The Dutch reform does not tax crypto twice. It taxes crypto earlier.
Consider two portfolios of identical economic substance. One holds a euro-denominated equity index fund. The other holds bitcoin. From 2028, the first portfolio is taxed at the moment of sale โ realized gain, one event, one measurement, one number on a broker statement. At least until 2030, the second portfolio is taxed on unrealized gain โ an annual measurement of a position that has not been sold, may never be sold, and in the interval may have round-tripped to zero.
The differential is not a rate. Rates can be argued, capped, indexed, exempted, phased. The differential is a clock. One asset class is measured once, at exit. The other is measured every year, in perpetuity, against a cost basis that must be reconstructed from an append-only ledger that was never designed to record it.
The entire public record supporting this claim is four data points long. One unnamed industry body. One wire service. No statutory text. No rate schedule. No exemption threshold. No loss carry-forward language. No effective-date clause beyond two round numbers: 2028 and 2030.
That thinness is the first thing an auditor notices. Silence is the strongest proof of truth โ but silence is also where lobbying lives. A claim that a policy "may trigger selling" is not a finding. It is a hypothesis wearing the clothes of a finding. The work is to separate the two, because the two have very different consequences for anyone deciding where to hold an asset.
The substrate: Box 3 and why the reform exists at all
To read the Dutch proposal correctly you have to understand that it is not a crypto policy. It is a repair job on a tax regime that a court already broke.
The Netherlands does not tax personal investment income through a conventional capital gains line item. It taxes it through Box 3 โ one of three boxes in the personal income system, covering savings and investments as a single aggregate. The historical method was notional: the tax authority imputed a fictional rate of return to the aggregate portfolio, split between a deemed savings component and a deemed investment component, and levied a flat rate on that fiction. The actual composition of the portfolio mattered only at the margin, and the actual return did not matter at all.
This worked while interest rates were positive. It failed catastrophically when they were not. In December 2021 the Hoge Raad โ the Dutch Supreme Court โ held that taxing a notional return that exceeded the actual return breached Article 1 of Protocol 1 of the European Convention on Human Rights, the provision protecting the peaceful enjoyment of possessions. A tax on a gain that never existed is expropriation with extra steps.
History verifies what speculation cannot. The legislator did not appeal to first principles; it appealed to arithmetic. A transitional regime was bolted on, and the stated destination became taxation of werkelijk rendement โ actual return. That destination has been pushed back repeatedly. Any proposal that claims to tax actual returns is therefore operating inside a framework that is already several years late and politically bruised.
This matters because it tells you what the reform is optimizing for. It is not optimizing for fairness between asset classes. It is optimizing for defensibility in court. Every category that can be measured reliably lands in the realized-gain bucket, because realized gain is a fact. Every category that cannot be measured reliably stays in the mark-to-market bucket, because a valuation at a timestamp is also, technically, a fact โ a weaker one, but a fact nonetheless.
The second substrate is European. DAC8, the eighth administrative cooperation directive, brings crypto-asset service providers into the automatic exchange of information network. It applies from 1 January 2026, with the first exchanges in 2027 covering the prior year. CARF, the OECD framework it implements, is structurally the same idea exported globally. MiCA supplies the licensing layer and, crucially, a passport: a firm authorised in one member state may serve the whole union.
Read those three instruments together and the Dutch proposal stops looking arbitrary. DAC8 supplies the data. MiCA supplies the licensed intermediary that must produce it. The tax code supplies the liability. The three-year window between 2026 and the 2028 start date is not padding. It is the interval required for the reporting infrastructure to produce a first full year of usable records.
Structure outlasts sentiment. The architecture here was built in Brussels and Paris before it was ever debated in The Hague.
The loop I actually tested
I have spent enough time inside accounting-adjacent smart contracts to have a strong prior about where mark-to-market systems break. In 2020 I reviewed early cToken implementations for a small team, looking specifically at interest accrual. The finding that mattered was not a reentrancy bug. It was an overflow in the rate accumulation path that propagated across twelve lending pools, because each pool inherited the same accrual assumption and none of them had been tested against a rate regime the original authors had not imagined. The dollar figure attached to that finding was large, but the lesson was structural: the failure mode of an accounting system is never the formula. It is the assumption the formula silently embeds.
Cost basis is the same class of object. It is a state variable. It has an initial condition โ the acquisition price in euro โ and a transition function โ every subsequent event that changes the position. In a brokerage account, the custodian owns the state variable, the regulator mandates its format, and the taxpayer simply inherits it. In a self-custody crypto portfolio, the state variable is owned by nobody. It is distributed across wallets, chains, contract logs, and the taxpayer's memory.
Consider what a single position can accumulate over one tax year. A purchase on a centralised exchange. A withdrawal to self-custody. A bridge to a second chain, which mints a representation and burns the original โ an economically neutral event that generates a taxable-looking transfer in the raw logs. A deposit into a liquidity pool, which mints an LP token whose price is not quoted anywhere standard. Accrued fees inside that position, denominated in two volatile assets, auto-compounded. A governance token distributed as an airdrop with no cost basis and no liquid market. A rebasing token whose balance changes without any transfer event at all. A wrapped variant trading at a persistent discount to the underlying.
At the end of the year, the tax authority requires a euro valuation of that position and a euro-denominated cost basis to net against it. Neither number exists in the protocol. Both must be synthesised.
The 2018 refund contract work taught me the same lesson from the other direction. I spent three months on withdrawal logic in an ICO refund contract because three edge cases in the ordering of state updates could have locked out roughly fifty thousand claimants. The bug was not in the math. It was in the sequence. Tax compliance has the same shape: the liability is computed correctly and the sequence of evidence needed to support it does not exist.
And that is the point where the proposal stops being a policy question and becomes an engineering one. Complexity hides its own failures. A mark-to-market regime assumes a valuation oracle and a cost-basis ledger. Neither is specified.
Three execution problems the proposal has not answered
First: the valuation source. An annual tax on unrealized gain requires a price at a defined instant. The proposal as reported does not name the instant. Midnight UTC is not midnight CET. A December 31 close in one timezone is a December 31 open in another, and for a volatile asset the difference is not decorative โ it is the entire tax base. Nor does the proposal name the venue. A token listed on three exchanges with three order books and three last-trade prices has three defensible valuations, and the widest spread is usually where the thinnest liquidity sits. For low-cap tokens, bridged representations, and LP positions, there may be no continuous market at all. In that gap sits a manipulation surface: a thin book marked at year-end is a price anyone with inventory can move.
Second: cost basis in the self-custody tail. DAC8 gives the tax authority visibility into custodial flows. It gives nothing for wallets no service provider touches. Chain analytics can infer ownership probabilistically. Inference is not assessment. The result is a regime whose enforcement intensity depends on where an asset is held, which is a structural incentive to hold assets where enforcement is weakest. That incentive was not designed into the proposal. It falls out of it.
Third: the cash-flow mismatch. This is the part that converts a tax question into a solvency question for individual holders, and it is the part most consistently underweighted.
A non-yielding spot position โ bitcoin, unstaked ether โ generates a liability with no cash inflow. The holder can sell a fraction of the position to fund the tax. That is a real cost, but it is survivable.
A staking position is worse. Rewards arrive as tokens, frequently inflationary, frequently illiquid at the margin. The liability is denominated in euro and computed on both the rewards and the appreciation of principal. The taxpayer is being taxed on units of a volatile asset whose euro value at the payment date is not the euro value at the tax date. Between those two dates the liability is fixed and the funding source is not.
A DeFi liquidity position is worse again. Fees accrue inside a contract, often auto-compounded, often in two assets, and the position's impermanent loss is a real economic loss that the mark-to-market framework does not obviously recognise as a deductible event because no disposal occurred.
An airdrop is the pathological case. Receipt is a taxable event at market value in most mark-to-market designs. There is no cash. There is frequently no market. The recipient may be taxed on an asset they cannot sell to pay the tax on holding it. And a vesting or locked allocation is worse than the airdrop: it is a position with a paper gain, no transferability, and an annual assessment attached. The only escape routes are OTC discount sales or default.
Now run the volatility arithmetic, because volatility is where the asymmetry compounds.
A position of โฌ100,000 gains 60% in year one. Mark-to-market produces a โฌ60,000 taxable gain. At a Box 3 rate in the region of 36%, the liability is โฌ21,600 โ payable in cash, in the year the gain exists only on paper. In year two the position falls 50%. It is now worth โฌ80,000. Cumulative realized profit: zero. Cumulative tax paid: โฌ21,600. That is roughly 27% of the original principal, extracted from an investment that never produced a euro of realized return.
The holder funded that liability by selling into the position that subsequently halved. The tax was computed on a peak that did not persist, and no loss carry-forward provision has been disclosed that would return it. This is the mechanical reason mark-to-market regimes are rare for retail investors and standard for derivatives desks: a derivatives desk can hedge the mark. A retail holder cannot hedge an annual valuation of an unhedgeable position.
Evidence does not negotiate with any of this. The arithmetic holds regardless of who is lobbying.
The lobbying structure, read cold
The reported claim is that excluding crypto from the new realized-gain regime means crypto continues to bear the heavier, earlier annual charge, which makes the after-tax cash flow from holding crypto worse than from holding equities, which may cause holders to sell or relocate.
That chain is directionally correct and it is also not the ask. The industry position is not the abolition of crypto taxation. It is parity: fold crypto into the realized-gain bucket alongside equities. The "sell-off" framing is a pressure device aimed at a legislative window, and it is aimed there because a tax authority's priority is protecting the base, not stabilising prices.
The most informative detail in the whole disclosure is the company crypto keeps. It is grouped with gold and foreign exchange, not with speculative ventures or gambling products. That grouping is a tell about the legislator's mental model. Gold and FX share one property with crypto and one property with each other: they are highly liquid, trivially movable across borders, and resistant to cost-basis tracking. The classification is not moral. It is an assessment of traceability.
Which means the debate about whether this is "anti-crypto" is the wrong debate. The productive question is whether the traceability assumption survives contact with self-custody. It probably does not, and the legislator probably knows it, which is why the timeline is anchored to the DAC8 reporting cycle rather than to an independent policy judgment.
There is a second-order effect worth naming. A regime that taxes positions rather than trades makes the custodial intermediary structurally more valuable, because the intermediary can pre-compute basis, withhold, and report. That is a subsidy to centralised custody dressed as a tax code. It runs directly against the self-custody thesis, and nobody involved in the public argument has said so.
The contrarian read: the damage is not where the headline says
The consensus reading of this story is that the Netherlands is tightening, crypto holders will sell, and the selling is the story. I think that reading is wrong in three specific ways, and each way is measurable.
First, the sell-off prediction is unfalsifiable as stated. The reported claim is conditional: if crypto is excluded, then holders may sell. Two conditions, neither quantified, delivered by a party with a clear interest in the answer. There is no volume data, no cohort data, no estimate of affected holdings. A prediction with no quantity attached cannot be wrong, which means it cannot be evidence. Treating it as a market fact is a category error.
The actual harm shows up elsewhere: in compliance cost. Reconstructing a multi-chain, multi-wallet, DeFi-active cost basis to a standard that survives audit is expensive. That expense is a hidden tax rate that nobody quotes in the political argument. For an active DeFi user, the accountant's invoice and the risk of a misreporting penalty may exceed the nominal liability. That is the real regressive edge of this design, and it falls hardest on the mid-sized holder โ too large to ignore the rules, too small to hire a desk.
Second, the reform may be net favourable and the coverage is not written that way. The current arrangement taxes a notional return. The proposal, on its face, moves toward taxing actual return. If crypto eventually lands in the realized-gain bucket, the outcome is strictly better for holders than the status quo: tax deferred until disposal, losses recognised at disposal, no annual cash drain. The public argument has inverted this. A proposal whose worst case is "you keep the bad regime" and whose best case is "you get the good one" has been reported as pure downside. That is a framing failure, and framing failures are where the mispricing lives.
Third, and least discussed: the reform is self-eroding. A tax on unrealized gain in a jurisdiction where the holder can move the asset โ or move themselves โ does not raise the projected revenue. It raises the projected migration. MiCA's passport makes the platform side frictionless. Self-custody makes the asset side frictionless. The only thing left anchoring the base is residence, and residence is a filing, not a fact. So the policy taxes the least mobile participants hardest and calls it a base.
There is a legal weak point too, and it is not decorative. The Box 3 litigation succeeded precisely because a fictional return exceeded real return. A regime that taxes unrealized appreciation on one asset class while taxing the same economic substance on another only at disposal invites the same argument in a new form: unequal treatment of comparable positions, challenged under the same property-rights provision that already produced one taxpayer-favourable judgment. If that challenge succeeds, the correction is retroactive refunds โ which creates a fiscal incentive for the state to legislate first and settle later.
Pressure reveals the cracks in logic. The crack here is that the proposal's own stated principle โ tax actual return โ is the argument against taxing unrealized return on volatile assets.
The last gap is the one that should restrain everyone. There is no named industry body. A lobby representing five exchanges and a lobby representing two hundred thousand retail holders produce the same press release and mean entirely different things. The disclosure does not let you tell which one you are reading. That is the single most important omission in the entire story, and no amount of downstream analysis closes it.
What to watch, and what would change the answer
The article count on this story will keep rising. The information content will not, until a statute appears. What matters is narrow and specific.
Watch the Tweede Kamer record for the actual bill text, and specifically for four clauses: the rate, the annual exemption threshold, loss carry-forward, and the classification of crypto. Those four numbers and one category decide whether the impact is a rounding error or a genuine relocation event. The legislative track matters more than the wire copy, in both chambers and in the Council of State's advice on the draft.
Watch whether crypto moves from the third bucket into the second. If it does, the current anxiety is obsolete and the correct posture is relief. If it does not, the binding constraint is not the tax โ it is the cost basis.
Watch DAC8 implementation through 2026 and the first exchanges in 2027. Enforcement capacity is a function of that pipeline, not of the statute's ambition. Then watch the local order books, which will tell you whether "may sell" was ever anything other than a sentence.
Patience is a technical requirement. The schedule in question runs to 2030. Anyone repricing a portfolio today on the basis of a four-point wire story is not trading a policy โ they are trading a headline about a proposal about a proposal.
One question remains open, and it is the one that decides the whole file. If the legislator's objection to crypto is that its cost basis cannot be traced, and if that objection is soon to be answered by the same reporting infrastructure the legislator is building, then what exactly is the reason for keeping crypto in the harsher bucket after the traceability problem is solved?
