Two numbers in the Bank of Russia's draft crypto capital framework matter. Only one made it into the headline. The first is 1% โ an exposure ceiling for crypto, expressed as a share of a credit institution's own funds. The second is 1,250% โ the risk weight applied to proprietary crypto positions and to any client position where the bank carries the loss. Coverage sold the first. The capital math lives in the second.
Under an 8% minimum capital adequacy ratio, a 1,250% risk weight requires capital equal to 100% of the exposure. It is a full deduction. Not a haircut, not a buffer โ a write-off of regulatory capital for every unit of proprietary crypto a bank holds. Framed that way, the "harsh 1% cap" everyone is debating is almost decorative. The binding constraint is the weight, and the weight is a prohibition wearing the vocabulary of prudence.
The proposal is a regulatory capital rule, not a product. Two ratios anchor it. N31 measures crypto exposure at the level of an individual credit institution, with the denominator set to that institution's own funds. N32 measures the same exposure at the consolidated banking-group level, denominated in group capital. The design is legible: bind the denominator to the entity that actually bears the risk, so a group cannot inflate its tolerance by shuffling capital between subsidiaries. It is the kind of structural pairing I look for first when I review a mechanism โ the accounting unit and the risk-bearing unit are deliberately aligned, not left to be reconciled later.
The rule remains a draft. Publication is slated for the fourth quarter of 2026, effectiveness ten days later, and mandatory N31/N32 reporting from January 2027. A rule that does not bind for two years is a signal, not a fact. History repeats in the ledger, not the news โ and there is no ledger entry yet. Treat everything that follows as the shape of an intent, priced at zero until the final text prints.
The reference frame is the Basel Committee's 2022 crypto standard. Basel splits crypto into Group 1 โ tokenized assets passing strict classification tests โ and Group 2, the catch-all that includes bitcoin. Group 2 exposures are capped at 1% of Tier 1 capital; the least-qualified slice, Group 2b, carries a 1,250% risk weight. Russia's 1% ceiling and 1,250% weight map onto that template with high fidelity. This is not improvisation. It is localization of a global standard, with two native additions: a custodial carve-out and the inclusion of foreign digital instruments in the numerator.
The numerator is where the engineering lives. Covered exposure is not limited to direct coin holdings. It extends to direct and indirect investments, price-linked derivatives, and any loan, bond, guarantee, repo, or credit line whose settlement or value depends on crypto prices. This is anti-circumvention architecture. It closes the obvious escape routes โ a bank that wants crypto beta cannot book it through a credit facility or a structured note and pretend the exposure is something else. I have seen this failure mode before, from the inside. In my 2021 Zerion liquidity-mining assessment, I traced 15,000 transactions and found that the participants who believed they held no token risk were precisely the ones holding it through wrappers. Regulators who do not define the numerator precisely end up regulating the label, not the risk. Russia defines it.
The inclusion of foreign digital instruments deserves its own paragraph, because it is the quietest signal in the draft. By capturing offshore stablecoins and non-resident digital rights in the exposure calculation, the rule limits a bank's ability to gain crypto exposure through instruments issued outside the ruble system. Read alongside the state-bank key arrangements elsewhere in the proposal, the design points toward a controlled domestic channel rather than an open one. This is capital control expressed as capital adequacy โ the same instrument, a different face. Volatility is being measured, but the measurement is also a perimeter.
Netting and hedging get the same tight treatment. Long and short positions may offset only within qualifying lower-risk categories, and only subject to conditions on collateral, settlement, maturity, freezing, and liquidity risk. Direct holdings and higher-risk exposures cannot be fully hedged away. Contrast that with Canada, whose 2027 capital rules provide relief for positions hedged across exchanges. Russia chooses the conservative branch and refuses the relief. The math holds until the incentive breaks โ and here the design deliberately refuses to let banks arbitrage the arithmetic by pairing a long onshore with a short offshore and calling the net zero.
Then comes the split that defines the whole framework. Non-liable client custodial positions โ crypto the bank holds but does not itself stand behind โ receive a 50% risk weight. Proprietary exposure and any client position where the bank carries the loss receive 1,250%. Everything hinges on a single legal question: who bears loss upon seizure or restriction? The draft invokes "digital depository institutions" as an eligible role, and the answer to that question decides whether a given position lands in the 50% bucket or the full-deduction bucket. There is no middle category. The framework is a cliff, and the cliff is drawn in prose.
Read the two weights as an incentive gradient and the policy intent is unmistakable. The framework makes custody cheap and speculation expensive. A custodial book costs roughly 4% in capital โ 8% capital ratio ร 50% weight. A proprietary book costs 100%. For any bank optimizing capital, the choice is not close. The rule channels institutions into the role of pipes, not principals โ precisely the design a skeptical central bank would choose. In a bear market where balance-sheet survival matters more than upside, this is a defensible regulator's instinct: keep crypto volatility entirely outside the capital that protects depositors. Volume masks the insolvency structure, and the fastest way to strip volume is to make it capital-prohibitive.
The 1% cap compounds that gradient. If exposure cannot exceed 1% of a bank's own funds, then even a bank determined to hold crypto on-book faces a commercial ceiling so low that the business line barely justifies its compliance cost. The cap does not need to bite often; it needs only to make the scale unattractive. Between the exposure ceiling and the risk weight, proprietary crypto at a Russian bank is not regulated into safety โ it is regulated out of existence. The two mechanisms are redundant on purpose.
But the framing that this "protects customer assets" deserves a forensic pass. Audits verify logic, not intent. The logic here says the protection is conditional, not absolute, and it rests on three preconditions holding at once. First, the bank must genuinely not carry the client's loss โ if it does, the position is marked at 1,250% and the "custody" is proprietary risk in disguise. Second, the depository must remain solvent and operationally sound. Third, the loss-bearing classification must survive legal dispute, because the draft hands supervisors a qualitative judgment with no bright-line test.
That third condition is the soft underbelly. Economically the rule is coherent โ 50% versus 1,250% is a clean binary. Legally it is blurry. "Who bears loss" is a contractual question with a hundred edge cases: when does a frozen withdrawal become a bank loss? When does a custodian's operational failure become the bank's liability? The framework delegates those determinations to supervisory judgment, which means the effective capital charge for a given position may not be knowable in advance. Complexity of that kind is a known failure surface. My fault-proof review of the Arbitrum One bridge found a latency bottleneck that theoretical models had not surfaced; the gap between a rule as written and a rule as enforced is where risk hides. Risk is a feature, not a bug, until it is neither disclosed nor priced.
And then there is the concentration problem. The draft contemplates state banks holding keys. Stack that against the 50% custodial weight and the end-state is not a distributed custodian network but a small number of systemically important, state-aligned depositories holding client crypto. That converts an operational-risk question into a systemic one. Consensus is code, but code is fragile โ and key custody concentrated in three institutions is not code at all. It is a single point of failure with a central-bank logo.
The 50% weight is low enough to allow custody; it may be too high relative to the near-zero weight of cash and sovereign paper to make custody attractive. If compliance costs are heavy โ and N31, N32, plus turnover reporting from 2027 are heavy โ banks may simply decline the business. The compliant channel stays cold while users route around it.
Two more threads are worth pulling. The reporting regime โ N31, N32, and turnover data from January 2027 โ is the only channel through which the rule becomes verifiable. A capital rule that cannot be observed cannot be enforced, and turnover reporting is the tell that the central bank expects to monitor flow, not just stock. That is a surveillance posture, and it aligns with the rest of the design. Second, the correlated-risk question I stress-tested in my 2025 EigenLayer restaking simulation applies here in miniature: a rule that treats each bank's exposure as independent will understate the systemic effect when a handful of state depositories share the same operational assumptions. Individual compliance does not imply collective safety. Correlated custodians are correlated failures.
That is the likely equilibrium, and it should be said plainly. The 1% cap plus the 1,250% weight remove banks from proprietary crypto almost entirely. Custody survives only if the economics clear a 4% capital hurdle plus a reporting load. Russo-crypto demand is driven by sanctions evasion and store-of-value needs; it does not vanish when banks step back. It migrates. The rational outcome is a two-track system: a quiet, monitored, state-adjacent channel, and a busy, gray, offshore one. Watch the final text on key custody and on the loss-bearing test โ not the 1%. Liquidity is borrowed time, and Russia is borrowing it against a channel it cannot yet see. The window to read the final text is the third quarter of 2026, before publication. Until then, the 50-versus-1,250 split is a hypothesis, not a rule โ and the market is pricing it as neither.