tier 1 at current prices, smaller validators are further underwater, and even the #3 pool's commission (~EUR 39k) sits below its operator's stated Tier-2 costs. Notably, none of the three indexer operations is actually funded by the subsidy: one runs on foundation budget, one on a commercial API business (its effective commission of ~190k IOTA/year ~ EUR 6k clearly does not pay for an indexer), one on a top-3 commission stream that still falls short. Funding all three explicitly at EUR 50k each would cost ~4.7M IOTA/year - under 2% of current emission. A stake-proportional subsidy is simply not aimed at operations: it pays for stake, wherever it sits. Any reduction must therefore restructure who receives the subsidy, not just shrink it - that is what the coupling below does. Differentiated services on top of the security footprint are better served by targeted instruments (ecosystem/community funding), which can distinguish between operations the way a per-stake formula cannot.
Worth stating explicitly: this break-even is drift, not design - and the squeeze happened after the committee had committed. Most of today's validators onboarded in the first months after genesis (May 2025), when the token traded around USD 0.20; the fiat value of the same IOTA-denominated reward stream has since compressed roughly 5x. Over the same period the hardware class validators need became drastically more expensive in the AI-driven memory shortage (server DRAM contract prices up ~90% in Q1 2026 alone, NAND wafer prices up ~2.5x year-over-year), and chain-state growth pushes storage needs up over time. The squeeze has since reached retail hosting: Hetzner repriced on June 15, 2026 (its second increase that year) - dedicated machines meeting the validator's CPU/RAM spec now list at roughly EUR 557-612/month net (EUR 663-729 gross) even at the discount end of the market, dedicated-vCPU cloud lines rose up to ~176% in one step, and the remaining discounted "limited" models are available only while cheap hardware lasts. A fixed 767k constant tracks neither side: validator margins today are whatever price and hardware drift happen to leave over. Nothing in the current protocol ties the subsidy to the cost of operating the network - the floor and coupling below introduce exactly that link.
Mechanism
1. Demand gate. While the trailing
computation_charge_burned / subsidyratio stays below a threshold, the subsidy steps down on a fixed schedule (per protocol upgrade or per epoch window). Rising burn halts further reductions automatically - the gate self-deactivates exactly when the fee market starts doing its job.2. Commission coupling - how validators keep covering costs. Pair each step-down with a higher protocol minimum commission:
Validator service income (
c x vp_share x 365 x S) stays constant whencscales with1/S. The floor activates only once the subsidy has actually stepped down - at today'sS_refthe protocol's commission rule applies unchanged. Worked example: subsidy 767k -> 250k (-67%) paired withc_min10% -> ~31%: the median validator's commission income is unchanged by construction; delegator APY drops from ~10.7% to ~2.6%; net inflation lands at ~1.8% p.a. - inside the target band. The cut is borne by staker yield, and early feedback (see comments) includes stakers explicitly fine with low rewards while usage is low.No-loss proposition. Under a step-down from
S_reftoSwith the coupling above, a validator whose effective commission today isc0sees its commission income multiply by(S / S_ref) x (c_min(S) / c0) = c_ref / c0(forc0 <= c_min(S) <= c_cap, at constant voting-power share). Since the highest effective commission on the network today is exactlyc_ref = 10%(the VP cap), no currently active validator's commission income falls: operators at 10% stay exactly constant, everyone below gains. Concretely, the #3 validator (8%) would earn ~+25% (~1.53M IOTA/year - at today's price almost exactly the EUR 50k Tier-2 figure named in the comments), and a 2.6%-effective operator nearly 4x. To be equally explicit about the flip side: yield on a validator's self-stake falls like any staker's - that is the point of the proposal, not a side effect. And where the protocol's VP-linked floor deliberately bites only for the largest pools (a last-resort anti-dumping guard, per its authors),c_min(S)behaves differently by design: at today's subsidy it binds nobody, and as the subsidy steps down it becomes uniform - near-zero commission (15 of 66 set at or below 3% today) stops being a viable competitive weapon at a reduced reward pool, so delegators increasingly choose on service quality and trust. Whether the smallest pools should instead be phased in gradually - preserving the introductory-promotion flexibility the current design intentionally gives them, at the cost of the income guarantee while they use it - is open question 8.3. Security floor / re-expansion. If staking participation falls below a floor (e.g. 30% of supply - calibration open), the subsidy steps back up. Combined with (2), the system searches for minimum viable issuance instead of hard-coding it: if ~2.6% delegator yield does not sustain participation, the mechanism finds that out and corrects itself. Note that delegator exit partially self-corrects even before the floor triggers - less total stake means higher per-stake rewards for everyone who remains.
4. Absolute viability floor. Even at 100% commission, the subsidy must keep a median validator at break-even on the Tier-1 footprint:
With Tier-1 costs of USD 20k/year ~= 546k IOTA at the July 14 price: S_abs ~= 170k IOTA/epoch (~1.3% p.a. gross). A week earlier, at USD 0.041, the same model gave ~152k - the floor moving ~10% with a week of price action is precisely why it cannot be a fixed on-chain formula. It is a governance-calibrated parameter, revisited per protocol upgrade, ideally informed by operators publishing real cost data (open question 4). A practical floor sits above
S_abs, e.g. ~250k/epoch (~1.9% p.a. gross, still inside the target band) with the correspondingc_minat ~31% - the gap is deliberate headroom so that full-service operations (Tier 2) have room above bare break-even.Delivery. Changes to
validator_target_rewardhandling in protocol config plus epoch-boundary logic in theiota-systemMove package, feature-gated behind a protocol version bump. No changes to consensus, committee selection, staking pool mechanics, or reward distribution paths beyond the commission minimum.What this does NOT change
Design variant: continuous usage-coupling
Instead of stepped reductions, tie the subsidy directly to the demand signal, capped at today's level:
Simpler to specify and harder to game; more reward volatility for validators, which the commission coupling and the floor absorb. Side effect worth naming: every reward recipient becomes directly invested in mainnet adoption. (Suggested in community feedback - see comments.)
A second alignment argument: operating costs themselves scale with usage (state growth, bandwidth, storage upgrades), so usage-coupled income tracks the very driver of validator costs - whereas a fixed subsidy is at its most generous, relative to cost, exactly when the chain is emptiest.
Alternatives considered
Precedents
Relation to community governance
A Phase-1 proposal ("Pause IOTA Token Inflation Until Sustainable Demand Is Established") has been submitted on govern.iota.org and is awaiting moderation - intended as a token-holder mandate track parallel to this technical discussion. I will cross-link once the forum topic is public.
Open questions
c_min(S)phase in by pool size, preserving the introductory-promotion flexibility the current commission design intentionally leaves to the smallest pools? (Raised by the IIP-8 authors in the comments.)Feedback very welcome - especially real cost figures from validators, and views from core contributors on whether the gating and coupling rules belong in protocol config or in the system Move package.
Changelog
votingPowerandeffectiveCommissionRatefields; added the no-loss proposition (with the self-stake caveat made explicit); added the June 2026 hosting repricing to the drift evidence; absolute floor recomputed at the July 14 price with sensitivity noted; open question 4 marked partially answered.All reactions