The anomaly was not the classification. It was the clause that was never written.
In February, the Russian government created a new tier in its power-supply reliability taxonomy. Category One belongs to hospitals, defense plants, and critical infrastructure. Category Three is the traditional floor for ordinary industrial consumers — guaranteed supply, subject to scheduled interruption at peak. The new Category Four sits beneath that floor, and its only members are cryptocurrency miners. The Ministry of Energy has since confirmed the operative detail: Category Four loads can be disconnected at any moment, without prior consultation. No trigger threshold. No notice period. No restoration procedure. Silence was the first warning sign.
I have spent enough time inside protocol specifications to distrust what a system advertises and trust only what it enforces. A governance document that lists a right but omits the remedy is not granting a right — it is reserving a discretion. That is the correct lens for this policy.
The taxonomy behind the switch
Russia's grid reliability categories are not marketing. They are operational dispatch rules, and they determine who gets shed first when generation falls short of load. Historically the system ran on three classes. Category One receives continuous, protected supply. Category Two tolerates brief, scheduled interruption. Category Three — the old floor — accepts longer interruptions but still carries a guarantee of eventual restoration. Category Four, created by government decree in February, is the first tier designed to carry no restoration guarantee at all.
This sits in apparent tension with Russia's 2024 legalization of industrial mining. The contradiction resolves once you stop reading legalization as endorsement. Legalization defines who may operate; the reliability classification defines whether they may operate continuously. Russia chose to legalize the business and to render its continuity revocable.
Russia's 2024 mining law did more than decriminalize the activity. It installed a registration regime, restricted retail participation, and layered a tax on realized proceeds. What it did not install was any guarantee of the power that makes the activity possible. That omission is now formalized in Category Four — and it is the tell. A state that intended to protect its mining industry would have secured its energy supply first.
The instrument is a single term: "technological connection." Category Four permits miners to connect to the grid even in regions where mining has been restricted for power shortage — but connection is explicitly decoupled from reliable supply. A miner can be physically wired into the network and still receive nothing it can count on. This is not a subtle distinction. It is the entire architecture of the policy compressed into two words.
Russia matters here because of scale. It is the second-largest source of Bitcoin hashrate, historically around ten percent of the global total, underwritten by cheap Siberian hydro. When the second-largest supplier of proof-of-work security has its power supply downgraded from guaranteed to discretionary, the change is small in global terms and structural for everyone inside it.
The political economy underneath is regional. Oblasts with surplus hydro and stranded gas have lobbied to host miners, while deficit regions — much of Siberia and the Far East in winter — have pushed back. Category Four is the federal compromise, granting access where power is abundant and reserving the right to sever it where power is scarce.
Interruptible load without the contract
Grid operators have a name for loads like miners: interruptible. A data center cannot be shed without consequence; a smelter cannot restart cheaply; a hospital cannot be shed at all. A mining farm can be switched off in seconds and restarted with no human harm. That elasticity makes miners the ideal demand-response resource — and it is precisely why the grid's logic for placing them at the bottom is internally coherent. This is not a bug in the policy. It is the design.
But conventional interruptible-load arrangements are contracts. They specify a trigger, a maximum curtailment frequency, a notice window, and — critically — compensation in the form of a discounted tariff for accepting the risk. Category Four supplies the curtailment right and none of the corresponding obligations. It is interruptible load without the contract: the operator's discretion is retained, the miner's remedy is deleted.
Three implementation details are missing from the public record, and each one is load-bearing. The restoration mechanism: once a Category Four load is shed, what guarantees it comes back, and on what timeline? A switching-frequency cap: repeated cycling destroys ASIC hardware and stresses power delivery components, yet no ceiling has been disclosed. And a self-generation carve-out: whether a miner may insulate itself with on-site generation or storage, or whether the category applies regardless of source. The proof is in the unverified edge cases.
The invariant that leaks
I model these things the way I modeled Curve's StableSwap invariant in 2020 — by writing the algebra and watching where the assumptions break. The relevant invariant here is simple. A miner's effective cost per unit of hashrate is nominal power cost divided by uptime:
C_eff = C_nominal / U
Under a bounded uptime regime, U has a floor. Miners finance capital expenditure against a distribution with a known lower bound, and net present value can be bounded accordingly. Remove the floor — let U be set by administrative discretion with no disclosed minimum — and the denominator can approach zero without limit. C_eff becomes unbounded, and with it the variance of the entire investment model. The grid's math holds; the miner's does not. When the math holds but the incentives break, capital does not argue — it leaves.
Reframe it as an option and the asymmetry sharpens. Category Four grants the grid a perpetual, one-sided option on the miner's uptime, struck at zero cost and exercisable without notice. The miner, by accepting the classification, has written that option for free. In any honest market, the writer of a free option is compensated by the premium — the discounted tariff that conventional interruptible-load contracts pay. Here the premium is withheld. The miner sells optionality and receives nothing, which is not a power contract but an uncompensated short position in its own continuity.
I spent 2024 stress-testing Solana's validator network to see where throughput claims broke under load, and the lesson generalizes: an untested system's advertised capacity is a hypothesis, not a property. Russia's Category Four has never been stress-tested publicly. No one outside the Ministry of Energy knows how often the switch is thrown, in which regions, or for how long. An untested curtailment regime is not a known risk — it is an unknown one masquerading as a policy.
This is the same failure shape I traced in the Ronin bridge. Ronin did not fail; it was engineered to trust. The consensus mechanism was sound; the vulnerability lived in off-chain validator signature verification, outside the audited perimeter. Russian mining has not failed either; it has been engineered to be interruptible. The vulnerability is not in the grid. It is in an off-chain administrative discretion that no audit can reach and no contract can bind.
The blind spot the market is missing
The reflexive read is that Russia's legalization is bullish for domestic mining and the classification is a minor operational footnote. That read inverts the actual risk. A prohibition is priced easily: it is binary, it can be hedged, and it can be litigated. An undefined curtailment right is not binary. It is a stochastic variable with no disclosed distribution — the worst possible input for a capital-intensive, long-horizon asset. Complexity is not a shield; it is a trap, and Category Four is complexity dressed as a simple administrative tier.
The deeper maneuver is political. By classifying miners rather than banning them, the state acquires a soft decommissioning tool. It can shed marginal miners during winter peaks without paying the political cost of an outright prohibition, and without ever naming an exit. The miners leave on their own, priced out by uncertainty rather than evicted by decree.
There is a second-order blind spot worth flagging. The same winter peak that triggers curtailment is the peak that AI data centers now compete for. If power is being rationed, the load with the strongest political claim — and the loudest revenue story — is not a mining farm. The category structure quietly establishes the priority order in which the state would rather allocate scarce electrons, and mining is not near the top.
The second-order effects land on the hosting and pool layer. Miners rarely run their own sites; they rent capacity under hosting contracts that assume a supply baseline. If the baseline is administratively revocable, every hosting agreement inside Russia reprices, and pool hashrate attributed to Russian facilities inherits the same variance. That variance propagates into difficulty adjustments and, at the margin, into fee markets — not through consensus failure, but through a slower, quieter erosion of the supply-side assumption that hashrate is geographically fungible.
The rational miner response is migration to energy the grid cannot ration — curtailed hydro, associated gas, off-grid generation. That path exists, but it is capital-intensive and slow, which is precisely why the policy bites in the interim. The miners who survive the transition will be the ones who stopped treating grid power as a given and started treating it as a counterparty.
Where this is going
This is a template, not an incident. Kazakhstan and Iran have run earlier versions of the same experiment, and the pattern — legalize the activity, reserve the right to curtail the power — is portable to any jurisdiction where mining and residential demand collide. I have argued that Layer 2 is merely a delay in truth extraction; here the delay has migrated from the block to the breaker. Watch three signals over the next two quarters: the publication of actual curtailment criteria, the trajectory of Russia's share of global hashrate, and whether a peer state copies the fourth-class mechanism outright. If two of the three move, the era of cheap-power mining arbitrage is being administratively closed — and the miners who survive will be the ones who never depended on the grid's permission to keep running.