TL;DR
- The evaluation asks whether or not Solana can materially cut back the rent-exempt requirement.
- There is no such thing as a proof of an instantaneous threat of state development: the tied-up capital wanted for a state-bloat assault stays excessive.
- Present hire seems to include substantial financial headroom.
- Quick-term account demand is simply weakly associated to adjustments within the USD-denominated price of hire.
- Most gross account-state allocation is transient reasonably than persistent.
- Current web storage development stays reasonable.
- Decrease hire would materially cut back onboarding prices for account-intensive functions, notably funds.
Introduction
State development is likely one of the elementary scalability constraints of a high-throughput blockchain. In contrast to transaction throughput, account state is persistent: each validator should repeatedly retailer, index, snapshot, and keep it. Consequently, extreme state development impacts not solely storage capability, but additionally I/O, snapshot era, restart instances, and long-term validator working prices.
The hire parameters have been calibrated below a special working setting. Since then, validator {hardware}, operational expertise with AccountsDB, and the financial worth of SOL have all modified considerably. Consequently, it’s affordable to revisit whether or not the present hire stays calibrated to the infrastructure constraints it’s meant to guard.
Hire discount has been mentioned in
How A lot Capital Would a State-Bloat Assault Require?
Below the assumptions of the mannequin described right here, a tenfold discount within the rent-exempt requirement would nonetheless require roughly $17 million of capital to devour the presently obtainable AccountsDB storage headroom. By doubling present really useful necessities (2TB), this goes to $51 million of capital. It is a significant deterrent, however it’s a capital requirement reasonably than a assured financial loss, and it doesn’t account for doubtlessly exterior attacker payoffs.
A decrease rent-exempt requirement reduces the capital that customers should immobilize to keep up official state. The identical discount, nonetheless, additionally lowers the capital required to maintain malicious state dwell. The related safety benchmark is subsequently the quantity of capital wanted to broaden AccountsDB sufficient to materially pressure validator infrastructure.
For an economically motivated attacker, the fundamental deterrence situation is:
the place is the profit obtained if the assault succeeds and is the related financial burden.
Within the case of a state-bloat assault, the time period “price” requires some care. The rent-exempt steadiness is just not usually burned: it’s capital locked contained in the accounts and may typically be recovered when these accounts are closed. The related amount is subsequently higher described as a capital requirement, along with its alternative price, execution charges, and any portion of the principal that might not be recovered.
A state-bloat assault doesn’t provide an apparent protocol-native payout. Disrupting or halting the community doesn’t straight switch worth to the attacker. Nonetheless, the potential profit can’t be assumed to be zero, since it might come up externally by brief positions, derivatives, liquidations, extortion, aggressive benefit, or non-economic goals. The aim of the next mannequin is subsequently to not show that an assault is unprofitable, however to estimate whether or not hire continues to impose a considerable financial barrier.
A easy attack-cost benchmark
Let the commodity storage price be:
Assuming that persistent state is replicated throughout roughly 1,000 validators:
Making use of a security multiplier for operational overhead and financial deterrence offers:
or roughly
This isn’t meant to characterize the precise marginal infrastructure price imposed by one extra byte. It’s a conservative benchmark for the financial barrier that hire ought to keep.
Comparability with present hire
Determine 1 exhibits the evolution of common USD worth per byte, with present rent-exempt requirement at roughly
Fig.1: USD worth per byte saved on Solana. Supply: Dune question.
It follows that
Thus, below these assumptions, a tenfold discount within the rent-exempt requirement would carry the capital requirement near the benchmark derived above.
How a lot capital would a state-bloat assault require?
The present AccountsDB storage recordsdata occupy roughly 495 GB (see Fig.2), in opposition to a really useful capability of 1 TB. The remaining storage-file headroom is subsequently
Fig.2: Left: whole dwell account knowledge (accounts_data_len), measuring the cumulative dimension of account knowledge presently saved on-chain. Proper: whole AccountsDB storage footprint (accounts_db-stores.total_bytes), together with the bodily storage occupied by AccountsDB recordsdata.
Nonetheless, hire is paid on allotted account state reasonably than straight on AccountsDB storage-file bytes. Utilizing the presently noticed storage amplification (see Fig.2):
the corresponding extra dwell account state could be:
On the lowered fee
This capital would wish to stay locked for so long as the malicious state stays dwell. The rent-exempt reserve is mostly recoverable when an account is closed. It ought to subsequently be interpreted as capital in danger or capital immobilized, reasonably than as an irreversible payment.
What Is Really Driving Solana State Development?
The attack-cost mannequin offers a theoretical certain on the capital required to keep up malicious state. We now flip to the noticed behaviour of account state below regular community exercise.
Does Account Exercise Reply to the Value of State?
Account creation doesn’t seem to reply mechanically to short-term adjustments within the USD price of state. The noticed correlations are weak and regime-dependent, though they don’t present a causal estimate of the impact of adjusting the hire parameters.
It is a related query since if account creation responds strongly to the USD-denominated price of hire exemption, lowering hire might speed up state development. If the response is weak, short-run development is extra prone to be dominated by application-specific demand.
The rent-exempt requirement is denominated in lamports, however its efficient USD worth adjustments with the value of SOL. We use this variation as a primary descriptive check of whether or not account exercise responds to adjustments within the USD price of sustaining state.
During the last 30 days, the connection between SOL worth and the variety of accounts seems weak. When computed on 1-hour proportion adjustments, the 24-hour rolling correlation fluctuates round zero, with no persistent constructive or adverse regime (cfr. Fig.3). This means that short-term account-count dynamics will not be strongly defined by simultaneous actions within the SOL worth.
Fig.3: High: SOL worth denominated in USD in contrast with N° of Accounts. Backside: 24h rolling correlation at 1h decision.
A lagged comparability offers a extra nuanced image, see Fig.4. Particularly, the sharp decline in SOL worth at the start of the pattern is adopted, with some delay, by a marked discount within the variety of accounts. Equally, the following restoration in SOL worth is broadly adopted by a restoration in account rely. Nonetheless, the lagged rolling correlation stays unstable and oscillates over time, indicating that the connection is just not stationary and shouldn’t be interpreted as a easy linear dependence.
The information subsequently recommend some low-frequency co-movement, however not a secure lead–lag relationship that may be summarized by a single correlation coefficient.
Fig.4: High: SOL worth denominated in USD and shifted by 84h in contrast with N° of Accounts. Backside: 24h rolling correlation at 1h decision.
We subsequent examine each day adjustments within the hire worth per unit of account state with the each day proportion change in newly created accounts (not-deduplicated), cut up throughout account classes.
The scatter plot in Fig.5 exhibits no clear linear relationship. Most observations are concentrated round small rent-price adjustments, whereas account creation stays extensively dispersed throughout classes. Even when the USD price of state adjustments materially, the corresponding response in new account creation is heterogeneous.
Fig.5: Scatter plot of each day change in USD per account state vs new account creation (not-deduplicated). Supply: Dune question.
The 30-day rolling correlations assist the identical interpretation, see Fig.6. The coefficients change signal over time and sometimes attain reasonably constructive or adverse values, however no account class enters a persistent high-correlation regime.
Fig.6: 30 days rolling correlation between each day change in USD per account state and new account creation (not-deduplicate). Supply: Dune question.
Thus, on the each day frequency, gross account creation is just not tightly coupled to contemporaneous adjustments within the USD-denominated rent-exempt requirement.
Throughout each the entire account inventory and gross account-creation flows, the noticed relationship with SOL worth is weak and unstable at brief horizons. That is per protocol-specific exercise and on-chain demand enjoying a bigger function than short-term variation within the USD price of state.
Nonetheless, this evaluation shouldn’t be interpreted as a causal estimate of hire elasticity. SOL-price actions will not be an exogenous rent-policy experiment, and the pattern doesn’t reproduce the magnitude of a possible policy-level hire discount. The evaluation might also miss delayed responses, nonlinear thresholds, and adjustments in account lifetime reasonably than account creation.
Day by day State Demand and Account Measurement Development
Current exercise requests roughly 10 GiB of recent account state per day in gross phrases, however gross allocation is just not equal to persistent storage development. Within the cohort analyzed, 75.5% of account-creation occasions shut in the identical transaction. On the 1-million-slot horizon (~4.7 days), roughly 0.61 GiB stays out of an preliminary estimated 10 GiB, similar to a byte-survival share of 5.65%. Essentially the most persistent class (small custom-program accounts) represents solely a small fraction of gross state demand. After accounting for closures throughout each new and pre-existing accounts, the web dwell account-state footprint grows by roughly 0.3 GB per day.
Gross account creation measures the instantaneous demand for state allocation, however it might considerably overstate how a lot state stays dwell.
The each day required-space collection in Fig.7 measures the quantity of account state allotted by newly created accounts. These figures will not be deduplicated by account handle: every creation occasion is counted individually, so repeated create–shut cycles involving the identical handle could contribute a number of observations.
Fig.7: Day by day requested house by account creation divided by class. Supply: Dune question..
Gross state demand is dominated by token-account exercise. On 11 July 2026, for instance, newly created accounts requested roughly 10.23 GiB of estimated AccountsDB state:
- SPL Token basic accounts: 8.60 GiB
- {custom} or different program accounts: 0.79 GiB
- Token-2022 extension accounts: 0.58 GiB
- small custom-program accounts: 0.22 GiB
- Token-2022 base accounts: 0.038 GiB
SPL Token basic accounts subsequently represented roughly 84% of gross state demand on that day. Nonetheless, this stream shouldn’t be interpreted as a ten GiB each day enhance in AccountsDB: a lot of the newly allotted state is closed virtually instantly.
We focus solely on accounts that have been successfully closed in the course of the remark window. For every closed account occasion, we compute the length between creation and closure and bucket this length in days. The determine subsequently exhibits the lifetime distribution conditional on closure, not the survival distribution of all newly created accounts.
The distribution is strongly concentrated within the zero-day bucket: amongst accounts which are finally closed, a big fraction is created and closed inside the identical day, see Fig.8. The remaining mass decays quickly, with solely a small tail of closed accounts surviving for longer durations earlier than being eliminated.
Fig.8: Period buckets of non-deduplicated closed-account occasions during the last 180 days. The chart contains solely accounts that have been successfully closed. Supply: Dune question.
These counts are not deduplicated by account handle. The histogram ought to subsequently be interpreted as a distribution of account-lifecycle occasions reasonably than a distribution of distinctive accounts. Repeated create-close cycles could contribute a number of observations.
A big fraction of the short-lived account exercise seems to be associated to trading-driven workflows. Particularly, the dominant contributors to this create–shut sample are:
the place FD1a…1Dfw9 corresponds to an arbitrage bot that makes use of idem…PVg3Y to create and shut accounts.
The affect of those accounts on the dwell AccountsDB footprint is proscribed, as a result of the related state is generally transient. After we take a current-state snapshot utilizing getProgramAccounts, the corresponding dwell account footprint is small: roughly 0.46 GB and 0.0008 GB, respectively. Subsequently, whereas these workflows could generate substantial short-term account churn, they don’t materially enhance persistent account-state storage.
The broader each day account-creation breakdown in Fig. 9 exhibits that the biggest supply of newly created accounts is the SPL Token account class. This means that gross account creation is dominated by token-account exercise reasonably than by persistent {custom} program state.
Fig.9: Day by day newly created accounts by account sort. Counts will not be deduplicated by account handle. Supply: Dune question..
Measuring Account Survival
To estimate persistence, we observe a one-day cohort of non-deduplicated account-creation occasions from opening to the primary noticed closure. When an handle is recreated, every create–shut lifecycle is handled individually.
At every slot horizon , we embrace solely accounts with no less than slots of observable historical past, avoiding the classification of lately created accounts as survivors just because they haven’t been noticed lengthy sufficient.
We apply the identical calculation to estimated AccountsDB bytes reasonably than account counts. Within the analyzed cohort, 75.5% of creation occasions shut in the identical transaction, and solely 5.7% of the initially allotted bytes stay after 1 million slots, roughly 4.7 days. Persistence is strongly account-type particular: small custom-program accounts survive longer, however characterize solely a small share of gross each day state demand.
Fig.10: Survival fee for accounts created. Supply: Dune question. This doesn’t change primarily based on the day, see right here.
The cohort evaluation explains why gross allocation quickly decays, however it’s not itself the net-growth calculation. We will measure the web development from the mixture live-state steadiness and in addition contains closures of accounts created earlier than the analyzed cohort. This ends in roughly 1M new account per day (on common), for an estimated house in account DB of roughly 0.3 GB / day.
Fig.11: Price of closed account for accounts created from 1st of June 2026 and 1st of July 2026. Supply: Dune question.
How Quick Is AccountsDB Rising?
We match the latest upward development within the AccountsDB Storage Information Measurement metric to estimate the present native development fee, see Fig.12. Over the chosen post-dip interval, the fitted development fee is roughly 0.23 GB a day. This needs to be interpreted as an area linear estimate reasonably than a structural long-term forecast. We then extrapolate the AccountsDB storage footprint over a two-year horizon below completely different multiples of the present fitted development fee.
Fig.12: Left Panel: Match of the present development fee of AccountsDB dimension, taking solely the quickly rising case. Proper Panel: Forecasted AccountsDB dimension over time below completely different development fee.
This estimate is just not straight interchangeable with the earlier 0.3GB/day. The roughly 0.3 GB/day determine is reconstructed from account creation and closure flows, whereas the 0.23 GB/day estimate is an area match to bodily AccountsDB storage recordsdata over a specific post-dip window. Variations could mirror the remark interval, storage cleanup and shrink behaviour, and the excellence between logical dwell state and bodily file allocation.
Below the baseline development fee, AccountsDB storage would attain roughly 665 GB after two years. At 2x the present development fee, it could attain roughly 835 GB; at 3x, roughly 1.0 TB; at 5x, roughly 1.35 TB; and below a intentionally aggressive 10x state of affairs, roughly 2.3 TB.
Because of this even a development fee considerably above the present noticed development would stay inside a couple of terabytes over a two-year horizon.
Fig.13: Account dimension and snapshot dimension.
Compressed full snapshot archives are considerably smaller than the uncompressed AccountsDB storage recordsdata. Within the noticed knowledge, the total snapshot archive dimension is roughly 22% of the AccountsDB Storage Information Measurement.
Making use of this ratio to essentially the most aggressive state of affairs proven above offers roughly ~460-510 GB. This means that compression materially limits snapshot development, however it doesn’t set up that 500 GB is ample for an operational snapshot quantity as soon as retention, incremental archives, momentary recordsdata, and security headroom are included.
What Occupies Solana State Immediately?
Dwell Solana state is concentrated in a couple of storage-heavy classes: SPL Token accounts are the only largest element, OpenBook/Serum stays one of many largest persistent DeFi state clusters, and NFT infrastructure nonetheless occupies a considerable share of dwell bytes.
Determine 14 highlights an vital distinction between account rely and state footprint. Particularly, OpenBook/Serum accounts characterize solely a really small share of all accounts, but they occupy roughly 30% of the dwell state within the snapshot. This strongly means that the primary driver is just not merely adoption, however bytes per account. That interpretation is per the protocol design: a Serum/OpenBook market is just not a single compact pool account, however a heavier state object requiring separate persistent accounts for the market itself, request queue, occasion queue, bids, and asks, with buying and selling exercise additionally counting on OpenOrders accounts. In different phrases, the orderbook structure shops substantial market construction straight on chain.
Fig.14: Solana Account Knowledge Measurement Snapshot at epoch 997.
The second main result’s that SPL Token is the biggest state class within the present snapshot. This isn’t shocking, since token accounts are the fundamental storage primitive used throughout wallets, DeFi, airdrops, launchpads, and token issuance extra broadly. A notable share of this token-state footprint seems to be associated to Pump.fun-style belongings: utilizing a suffix-based heuristic, roughly 30% of SPL-Token house seems related to Pump.fun-related tokens. This attribution needs to be handled cautiously, nonetheless, as a result of naming-based identification is imperfect and any issuer can generate a token with an identical suffix. The interpretation is subsequently indicative reasonably than definitive: the result’s greatest learn as proof that the memecoin issuance ecosystem is a significant contributor to dwell token-account state, not as a precise measurement of Pump.enjoyable’s distinctive footprint.
Lastly, the snapshot exhibits that the NFT stack stays a big supply of persistent storage. Most of this footprint seems to return from Metaplex Token Metadata, which is per the truth that NFT ecosystems require greater than token balances alone: additionally they keep metadata, assortment construction, and associated indexing objects. This makes NFTs structurally extra storage-intensive than easy fungible-token balances, even when transactional exercise is decrease than in DeFi or token-launch workflows.
What Decrease Hire Would Change for Funds
A discount within the rent-exempt requirement would scale back the SOL required to initialize new accounts by 90%. For account-heavy functions, this is able to materially decrease the capital wanted to onboard customers and keep persistent state. For fee flows, the price of creating a regular SPL token account would transfer from roughly $0.16 to $0.016 on the SOL worth used on this evaluation. This can have enormous implications on segments like funds.
On Solana, a pockets doesn’t maintain SPL tokens straight. It usually makes use of a deterministic Related Token Account for every token mint. If the recipient doesn’t but have the suitable account, the sender can create it in the identical transaction because the switch, however should additionally fund its rent-exempt steadiness. As soon as that account exists, subsequent funds in the identical token don’t require creating it once more. Account creation is subsequently primarily a one-time onboarding price per pockets and mint, reasonably than a value paid on each fee.
On the pattern SOL worth of roughly $77.94, the present hire deposit for the standard SPL token account is:
With a discount it turns into $0.0159.
This has an enormous implication on segments like funds. If we examine the lowered distribution of hire with Tempo, a payments-first Layer 1 purpose-built for stablecoin funds, we see how Solana locations prices for account creation within the decrease tail of Tempo’s noticed account-creation prices, see Fig.15. Regardless of Tempo offers stablecoin-denominated fuel charges and devoted fee blockspace, explicitly concentrating on predictable fee prices with out requiring customers to carry a unstable fuel asset, we are able to additionally see how Solana will end in a way more secure price, exhibiting a lowered volatility.
Fig.15: Distribution of account-creation prices on Tempo and Solana below the present and 10×-reduced hire schedules. Solana values mix transaction charges with the rent-exempt deposit; the latter is recoverable when the account is closed, whereas Tempo fuel represents an execution expense. Supply on Dune.
