Bittensor’s Gamma: Can Subnets Build a Closed-Loop Economy?

A token that would stop hitting the market

Bittensor spent most of 2026 arguing with itself about a single number: the gap between what its network earns from customers and what it pays out in token emissions. Every block, the protocol mints fresh TAO and dozens of subnet tokens, and most of that reward eventually gets sold so miners and validators can cover their costs. That steady outflow is the quiet tax on every holder, and it is a big reason a network that keeps shipping working AI products has still watched its token sag. TAO trades near $259.77, roughly 66% below its March 2024 peak of $757.60 and down close to 9% on the week, for a market value around $2.95 billion, according to CoinGecko.

At the Exploit Summit in Montreal on 28 September, the OpenTensor Foundation put a different idea on the table. Instead of emissions flowing out to the market and coming back as sell pressure, what if subnets spent those emissions on each other? The proposal is called Gamma, and it is the most consequential change to Bittensor’s token design since dynamic TAO arrived in early 2025. It is also, for now, mostly a slide and a promise. This piece takes Gamma apart: what it is, why the network thinks it needs it, and the question it cannot answer yet, which is whether an internal currency creates real demand or simply moves a subsidy around in a circle.

What Gamma actually is

Gamma is a credit, not a coin you trade. A subnet creates Gamma by burning its own alpha token, the per-subnet asset that dynamic TAO gives every project. The quantity is fixed to a dollar value at the moment of the burn. A KuCoin research explainer describes it as a usage credit you mint by burning subnet alpha, “priced against a stable oracle basket, so $5 of Alpha becomes 5 Gamma that never changes in quantity once minted.” Once created, a unit of Gamma holds its nominal value; it does not float the way alpha does.

Three design choices make Gamma unusual. First, it is a one-way trip. As the same explainer puts it, Gamma “can’t be converted back to Alpha and only the subnet owner key can accept it, which means 100% of the value capture flows through Alpha.” Burning alpha to mint Gamma permanently removes that alpha from circulation, so every Gamma payment doubles as an alpha buyback and burn. Second, only a subnet owner’s key can receive it, which makes Gamma a settlement instrument between projects rather than a retail token you would hold in a wallet. Third, it is meant to run on chain as one atomic transaction from any chain, with no invoices, no API keys, and no off-chain trust between the two subnets.

Put plainly, Gamma turns the informal buyback-and-burn that a handful of profitable subnets already do by hand into a native payment rail. A subnet that wants compute from another subnet would no longer have to sell its treasury on the open market, convert to fiat, and wire it; it burns alpha, mints the credit, and pays the provider directly on chain. The reward token stops being something you liquidate and starts being something you spend.

The obvious question is why a subnet would not simply pay in a stablecoin or in TAO itself. The answer is the whole point of the design. Paying in USDC would move dollars but do nothing for the seller’s token; paying in TAO would spread any benefit across the entire network rather than the subnet that did the work. Gamma forces every payment through an alpha burn, so the act of buying a service is also an act of tightening the seller’s own supply. It is less a currency than a purpose-built settlement token that happens to resemble one.

Why Bittensor needs an internal currency

To see why the foundation is reaching for this, follow the money. The protocol emits roughly 3,576 TAO a day across about 129 active subnets, per a mid-October ecosystem roundup, split 18% to the subnet owner, 41% to miners, and 41% to validators and their delegators, all paid in alpha, as the Taostats documentation lays out. Miners and validators carry real costs, GPUs, power, and payroll, and most of them meet those costs by selling the alpha and TAO they earn. Emissions, in other words, are manufactured sell pressure by design, and the only durable offset is customers.

Here the picture is better than skeptics once allowed, but still thin. SubConnect’s first Bittensor Revenue Index counted 24 subnets generating between $28 million and $35 million in annualized external revenue, with compute and infrastructure subnets producing roughly 78% to 82% of it, and 14 of the 24 using customer income to buy back their own tokens, Crypto Briefing reported. Set that against annual emissions worth something north of $300 million and the external revenue works out to somewhere around 9% to 12% of what the network pays out every year.

Independent trackers are harsher still. Analytics shop Pine, surfaced by ownyourmind, has noted that Bittensor’s single largest subnet collects around $52 million a year in emissions while booking only a couple of million dollars in verifiable external revenue, a subsidy ratio in the range of 22 to 40 times, and puts the network’s break-even inference price near $1.41 per million tokens against about $0.88 at competitive centralized providers. The uncomfortable one-line summary is that most subnets are, in effect, paid to exist rather than paid to serve. Gamma is the foundation’s attempt to change the verb. The same value-capture problem shadows every decentralized-compute token; Akash has spent the year wrestling with whether its burn actually captures the demand it serves.

The full-stack intelligence network

Gamma does not arrive on its own. The foundation framed it as the plumbing for a larger ambition it calls the full-stack intelligence network, a map in which subnets specialize into six layers, data, compute, storage, pre-training, inference, and AI agents, and then buy services from one another instead of each project rebuilding the whole stack, according to Crypto Briefing. An agent subnet would rent inference from an inference subnet, which rents GPUs from a compute subnet, which parks its data on a storage subnet, with each handoff settled in Gamma. The analogy the foundation keeps reaching for is cloud computing, where a startup stitches together storage, compute, and APIs it never built, except here the vendors are other crypto-incentivized networks.

The rhetoric in Montreal was deliberately grand. The event drew around 400 attendees and more than 70 subnet teams to New City Gas over two days, Yahoo Finance reported. “The question is no longer whether machines can be made intelligent,” said Etienne Leroy, Director of the OpenTensor Foundation and Co-Director of Bittensor Commons. “It is who gets to build the systems that produce that intelligence, and on what terms.” He was careful to add that the foundation was not claiming to have settled that, only putting the argument on a stage in a city with deep roots in AI research. The same roadmap restates the project’s long-held goal of handing control to a democratized governance model by December 2027.

From dTAO to Gamma: a three-token stack

Bittensor’s token model has added a layer about once a year, and Gamma would make three. TAO is the base asset, capped at 21 million like Bitcoin and subject to a supply-triggered halving; the first, in December 2025, cut the block reward from 1 to 0.5 TAO, the same kind of scarcity logic that stock-to-flow models try to price for Bitcoin. Alpha tokens arrived with dynamic TAO in February 2025, giving each subnet its own market-priced asset whose price sets that subnet’s share of emissions, as CoinGecko’s subnet primer explains. Gamma would sit on top as a stable, non-tradeable credit used only for payments between subnets.

The three are easy to run together in your head, so it helps to line them up.

The key move sits in the bottom two rows. Gamma is the first piece of the stack designed not to be sold.

How a Gamma payment would work

Picture a subnet building an autonomous trading agent. Today it needs inference to run models, storage to keep state, and specialized market data. It can build all three badly in house, or it can pay three other subnets that already do them well. The second option has been awkward, because paying meant selling alpha for cash and managing off-chain contracts with teams that may be pseudonymous. Under Gamma, the agent subnet’s owner routes the payment natively: burn some alpha, mint the credit, and send it to each provider’s owner key in a single on-chain action.

The more interesting wrinkle is automation. The KuCoin explainer notes that a subnet owner can redirect part of miner emissions to automatically buy other subnets’ Gamma, so a product team can pull in inference, compute, and storage from across Bittensor without raising outside capital. A subnet’s emission stream, instead of being dumped for dollars, becomes a procurement budget. Seven infrastructure and compute subnets, including Chutes, Targon, Lium, Affine, Engy, GM, and Hippius, have been floated as the likely first Gamma issuers, the sellers of compute that everyone else would pay, Startup Fortune reported. On the other side of the ledger would sit the application and agent subnets that consume those services.

Because settlement is on chain, and Bittensor’s tokens already travel to other networks (subnet alpha reaches Base through a Chainlink bridge), each new rail is also a new place for value to leak, a reminder that cross-chain plumbing remains crypto’s richest honeypot. If it works as drawn, the money that used to leave the system as sell pressure instead circulates inside it, and every lap burns alpha. That is the bull case in one sentence. The bear case starts with the same sentence.

The flywheel, if it spins

Strip away the vocabulary and Gamma is a buyback engine with a throttle. Every credit minted burns alpha, and because Gamma cannot be redeemed back into alpha, that burn is permanent. The more services subnets buy from each other, the more alpha comes out of circulation, which, through the dynamic TAO pools that tie each alpha token back to TAO, should tighten supply across the whole stack. The foundation’s pitch is that this quietly converts routine operating spend into continuous, automatic buy pressure, the mirror image of the sell pressure emissions create today.

The math only flatters the token if two things hold. The internal spend has to be large relative to emissions, and it has to be funded, at least in part, by money from outside the network. A flywheel spun entirely by recycled emissions adds no net demand; it just relabels a subsidy as turnover and burns a little alpha on the way around. That is the same trap as the revenue debate, seen from the supply side instead of the demand side. A burn financed by a paying enterprise is deflationary in a way that matters; a burn financed by freshly minted emissions is closer to moving money from one pocket to another and setting a small part of it on fire.

This is why the decision to route 100% of value capture through alpha is both clever and revealing. It guarantees that if real demand ever shows up, alpha holders capture it cleanly, with no leakage to a separate speculative asset. It does nothing to guarantee the demand itself. Gamma is, in effect, a very well-built pipe; whether anything valuable flows through it depends on the part of the business the proposal does not control.

The circular-demand problem

Here is the hard question Gamma raises. If a subnet funds its purchases by redirecting its own emissions, and the subnet it buys from also lives on emissions, then Gamma can move a subsidy around the network and make it look like commerce. A compute subnet that receives Gamma from ten emission-funded application subnets has booked ten customers, but none of them paid with money that came from outside Bittensor. The revenue is real in the narrow sense that alpha was burned and value changed hands. It is not real in the sense that matters for a token’s long-term price, which is net dollars flowing in from people who are not already inside the system.

This is why the distinction the SubConnect index draws, between external revenue and emissions, does not disappear under Gamma; it just gets harder to see. On chain, a Gamma payment looks identical whether the buyer earned its budget from a paying enterprise or minted it out of subsidy. The one metric that would separate the two, the share of Gamma ultimately backed by external cash rather than recycled emissions, is not something the proposal measures yet. Without it, a wave of inter-subnet Gamma volume could be read as a sign of a thriving economy or as a subsidy chasing its own tail, and the chart would look the same either way.

There is a rough test an analyst can apply in the meantime. Follow a unit of Gamma backward: if the alpha burned to mint it was bought with revenue from a named, external customer, the demand is real; if it was bought with alpha the subnet received as emissions, it is subsidy wearing a new coat. Doing that at scale is hard, which is precisely why self-reported numbers have been so slippery in this ecosystem, and why a figure that nets out the circular flows, rather than one that simply counts gross Gamma volume, is the only number worth trusting.

Skeptics have already framed Gamma as treating a symptom rather than the cause. The cause, they argue, is that most subnets have very few outside customers, and a slick internal payment rail does not conjure them. The optimistic reading runs the other way: cheap, frictionless internal settlement is exactly what lets a subnet specialize deeply enough to eventually be worth paying from outside, the way cloud primitives made it possible for startups to exist at all. Both can be true at the same time. The thing to watch is which one the data supports a year from now.

What the revenue numbers actually say

Because Gamma’s whole premise rests on real demand, the revenue question is not academic. The most credible public accounting is SubConnect’s index, and it rewards a close read rather than a headline glance.

A few things stand out. The money is concentrated: compute and infrastructure subnets produce roughly four-fifths of the total, and the top three alone account for about half. Real enterprise names have shown up, including PwC France, Dropbox, and an unnamed NYSE-listed REIT, per Crypto Briefing’s roadmap coverage. And the behavior Gamma wants to encourage, using income to buy back the token, is already happening at 14 of the 24 revenue subnets, which is the detail that makes the proposal feel like a formalization of something organic rather than a top-down experiment.

The caveats matter just as much. Independent trackers note that the most-emitting subnet, Chutes, self-reports throughput of roughly 160 billion tokens a day, while the externally verifiable figure (traffic routed through the aggregator OpenRouter) implies a revenue band closer to $1.1 million to $5.6 million, well under its emission take, per ownyourmind. The gap between claimed and verified demand is the oldest story in this network, and Gamma does not close it. If anything, a token that makes internal payments effortless raises the bar for proving that a given dollar of revenue came from a genuine, outside customer.

Gamma and Root Reborn: two fixes for one leak

Gamma is not the only 2026 mechanism aimed at the sell-pressure leak, and seeing how it fits matters. Root Reborn, which reached mainnet in August and switched its headline feature on in later releases, attacked the problem from the validator side: it stopped the network’s root dividends from auto-selling to TAO every block and instead let them accrue in per-validator escrowed baskets that holders realize only when they claim, turning validators into something like fund managers who allocate across subnets, as CoinDesk described when the proposal first surfaced. The earlier emission gate throttles rewards to subnets that cannot show usage at all.

Gamma attacks the same leak from the subnet side. Where Root Reborn changes when and how validator yield hits the market, Gamma changes whether subnet emissions hit the market at all, by giving them somewhere else to go. The three are meant to be complementary: the emission gate decides who gets paid, Root Reborn decides how validator pay is realized, and Gamma decides whether subnet pay is spent internally or sold. The yield mechanics rhyme with debates in liquid staking, where who controls the validator and how rewards are realized shapes what holders actually keep, a tension laid bare in the Lido, Rocket Pool, and Frax comparison.

Taken together, they describe a protocol trying, upgrade by upgrade, to convert a high-inflation incentive machine into something that behaves like an economy. Whether that many moving parts can be reasoned about by the people who actually hold the token is a separate question, and the critics have not been quiet about it.

The peg nobody has specified yet

Gamma’s stability is doing a great deal of work in the design, and it is the part with the least detail attached. A unit of Gamma is supposed to hold its value because it is priced against a stable oracle basket at the moment of minting. That short phrase hides at least three unanswered questions. What is in the basket: dollars, a stablecoin, a weighted set of assets? Who runs the oracle that prices it, and what happens if that oracle is wrong or gets manipulated during a volatile stretch? And if Gamma is minted at a fixed dollar value but the alpha burned to create it later collapses, does anything reconcile the two, or is the loss simply borne by whoever burned?

Pegged credits are not new, and neither are their failure modes. Holding a unit of account steady while the collateral behind it moves is the central problem of stablecoin design, and the market has spent years learning that the mechanism matters more than the promise, a lesson written all over the way stablecoin AMMs are engineered to hold a peg. Bittensor would be importing that problem into the heart of its inter-subnet economy. Done well, an oracle-pegged credit gives subnets a predictable price to plan around. Done badly, it becomes a fresh attack surface in a network that has already had its share of exploits and contested upgrades.

None of this means the idea cannot work. It is a reason to be wary of anyone who says it already does. As of mid-October, Gamma is a design reveal with a compelling diagram and no published specification for issuance, supply, redemption, or oracle governance. The agenda from Montreal listed a 20-minute deep dive led by co-founder Jacob Steeves, now also running the Affine subnet, but no numbers for how much alpha a subnet can burn, how fast, or who sets the basket.

Governance, and a mechanism that keeps changing

Who decides how Gamma works is inseparable from whether to trust it, because Bittensor’s recent history is a run of consequential rule changes pushed through quickly. That history is why some of the loudest voices in the room are not bullish.

Covenant AI, the team behind the Templar subnet, left the network in April 2026 and did not soften its language. Founder Sam Dare said the promise that no single entity controls Bittensor “is a lie,” and called the governance model “decentralization theatre,” alleging that core upgrades traced back to one party’s infrastructure, as The Block reported. The exit knocked TAO down about 15% intraday. Investors have raised a narrower but related complaint: the rules move. Mark Creaser, CEO of DSV Fund, has said that “a casino at least keeps the same rules all night,” arguing that Bittensor “changes the table mid-hand and calls it an upgrade,” while his colleague Siam Kidd called dynamic TAO “basically uninvestable” after one round of emission changes, per tao.media.

Gamma, however elegant, is another mid-game rule change, the third major reweighting of how value moves through the system in under two years. The foundation’s answer is the December 2027 handover: ship fast now, decentralize control later. That is a coherent plan, and also an admission. Until the handover actually happens, Gamma’s parameters, the basket, the oracle, the burn ratios, and who can change them, would be set by the same small group whose control the critics are objecting to. A payment rail is only as neutral as its rule-makers, and right now the rule-makers are few and known.

What Gamma means for the ETF question

The one audience Gamma pointedly does not reach is the one Wall Street is busy building products for. US investors still cannot buy a spot TAO ETF. Grayscale’s Bittensor Trust remains a private placement for accredited investors; its most recent filing, dated 7 October, reports 253,100 new shares issued and 3,247,300 outstanding as of 13 October, against aggregate consideration of 4,777.99 TAO worth about $1.45 million, all under a Regulation D exemption rather than a public listing, according to its SEC filing. A conversion to a listed NYSE Arca fund has been filed but not approved, and a thin, shutdown-slowed SEC has not acted.

Even once a US product lists, its holders would be walled off from exactly the mechanics this article is about. Grayscale’s prospectus bars the trust from staking, so an ETF holder earns no emissions, holds no alpha, and would never touch Gamma; they own price exposure and nothing else. Europe’s staked-TAO products go a step further (the Safello and Deutsche Digital Assets ETP stakes the underlying and rolls the yield into net asset value), but even that passes through staking rewards, not subnet-level participation in any internal economy.

The gap is worth stating plainly, because it cuts against the triumphant framing. If Gamma succeeds, the value it captures accrues to alpha and TAO that are held and used inside the network, by stakers, subnet owners, and validators. A passive investor holding a wrapper captures only whatever of that leaks into the spot price. The internal economy and the financial product are two different bets, and Gamma, by design, widens the distance between them.

What to watch next

Gamma is at the stage where the right posture is curiosity, not conviction. A handful of concrete signals will tell you whether it is becoming real or staying a slide in a deck.

• A published specification. The proposal still lacks public mechanics for issuance, supply, redemption, and oracle governance; a formal spec or code proposal is the first real milestone.
• First live Gamma payments. Watch whether the named infrastructure subnets (Chutes, Targon, Lium, Affine, and peers) actually transact, and at what volume, rather than demo once and stop.
• The external-versus-internal split. The number that matters is how much Gamma is ultimately backed by outside cash rather than recycled emissions; if nobody reports it, assume the worst.
• External revenue growth. SubConnect projects the revenue-generating subnet count could climb toward 35 to 40 by year-end; a credible move past the current $28 million to $35 million band would make Gamma’s premise more than theoretical.
• The December 2027 clock. Every month that passes with parameters set by the foundation is a month Gamma remains a tool of the governance the critics distrust.

Some outlets have reported an early-2027 target for Gamma going live, though the foundation itself has kept the timeline deliberately vague, and dynamic TAO’s own path from concept to mainnet took the better part of a year. For now, Gamma is the clearest statement yet of what Bittensor wants to become: not a token that pays people to show up, but a market where machines pay each other to do work. The idea is genuinely good. The proof is not in yet.

Frequently Asked Questions

What is Bittensor’s Gamma token?

Gamma is a proposed payment credit that would let Bittensor subnets buy services from one another. A subnet mints Gamma by burning its own alpha token; the credit is fixed to a stable value at mint, cannot be converted back to alpha, and can only be received by a subnet owner’s key. It is a settlement rail between projects, not a tradeable coin.

How is Gamma different from TAO and alpha tokens?

TAO is Bittensor’s base asset, capped at 21 million and traded on exchanges. Alpha tokens are per-subnet market assets created by dynamic TAO, and their price sets each subnet’s emission share. Gamma is neither: it is a non-tradeable, stable credit minted on demand by burning alpha, used only to pay for compute, inference, and storage between subnets.

Does Gamma fix Bittensor’s emissions sell-pressure problem?

It is designed to. By letting subnets spend emissions on each other instead of selling them for cash, Gamma could keep money circulating inside the network, and because minting burns alpha, every payment doubles as a buyback. The open question is whether that demand is real external revenue or recycled subsidy, which the proposal does not yet measure.

When will Gamma launch?

There is no confirmed date. Gamma was unveiled as a design concept at the Exploit Summit on 28 September 2026 and remains at the proposal stage, with mechanics such as its oracle peg and supply rules still unspecified. Some outlets report an early-2027 target, but the OpenTensor Foundation has kept the timeline vague.

Can I buy Gamma or get exposure through a TAO ETF?

No. Gamma is not tradeable and only subnet owner keys can hold it. There is also no US spot TAO ETF yet; Grayscale’s Bittensor Trust is a private placement for accredited investors and bars staking, so even that product gives holders price exposure only, not emissions, alpha, or Gamma.

By Marcus Okafor, HOGE Wire markets desk.