Grayscale’s TAO ETF Can’t Stake — Its Own Filing Says So

Article Brief

Key Takeaways

5 Points30s Read

1. 01No staking–Grayscale’s April 2, 2026 amended registration statement bars the Bittensor trust from putting any of its TAO into a staking protocol, with no assurance that ever changes.
2. 02Emissions follow stake–Under dTAO, each subnet has an alpha token and a liquidity pool, and a subnet’s share of new TAO is set by how much TAO is staked into its pool.
3. 03Merit is not the input–Grayscale’s own risk language notes emissions are allocated by relative staked TAO rather than by the quality of a subnet’s AI outputs.
4. 04The blocker is tax–Staking is gated on the trust keeping grantor-trust treatment, which needs a tax opinion, ruling, or guidance — not a protocol change.
5. 05Cost compounds–The 2.50% sponsor fee is paid in TAO, so units backing each share decline while unstaked holdings miss the emission flow.

Grayscale wants to turn its Bittensor product into a listed fund, and the paperwork says something the ticker never will. If the conversion works, the vehicle will hold TAO and do absolutely nothing with it.

That is not a swipe at Grayscale’s execution. It is a constraint written into the amended registration statement the sponsor filed on April 2, 2026, for the Grayscale Bittensor Trust — a Delaware statutory trust whose shares are quoted on OTCQX under GTAO and which intends to list on NYSE Arca under the same symbol once the SEC declares the filing effective, at which point the sponsor plans to rename it Grayscale Bittensor Trust ETF. The prospectus is unambiguous: no portion of the trust’s TAO will be put into any staking protocol, and there is no assurance that the trust, the sponsor, the custodian, or anyone associated with it will ever be permitted to do so.

For a bitcoin wrapper that would be a shrug. Bitcoin does not pay its holders either, and nobody expects an ETF to mine. Bittensor is built differently, and the difference is the entire story.

What the trust is actually a claim on

Strip away the branding and the vehicle is simple. It holds TAO. Its stated objective is for share value to track the value of that TAO by reference to an index price, less expenses, and it will not use leverage, derivatives, or anything similar to get there. Authorized participants create and redeem in baskets, in kind or through cash orders handled by liquidity providers.

The one recurring cost is the sponsor’s fee, which accrues daily in dollars and is then converted and paid in TAO. That detail matters more than it looks. Because the fee is settled in the asset itself, the quantity of TAO backing each share declines over time by construction — the prospectus says as much when it notes the TAO required to create or redeem a basket will gradually decrease. A holder is paying rent in the thing they are trying to accumulate.

Bittensor pays for AI work, and dTAO decides whose

Here is where a TAO wrapper stops resembling a commodity trust.

Bittensor was introduced in a whitepaper by the pseudonymous Yuma Rao, and its Yuma Consensus exists to grade user-submitted AI outputs. Participants called miners host AI models and offer them to the network. Those miners cluster into subnets, which the filing’s own glossary defines as self-contained incentive frameworks for miners to perform a specific AI task — text generation, image work, and so on. The chain underneath, Subtensor, runs on proof-of-authority.

In February 2025 the network shipped Dynamic TAO, or dTAO, and rewired the economics. Every subnet now has its own token, generically an alpha token, plus a liquidity pool. Backing a subnet means staking TAO into that subnet’s pool and receiving alpha in return. The share of newly minted TAO a subnet receives is governed by how much TAO sits staked in its pool relative to every other pool.

Read that mechanism twice, because Grayscale’s risk disclosure states the consequence more bluntly than most crypto commentary does. The filing warns that dTAO allocates emissions based on relative staked TAO “rather than on the quality of the Subnet’s underlying AI outputs,” and flags that incentives can concentrate in speculative or underperforming subnets and that influence can pool among a small number of large holders.

So the network’s AI subsidy is settled by a capital-weighted auction. Whoever stakes the most into a pool steers the most emission toward that subnet’s miners, whether or not those miners are producing the best work. That is the AI mechanism at the center of this asset, and it is the layer a spot wrapper cannot touch.

The dilution nobody puts on the fact sheet

TAO issuance is hard-capped at 21 million, a limit currently enforced in the network code. One TAO was minted per block at roughly twelve-second intervals, halving on a four-year cadence — the first halving was expected in October 2025, taking issuance to half a TAO per block, with another cut expected in 2029 and a final event projected for September 2069.

New units still enter circulation every twelve seconds, and they flow to the people running and backing subnets. Someone holding TAO without staking it does not share in that flow; their slice of the network quietly thins while the emission schedule runs. That is ordinary staking dilution, and it is precisely the argument that made staking such a live issue for other crypto funds — a fight TECHi covered when the SEC weighed staking inside Ethereum ETFs.

The asymmetry is sharper on Bittensor than on a generic proof-of-stake chain, because staking here is not a passive yield toggle. It is the act that routes subsidy to particular AI workloads. A holder who cannot stake is not merely forgoing a coupon. They are absent from the governance of where the network’s compute budget goes.

The blocker is a tax question, not a crypto one

The prospectus does not say staking is impossible forever. It ties any staking to what it defines as the staking condition, and that definition is thoroughly unglamorous: staking must not jeopardize the trust’s treatment as a grantor trust for US federal income tax purposes, and the trust needs either a written opinion from a tax advisor, a tax ruling, or tax guidance confirming that a grantor trust may do this.

Nothing about that turns on Bittensor’s engineering. It turns on how the IRS and Treasury characterize staking rewards received by a passive trust — the same knot that has shadowed every staking-enabled US crypto product. Until it is untied, the vehicle’s objective can only contemplate staking consideration in the conditional, which is exactly how the filing phrases it.

Anyone modeling an approval date should therefore separate two events that tend to get blurred together: a registration statement going effective and shares listing, versus the trust being permitted to stake. The first can happen without the second.

Someone still holds the switch

Bittensor’s rules are software, and the filing is candid about who can change them. The Opentensor Foundation and core developers can access and alter the network’s source code and are responsible for releases. Subtensor’s proof-of-authority design means nodes are admitted by the network administrator, and the foundation owns or controls a majority of them.

That concentration cuts in an awkward direction for a fund whose entire premise is passivity. The trust is not actively managed and has no strategy for the network’s development. Emission mechanics have already been reworked once in dTAO; if they are reworked again, a wrapper holding unstaked TAO absorbs the outcome without a vote or a hedge.

What the audited numbers already showed

The trust’s own financial statements give a sense of what holding this exposure has felt like. For the year ended December 31, 2025, principal market NAV per share fell from $8.76 to $4.24, a total return of negative 51.60%. Expenses ran at 2.50% of average net assets.

The token has not recovered since. TAO changed hands near $193 late on July 23, 2026, quoted at $193.34 by Financial Modeling Prep and $193.16 by Alpha Vantage as of 6:36 p.m. ET, a spread of under a tenth of a percent between the two feeds. That leaves it roughly 64% below its highest print of the trailing year.

That fee sits far above what large listed crypto funds now charge, a spread TECHi walked through when comparing Bitcoin ETF fee structures. Paying 2.50% in-kind, on an asset whose protocol is simultaneously issuing new units to stakers you are not among, is a compounding drag from two directions at once.

None of that forecasts the next twelve months. It does establish that the wrapper has already been tested through a severe drawdown, and that its cost structure was not built for a low-fee listed market.

Distribution is arriving ahead of the answer

The gap gets stranger when you look at what non-ETF channels shipped this week. On July 21, 2026, MEXC announced TAO staking built on validator infrastructure from Yuma, a Bittensor-focused firm and a Digital Currency Group subsidiary, opening plain TAO staking to a user base the exchange puts above 40 million across more than 170 countries.

Plain TAO staking, not subnet alpha — so even that route stops short of the pool-level exposure where dTAO actually allocates emissions. Still, the ordering is worth sitting with. A retail user on an offshore exchange can stake today. A US investor who waits for the regulated wrapper gets custody, an audited NAV, and a listing venue, while giving up participation in the mechanism that makes the asset productive.

That trade-off is not unique to Bittensor. It is the shape of most AI-crypto infrastructure right now, where the rails ship faster than the compliant access to them — a pattern visible in how agent payment rails reached merchants before the accounting caught up, and in how the ASI token’s rebound outran its actual agent-payment volume. Allocators weighing token exposure against equity exposure to the same theme have been running into the identical problem all year, which TECHi mapped in its comparison of crypto versus AI stocks.

Where this reading could break

Several things would weaken the argument above, and they deserve naming.

If Treasury or the IRS issues guidance letting grantor trusts stake, the staking condition is satisfied and the structural gap narrows quickly — the trust’s objective already contemplates TAO earned as staking consideration, so the plumbing is pre-drafted. Timing is unknowable, and the filing carefully promises nothing.

Subnet alpha exposure may also be something a conservative allocator should want to avoid. Alpha tokens float against TAO, depend on individual subnets continuing to operate, and carry liquidity risk that Grayscale spells out. An investor who wants the AI-network beta without the pool-level volatility might reasonably regard the wrapper’s limitation as a feature.

And if TAO’s price path dominates everything else, dilution at half a TAO per block becomes a rounding error against the move. Structural analysis of this kind matters most in flat and falling markets, which is where the asset spent 2025.

One caveat on scope: reporting elsewhere has referenced competing spot TAO filings from other issuers, but a search of EDGAR company records turned up only the Grayscale trust and an unrelated private fund. This piece therefore rests on the Grayscale documents, which are the ones on the public record.

What that record shows is narrow and specific. The most regulated way to own Bittensor is, by the sponsor’s own disclosure, the version that sits out the network’s core economic loop — and the fix is sitting in a tax opinion, not in a block explorer.