Bittensor’s Subnet Tokens Are Leaving Home. What You Own Now

The year Bittensor’s subnet tokens left home

For most of its life, Bittensor was a one-token story. You bought TAO, you staked it, perhaps you skimmed a subnet whitepaper, and that was the extent of what the market could price. In 2026 that changed. The network’s dynamic TAO upgrade turned every subnet into its own tradable token, and over the past year those tokens have started leaving the house they were born in. They now trade on Base through a Chainlink-powered bridge, they sit on Kraken’s order books, and they back a growing shelf of regulated products in Europe. The question for anyone holding one is no longer “what is Bittensor.” It is “what exactly do I own when a subnet token turns up somewhere other than Bittensor.”

The timing is not an accident. A rotation into artificial-intelligence tokens through late September and early October lifted the whole sector, and TAO rode it. The token changed hands near $285 in early October, with a market capitalization around $3.2 billion and a fully diluted value close to $6 billion, still roughly 62 percent below its March 2024 record of $757.60. Rallies pull eyes toward where tokens actually trade, and in 2026 that is a far longer list than it used to be. This piece maps the export rails one by one, then asks the harder question underneath them: whether a wrapped, bridged, exchange-listed subnet token is a claim on a real business or a prettier package around a subsidy.

What a subnet alpha token actually is

Bittensor is a network of specialized markets called subnets. Each subnet runs a competition: miners submit work, such as model inference, training contributions, data or predictions, validators score that work, and the protocol pays emissions to the participants who score well. Until early 2025 those payouts were denominated only in TAO, and TAO was the single asset the outside world could buy. Then came dynamic TAO, usually shortened to dTAO, which the project rolled out in February 2025 and which CoinGecko describes as giving every subnet its own token and its own market.

Here is the mechanism in plain terms. Every subnet now has an alpha token and an on-chain liquidity pool that pairs that alpha against TAO. When you want exposure to a subnet, you stake TAO into its pool and receive alpha in return, much like depositing into an automated market maker. The pool sets a price for the subnet’s alpha in TAO terms, and a smoothed version of that price, an exponential moving average, decides how large a share of network emissions the subnet earns. A subnet whose token the market bids up attracts more emissions; a subnet the market ignores earns less, and recent rule changes add emission-gate throttles that cut rewards to markets with thin demand.

The practical result is that Bittensor stopped being one asset and became a basket of them. The network supports 128 subnets today, with a cap set to expand toward 256 later in the cycle. Each of those subnets has a token with a live price, a liquidity pool, and a set of holders who are, in effect, betting on that subnet’s slice of the emission pie. Alpha is not a governance token in the Ethereum sense and it is not equity. It is a claim on a subnet’s future emissions and on whatever demand its token can sustain in the pool. That distinction matters once the token leaves Bittensor, because the thing you are wrapping or listing is itself a derivative of an incentive schedule.

One network, a hundred little markets

Put a hundred-plus of these tokens together and you get an ecosystem with its own market capitalization, separate from TAO. That number has swung hard with sentiment. By late March 2026 the combined value of subnet alpha tokens reached roughly $1.12 billion, about 27 percent of TAO’s own market cap, per CoinGecko data, and it briefly pushed toward $1.47 billion at the top of the March rally. By late August the tracked total had cooled to around $965 million, according to tao.media’s August snapshot. These are not blue chips. They are small, volatile markets that can double or halve on a single endorsement.

One such endorsement set the tone for the year. Nvidia chief executive Jensen Huang and investor Chamath Palihapitiya praised Bittensor’s distributed-training approach on the All-In Podcast in March, framing decentralized AI training as complementary to proprietary models, as CoinDesk reported. TAO rallied about 90 percent that month, climbing from roughly $180 to above $332, and the subnet tokens rallied harder. When the executive who runs the company selling the picks and shovels of the AI boom name-checks your network, the long tail of tokens attached to it tends to move first and ask questions later.

The leaders of the pack are the subnets that actually ship product. The table below shows a handful of them with approximate alpha-token market caps from tao.media’s August read. Note how far several had fallen from their March highs, a reminder that these are emission-linked markets, not revenue multiples.

Rail one: Project Rubicon bridges alpha to Base

The first serious attempt to give these tokens a life outside Bittensor is Project Rubicon, launched by a firm called General TAO Ventures and announced in November 2025. Rubicon takes a subnet’s alpha token and wraps it into an ERC-20 token called xAlpha that can live on Base, Coinbase’s Ethereum layer-2. The bridge is built on Chainlink’s Cross-Chain Interoperability Protocol, or CCIP, which Rubicon uses as the canonical path between Bittensor’s own EVM layer and Base, with a one-to-one mint-and-burn design and tooling audited by the security firm Hashlock. Once on Base, xAlpha tokens trade on Aerodrome, the chain’s largest decentralized exchange.

The launch cohort was 17 subnets, including Chutes (SN64), Targon (SN4), Lium (SN51), Macrocosmos’ IOTA (SN9), Ridges (SN62), Score (SN44) and Affine (SN120), each seeded with roughly $300,000 of initial liquidity split across its alpha token and a USDC pair. That is a modest float, and it matters later when we get to slippage and manipulation risk. But the intent is clear: make subnet tokens composable with the rest of DeFi so they can be traded, used as collateral, and paired into yield strategies without a holder ever touching Bittensor’s native staking flow.

The people behind it frame Rubicon as a coming-of-age moment. “Project Rubicon represents the crossover point where Bittensor can capitalize on its unparalleled intelligence network,” said Victor Teixeira, chief investment officer of General TAO Ventures, in the launch release. Johann Eid, chief business officer at Chainlink Labs, said the team was “excited to see Chainlink CCIP serve as the canonical cross-chain infrastructure powering Project Rubicon’s connection between Bittensor’s subnet alpha tokens and Base.” Even Jacob Steeves, Bittensor’s founder, offered a line of support: “Really glad General TAO Ventures is doing this, it’s a big deal for subnet tokens to reach other markets.”

Strip away the press-release gloss and Rubicon is doing something specific and new. It is turning a subnet’s internal incentive token into a standard DeFi asset on a chain where billions of dollars already change hands daily. If you believe Bittensor’s subnets will produce durable AI businesses, that is a bridge to a much deeper pool of capital. If you do not, it is a very efficient way to export volatility.

xAlpha, liquid staking, and what you actually hold

Rubicon’s own description of xAlpha is worth reading closely: it performs non-custodial liquid staking of a subnet’s alpha token. That phrase will be familiar to anyone who has held a liquid-staking token on Ethereum, where you deposit ETH, keep a tradable receipt, and still earn the underlying staking yield. The analogy is useful but incomplete. In Ethereum liquid staking, the yield comes from validating a base layer with tens of billions of dollars of economic security behind it; the mechanics and trade-offs of that model are something we have walked through in detail before. In Bittensor, the yield comes from subnet emissions, which are themselves set by the protocol and can be changed by it.

So what do you actually hold when you hold xAlpha? You hold a wrapped claim on a subnet’s alpha token, which is a claim on that subnet’s share of emissions, which is a function of the subnet’s smoothed token price and the protocol’s current rules. There are three layers of abstraction between an xAlpha balance on Base and anything a customer actually pays for. That is not a reason to avoid it, but it is a reason to understand it. A price can rise because the subnet is winning real customers, or because emissions are flowing and holders are compounding, or simply because the token is thin and someone is bidding. From the outside, those three look identical on a chart.

The liquid-staking comparison also raises the question of peg. A liquid-staking token is only as good as the market’s confidence that it can be redeemed for the underlying. xAlpha depends on the integrity of the CCIP mint-and-burn path and on enough liquidity in the Aerodrome pool to let holders exit near fair value. When both hold, xAlpha tracks alpha. When either wobbles, the wrapper can trade at a discount that has nothing to do with the subnet’s performance. This is the standard risk of any bridged, wrapped asset, and it does not disappear because the underlying is fashionable.

Rail two: Kraken opens the centralized door

The second rail is older-fashioned and, for most investors, more consequential: a major centralized exchange simply listing the tokens. On June 29, 2026, Kraken did exactly that, opening trading in a set of subnet alpha tokens that included Chutes (SN64), Targon (SN4), Lium (SN51), Ridges (SN62) and several others. Cryptobriefing called it the first time a major centralized exchange had opened the door to these specialized, AI-focused assets.

Why does a centralized listing matter more than a DeFi bridge for the median holder? Because it removes the single biggest barrier to owning these tokens, which was comfort with Bittensor’s native staking and wallet tooling. Before Kraken, buying a subnet token meant acquiring TAO, moving it into the right wallet, finding the subnet’s pool, and staking into it to receive alpha. After Kraken, it means clicking buy on an exchange that millions of people already use. The listing brought centralized order-book liquidity and cleaner price discovery to assets that previously traded only through on-chain AMMs.

The trade-off is the one that always comes with a custodial venue. When you buy a subnet token on Kraken, you typically hold an exchange balance rather than the on-chain alpha itself, which means you are trusting the exchange’s custody and its listing decisions. A token an exchange can list is a token it can delist. And a plain exchange balance generally does not stake into the subnet pool on your behalf, so unlike native alpha or Rubicon’s xAlpha, the simple exchange route can leave the emission yield on the table. Convenience and composability pull in different directions, and the two rails sit at opposite ends of that rope.

Is there anything under the token? Revenue, emissions, and the subsidy

All of this plumbing assumes the tokens are worth plumbing. So it is worth being blunt about where subnet value comes from today. The honest answer is that most of it still comes from emissions, not customers, though the gap is starting to close in a few places.

The most careful public attempt to measure real money is the SubConnect Bittensor Revenue Index, summarized by Cryptobriefing in September. It identified 24 subnets generating roughly $28 million to $35 million in annualized external revenue, with compute and infrastructure subnets accounting for around four-fifths of that. The top earners were Lium at an estimated $8 million to $10 million, Targon at $5.5 million to $6 million, and Chutes at $4 million to $5 million, with named enterprise customers reported to include PwC France, Dropbox and an unnamed NYSE-listed real-estate trust. Fourteen of the 24 were using some of that revenue to buy back their own alpha tokens, which is the closest thing the ecosystem has to a cash-flow-to-token link. We looked at whether that kind of buyback loop actually supports a token price in a separate study of perp-DEX buyback flywheels, and the short version is that it only works when the revenue is both real and large relative to emissions.

That is the catch. Set $28 million to $35 million of external revenue against emissions that run on the order of $300 million a year at current prices, and the ratio is sobering: real customer revenue is somewhere around 9 to 12 percent of what the protocol pays out. Independent trackers are harsher still on individual subnets. The flagship inference subnet, Chutes, is run by Rayon Labs, a team we profiled in September; its externally verifiable revenue through OpenRouter has been measured in the low single-digit millions even as the team has claimed figures approaching $10 million, and analysts have pegged its effective subsidy at anywhere from 20 to 40 times customer revenue. The point is not that Bittensor has no business. It is that the business is young, concentrated in a few compute subnets, and dwarfed by the emissions that currently set prices.

It helps to know how the subsidy is built. Bittensor caps TAO supply at 21 million, the same ceiling as Bitcoin, and emits new TAO on a Bitcoin-style schedule that halves at supply milestones rather than on a fixed calendar. The first halving arrived in December 2025 when issuance crossed 10.5 million TAO, cutting the daily reward in half to roughly 3,600 TAO; the next halving waits at 15.75 million. Within each subnet, those emissions are split by rule, with subnet owners, miners, and the validators and their delegators each taking fixed shares of 18, 41 and 41 percent, and the rewards are paid in alpha rather than TAO. Every xAlpha on Base and every subnet balance on Kraken ultimately traces back to this schedule. When you model one of these tokens, you are modeling an emission curve first and a business second.

The export map: where each token trades, and what you give up

Pull the rails together and a holder in 2026 has five broad ways to get subnet exposure, each with a different answer to two questions: do you earn the emission yield, and whom do you have to trust. The table lays them side by side.

The pattern is that convenience and yield rarely come together. The native route and Rubicon keep you close to emissions but demand on-chain competence or trust in a bridge. The exchange route is the easiest to use and the most likely to drop the yield. And the regulated wrappers, which we come to below, trade the yield for a brokerage account and a prospectus. No single row is strictly best; the right one depends on whether you are optimizing for ease, for yield, or for the ability to hold the thing inside a regulated account.

Bridge risk, thin liquidity, and the cost of leaving

Every export rail adds a layer of risk that did not exist while the token stayed home. Rubicon’s choice to build on Chainlink CCIP is a deliberate attempt to minimize the worst of it. Cross-chain bridges have been the single most lucrative target in crypto history, and the reason is almost always the same: the bridge holds or mints value based on a message from another chain, and if an attacker can forge that message, they can print the asset. A canonical mint-and-burn design with an established interoperability protocol and a published audit, in this case by Hashlock, is a stronger starting point than the custom, lightly reviewed bridges that produced the big exploits. The deeper trade-offs of the cryptography that secures custody and cross-chain messaging are worth understanding on their own terms, and we unpacked them with the team at Trail of Bits earlier this year. None of that makes a bridge risk-free; it makes the risk legible.

The subtler problem is liquidity. Roughly $300,000 of seed liquidity per subnet is thin enough that a modest trade moves the price, and a thin pool is an invitation to slippage on the way in and the way out. It also makes the wrapped token easier to push around, which matters because a subnet’s smoothed price feeds back into its emission share on Bittensor. A manipulated xAlpha pool on Base is not just a trading nuisance; in principle it is a lever on the native incentive system, although the moving-average smoothing and the separation between the Base pool and the native pool blunt that link. The honest summary is that these are young, shallow markets wearing grown-up infrastructure, and position sizing should reflect the pool depth, not the branding.

Then there is the wrapper-peg issue from earlier, now with teeth. If confidence in the bridge drops, or if a subnet’s native alpha sells off faster than arbitrageurs can rebalance the Base pool, xAlpha can detach from alpha. Holders who assumed a wrapped token was a perfect proxy for the underlying can discover the gap at the worst possible moment. This is not unique to Rubicon; it is the cost of composability, and it is the reason that a token trading on Base is a feature and a risk in the same breath.

The TradFi rail: ETFs that cannot stake

While crypto-native rails carried subnet tokens outward, traditional finance built its own on-ramp to TAO itself, and it comes with a strange asymmetry. In the United States, Grayscale has been working to convert its Bittensor Trust into a spot exchange-traded fund that would list on NYSE Arca under the ticker GTAO, with Coinbase and BitGo named as custodians and a management fee around 2.5 percent, per its SEC filings. The product’s structure holds the asymmetry: its registration statement bars the trust from staking the TAO it holds. An investor who buys the US fund would get price exposure and nothing else, no emissions, no alpha, no governance, while still bearing the full dilution of a token that inflates to pay stakers. As of early October there was still no approval, with a thinned-out SEC and a government shutdown slowing the pipeline, a backdrop we have been tracking in our coverage of who actually approves crypto ETFs now.

Europe took the opposite approach. Deutsche Digital Assets and the Swedish firm Safello launched what they call Europe’s first staked Bittensor product, the Safello Bittensor Staked TAO ETP (ticker STAO), which is 100 percent physically backed, held in cold storage, and actively stakes the TAO it holds, passing the staking yield into the product’s net asset value at a maximum fee of 1.49 percent. It trades on the SIX Swiss Exchange, Nasdaq Stockholm and, since mid-2026, Euronext Paris. So a European investor can buy a regulated wrapper that captures Bittensor’s emissions, while the proposed American wrapper, if it ever lists, cannot.

This is the staking paradox in one comparison. The product in the more permissive-sounding jurisdiction gives you the yield; the product in the jurisdiction with the deeper capital markets, so far, does not. For a token whose entire economic model is built on emissions, that is not a footnote. It is the difference between owning the business and owning a shadow of its price.

The Howey question travels with the wrapper

Wrapping a token, bridging it, or listing it on an exchange does not change what it is under the law, and that is the uncomfortable part of the export story. US securities analysis turns on the economic reality of an instrument, not the label or the format attached to it, a principle the SEC reiterated in its 2026 guidance on crypto assets. Under the Howey test, an arrangement looks like a security when people invest money in a common enterprise and expect profits from the efforts of others. Independent researchers have noted that Bittensor’s staking-and-delegation structure maps uncomfortably closely onto that definition: stakers delegate TAO to validators and receive a share of emissions generated by validators scoring miners, which resembles arrangements the SEC has previously treated as securities.

If that analysis is live for TAO, it is arguably sharper for subnet alpha tokens. An alpha token’s value is explicitly a claim on emissions produced by the coordinated effort of a subnet’s miners and validators. Wrapping it as xAlpha on Base, or listing it on Kraken, does not soften that; it arguably strengthens the expectation-of-profit-from-others prong by making the token easier to buy for people who will never run a node. The SEC has not ruled on subnet tokens, and nothing here is a prediction that it will. But holders should not assume a token has escaped securities scrutiny simply because it now has an ERC-20 address and a USDC pair.

Europe offers a cleaner, if stricter, frame. Under the bloc’s crypto rulebook, TAO and its subnet tokens would most likely be treated as ordinary crypto assets rather than as e-money or asset-referenced tokens, with the heaviest obligations falling on the platforms that offer them. We walked through what that regime actually requires of investors and venues in our guide to MiCA in 2026. The contrast is telling: Europe has a box to put these tokens in and a staked ETP already trading, while the United States has an unresolved classification and an ETF that is afraid to stake.

The governance shadow over the whole market

There is one risk that sits underneath all the others and travels with every wrapped token, every exchange balance and every ETP: the network that issues these tokens is not yet as decentralized as its marketing implies, and its own founder says so. Jacob Steeves and co-founder Ala Shaabana stepped down from executive roles at the Opentensor Foundation in February 2026, remaining as contributors, and laid out a roadmap to hand control to the community by the end of 2027, as Cryptobriefing reported. That is a candid admission that full decentralization is a destination, not a present-tense fact.

The sharpest version of the critique came in April 2026, when the team behind the Templar subnet, operating as Covenant AI, quit the network and shut their subnets. Founder Sam Dare told The Block that the promise that no single entity controls Bittensor “is a lie,” adding, “It is not. It is decentralization theatre.” TAO fell about 15 percent in the hours after the exit, sliding from roughly $338 to $285. Whatever one makes of the specifics, the episode showed how tightly subnet-token prices are coupled to confidence in the core network’s governance.

Investors who underwrite these tokens for a living have made the same point in blunter language. Mark Creaser, chief executive of the DSV Fund, told tao.media that the protocol’s habit of changing its emission rules mid-stream makes the tokens hard to value: “A casino at least keeps the same rules all night. Bittensor changes the table mid-hand and calls it an upgrade,” he said, while his colleague Siam Kidd called dTAO “basically uninvestable.” The through-line is that a subnet token is a claim on an incentive schedule that a small group can still rewrite. Exporting the token to Base or Kraken widens the audience for that claim; it does not change who holds the pen.

What to watch next

A few concrete markers will tell you whether the export of subnet tokens is maturing into something durable or just spreading risk to new venues.

• A US ETF decision. An approval of GTAO, with or without a staking carve-out, would be the clearest signal that regulators have gotten comfortable with TAO’s classification. A denial, or continued silence, keeps the no-staking wrapper as the only US option.
• Rubicon liquidity. Watch whether xAlpha pools on Aerodrome deepen well beyond their $300,000 seed. Thin pools that stay thin are a sign the DeFi rail is a demo; growing pools suggest real demand for composable subnet exposure.
• Revenue catching emissions. The SubConnect index put external revenue near 9 to 12 percent of emissions. If that climbs toward a third, the buyback loop starts to matter; if it stalls, the tokens remain emission plays.
• The decentralization clock. The 2027 handover is the governance catalyst. Concrete steps toward validator competition and community control would ease the decentralization-theatre critique that still shadows every token.
• More exchange listings. Kraken opened the door; whether larger exchanges follow will decide how much of the subnet-token market ever reaches mainstream buyers.

The larger story is that Bittensor spent 2026 learning to sell its subnet tokens to the outside world. The rails are real, the engineering is serious, and the convenience is genuine. What has not changed is the thing underneath the wrapper: a young network of AI markets, subsidized heavily by its own token emissions, still steering toward the decentralization it already advertises. The tokens left home this year. Whether they can stand on their own, away from the subsidy that raised them, is the question the next year will answer. For a fuller picture of how the network is trying to turn emissions into an actual economy, our look at Bittensor’s revenue era and the Gamma proposal is a useful companion to this one.

Frequently Asked Questions

What is a Bittensor subnet alpha token?

It is a token unique to one Bittensor subnet, created by the dynamic TAO upgrade in February 2025. Each subnet has its own alpha token and an on-chain pool that prices it against TAO; staking TAO into a subnet returns its alpha, and the token’s smoothed price helps set how much of the network’s emissions that subnet earns. Alpha is a claim on a subnet’s emissions and demand, not equity or a governance share in Bittensor.

How can I buy Bittensor subnet tokens in 2026?

There are three main routes. Natively, you stake TAO into a subnet’s pool on Bittensor to receive its alpha. Through Project Rubicon, subnet tokens are wrapped as xAlpha ERC-20 tokens on Base and traded on Aerodrome. And since June 2026, Kraken lists several subnet alpha tokens directly, which is the simplest route but usually does not pass through the staking yield. Each option trades convenience against yield and custody.

What is Project Rubicon, and is it safe?

Project Rubicon, launched by General TAO Ventures in November 2025, bridges Bittensor subnet alpha tokens to Base as liquid-staked xAlpha using Chainlink’s CCIP, with a one-to-one mint-and-burn design audited by Hashlock. It is a more conservative bridge design than the custom bridges behind past exploits, but it still carries cross-chain and liquidity risk, and the seed pools are thin, so wrapped tokens can trade away from the underlying.

Does a Bittensor ETF let me earn staking rewards?

It depends on the jurisdiction. The proposed US Grayscale GTAO ETF is barred by its filing from staking, so it would offer price exposure without emissions. Europe’s Safello Bittensor Staked TAO ETP (STAO) does stake the TAO it holds and passes the yield into its net asset value at a fee around 1.49 percent. That split is often called the staking paradox.

Are Bittensor subnet tokens securities?

Their status is unresolved. US securities analysis looks at economic reality rather than form, and Bittensor’s staking-and-delegation model maps closely onto the Howey test, which some researchers argue applies even more directly to subnet alpha tokens than to TAO. Wrapping a token as xAlpha or listing it on an exchange does not change its legal character. No US ruling has been made, so holders should treat the question as open.

By Marcus Okafor, HOGE Wire markets desk.