On October 6, the IRS published Revenue Procedure 2026-20. Fourteen safe harbor conditions for qualified trusts that stake Proof-of-Stake assets. Five have surfaced in public summaries. Nine remain buried behind tax databases and law firm memos.
The number that matters is not fourteen. It is two. The IRS split staking into two taxonomies โ passive property preservation and active business conduct. Every other requirement flows from that binary.
For a grantor trust, the difference is existential. Under IRC sections 671 through 677, a trust passes income through to beneficiaries and avoids entity-level taxation. Stake the wrong way, and the trust is re-characterized as a business entity. Pass-through status dies. Corporate treatment begins. The IRS did not cut the tax rate on staking rewards. It re-characterized the activity itself. That single phrase โ "property preservation activity" โ is the entire game.
Let me show you why the market is misreading it.
Context: How a Tax Procedure Repriced a Product Class
I spent two weeks in 2020 reverse-engineering the automated market maker mechanics of Uniswap V2 and Curve. The lesson stuck: the number that moves institutional capital is rarely the headline number. It is the structural constraint underneath it.
Digital asset trusts have carried a quiet structural constraint since PoS chains scaled. A trust holds assets for beneficiaries. If it stakes, validators run, rewards accrue, and the trust starts to look operational. The IRS never formally declared staking a trade or business. But the ambiguity was enough. Conservative issuers stayed out. Auditors flagged the exposure. Product lawyers could not underwrite the risk.
Revenue Procedure 2025-31, published in November 2025, offered principles-based guidance. Translation: nobody knew the operational rules. It told issuers that staking might be permissible without telling them how to structure it. That is not a green light. It is a yellow one with no timeline.
Revenue Procedure 2026-20 replaces it with a checklist. Fourteen items. Binary compliance. Satisfy them all and you sit inside the safe harbor. Miss one and you fall back into case-by-case judgment โ which is precisely the uncertainty you were trying to escape. A safe harbor is not a permission slip. It is an off-ramp from ambiguity, and off-ramps have lane markings.
The procedure applies to tax years ending on or after November 10, 2025. Read that date twice. The applicability window opens before the publication window. Either the IRS intended retroactive relief, or a drafting inconsistency slipped through. I raised this with two tax counsel contacts. Both gave the same instruction: verify against the official text before you file. Retroactive applicability is not a footnote. It is a liability.
Now the core.
Core: The Fourteen-Point Architecture
The mechanism is activity characterization, not rate relief. Staking rewards remain taxable income. Nothing in 2026-20 touches Revenue Ruling 2023-14, which governs the timing of reward recognition. The procedure answers exactly one question: does staking destroy the trust's passive identity? Read it any other way and you will draw the wrong conclusions.
The disclosed requirements reveal the intended beneficiary with unusual precision.
Shares must be listed on a national securities exchange. This is not a general trust provision. This is a spot ETF provision. Non-listed digital asset funds fall outside the safe harbor entirely. I have audited enough filing structures to recognize a product-specific carve-out. This one is drawn around listed wrappers.
The trust may hold only a single asset class. Single-asset trusts qualify. Multi-asset baskets do not. An index-style crypto trust cannot use this safe harbor even if every other condition is satisfied. That exclusion is deliberate. Multi-asset staking introduces allocation and rebalancing decisions. Rebalancing looks like management. Management looks like a business.
Staking rewards must be distributed, not hoarded. The trust cannot retain and re-stake. Distributions flow to beneficiaries. This is an anti-active-management design. Retained earnings signal a going concern. Distributed income signals a passive rent collector. The IRS is choosing which one it wants the trust to resemble.
A qualified custodian must control the staked assets. Not self-custody. Not a DeFi vault. A licensed custodian with a charter and an examination record. This requirement alone eliminates most crypto-native staking arrangements.
The liquidity policy must be approved by the SEC. This is the detail most analysts skipped. The IRS is not acting alone. Tax characterization sits with the IRS. Product structure and redemption mechanics sit with the SEC. Entry into this safe harbor requires clearing both gates. Two regulators, one door.
Nine requirements remain undisclosed in the summaries I reviewed. Based on the pattern, I expect they cover staking concentration limits, validator selection constraints, conflict-of-interest disclosure, and delegation boundaries. My confidence on that inference is moderate. Do not build a compliance model on it. Pull the official text.
There is a second-order observation worth stating. Network congestion does not change tax characterization. The protocol's congestion affects reward timing and fee economics, but the IRS is indifferent to your validator's performance. What it cares about is control. Who holds the keys, who selects the validators, who decides to stake or not. The fourteen requirements are a control audit wearing tax language.
Who Actually Wins
The beneficiary map is narrow, and it is not the coin.
Listed single-asset trusts win. That means the largest spot ETH products and any spot staking vehicle already trading on a national exchange. They can now legally fold staking yield into the wrapper. That is a real product upgrade.
Qualified custodians win harder. The custodian requirement converts custody from a service into a gate. Value concentrates at the chokepoint. I have watched this pattern before. In 2021, I traced the metadata pinning infrastructure of three leading NFT marketplaces and found that 40% of "permanent" tokens depended on centralized servers vulnerable to takedown. The lesson was identical: control the dependency, capture the margin. The same logic applies here. Custody is the dependency. Custody captures the margin.
Staking-as-a-service providers win indirectly. Institutional staking demand migrates toward licensed counterparties with audited key management. Unlicensed operators lose share.
Non-listed funds lose. Multi-asset vehicles lose. Offshore structures lose. They are not banned. They are simply excluded from the off-ramp, left in the fog of case-by-case judgment.
And the individual staker? Unaffected. This procedure governs trusts. It does not touch your personal validator, your liquid staking position, or your DeFi allocation. Anyone telling you otherwise is selling something.
Contrarian: This Is a Product Rule, Not a Price Rule
Here is where the narrative breaks.
Social feeds will frame 2026-20 as bullish for ETH. That framing is wrong on the timeline. The transmission path is slow and multi-hop: product attractiveness rises, institutional allocation follows, demand responds. Each hop takes quarters, not hours. The immediate price impact is close to zero. This is not a catalyst. It is a constraint removal, and constraint removals do not print candles.
Worse, the no-hoarding requirement is a yield ceiling. A trust that must distribute staking rewards cannot compound them internally. A personal staker can re-stake continuously. Over a multi-year horizon, that compounding gap is material. The trust wrapper trades tax certainty for yield drag. I have not seen a single analyst price this. The headline says "staking allowed." The fine print says "staking allowed, then distributed, no compounding."
Then there is the iteration risk. Revenue Procedure 2025-31 lasted roughly one year before replacement. Rules in this area are not stable. Today's compliance is not tomorrow's compliance. Any issuer building a permanent product strategy on 2026-20 should assume a 2027 revision and budget for it.
The information asymmetry cuts against the reader too. Five of fourteen requirements are public. Nine are not. If you are modeling trust economics from a summary, you are modeling from a fragment. The network's congestion, the validator queue, the reward curve โ none of that matters if you cannot confirm the nine hidden conditions.
One more contrarian read. The "property preservation" label may leak beyond tax law. If staking is passive for tax purposes, defense counsel in a securities matter will cite it. I am not claiming it controls a Howey analysis. I am saying it becomes a data point that regulators and litigants reach for. Characterizations travel.
Takeaway: Watch the Filings, Not the Charts
The signal to track is not the ETH price. It is the amendment stream.
If spot ETH trusts begin filing staking addenda, the safe harbor is being used. If custodians announce institutional staking mandates tied to listed products, value is migrating to the chokepoint. If a multi-asset trust petitions the IRS for inclusion, the next revision is already forming. If the nine undisclosed requirements include a staking concentration cap, expect product redesigns before year-end.
The IRS chose to channel compliant staking rather than suppress it. That is constructive. It is also conditional. Fourteen conditions, two regulators, one unresolved date inconsistency, and nine unknown rules.
The question for the next quarter is not whether staking is legal inside a trust. It is whether the trust wrapper can compete with the staker who never had to ask permission. In a bear market, survival is a function of structure, not sentiment. Structure just changed. Sentiment has not caught up.