TAO’s Price Didn’t Move. The Network Did This Year.

This piece builds on a recent essay by Sami Kassab, Managing Partner at Unsupervised Capital, reflecting on Bittensor’s past year. The figures cited here are his, drawn from Unsupervised Capital’s internal dashboard. The framing and commentary are ours.

If the TAO chart is the only thing you’ve watched this year, you could be forgiven for thinking nothing happened. TAO is trading roughly where it opened in 2026. No breakout, no collapse: a sideways year that leaves holders quiet in the group chats.

The chart, though, has been telling you the least interesting part of TAO’s story. Underneath a price that went nowhere, Bittensor has changed a great deal, both in what subnets are doing and in how the protocol itself works.

Here’s what changed, and why it matters to you, whether you hold TAO, stake into subnets, mine, or build.

Start with what subnets are earning

At the end of 2025, the subnet ecosystem was generating roughly $20M in annualised revenue. Today, according to Unsupervised Capital’s internal tracking, that figure is closer to $54M. That’s well over double in less than a year.

What makes this more than a nice headline is the conditions it happened under. Subnets faced two headwinds at once:

1) A flat TAO price. Miners are paid in TAO, so a price going nowhere meant no dollar-value boost to their rewards.

2) The first halving. In December 2025, Bittensor’s first halving cut TAO issuance in half.

In other words, subnets had less economic incentive to produce, and they produced more revenue anyway.

That’s the opposite of what a network living purely on emissions looks like. And revenue is the clearest sign that subnets are selling something people actually want to pay for.

The one number we’d ask you to keep an eye on

Every day, new TAO enters circulation. For anyone holding the token, the key question is simple: who absorbs it?

Sami give a partial answer:

1) Buybacks by tracked subnet teams now offset 4.7% of TAO emissions.

2) Add subnet and miner registration fees plus transaction fees, and total demand offsets reach 10.9% of emissions.

Put plainly, roughly one in every nine new TAO is now matched by demand generated from activity on the network itself.

This is one of Bittensor’s most important metrics. The long-term goal is for economic activity across the network to create enough demand for TAO to offset new issuance entirely.

At 10.9%, Bittensor isn’t close to that yet. But the direction is what matters, and the trajectory is highly encouraging.

Who’s doing the heavy lifting

Much of the buyback activity comes from infrastructure subnets: those providing GPU compute, inference, storage and other foundational services.

As these products have become more competitive over the past year, two things have happened. Subnets have started buying services from each other, and they’ve begun attracting customers from outside Bittensor too. Sami describes this as a “network state” forming. A simpler way to see it is an internal economy where subnets are acting as one another’s suppliers and customers, with outside money starting to flow in.

This makes sense because an ecosystem whose participants depend on each other is harder to dislodge than a collection of standalone projects each chasing emissions.

2026 built the roads. 2027 is about what drives on them.

2026 will be remembered as the year of infrastructure, and 2027 as the year of intelligence.

By “intelligence”, subnets producing genuinely useful AI models and agents, built on top of the infrastructure laid down this year. Some subnets are already showing early signs of producing Pareto-optimal models, including Affine (SN120), IOTA (SN9), Ingot (SN97), Reliquary (SN81) and Teutonic (SN3). (In simpler terms, models that can’t be beaten on one measure, such as performance, without giving something up on another, such as cost or size.)

The rules of the game changed as well

While subnets were growing, the protocol underneath them was being reworked. Let’s explore two major changes:

1) How emissions are shared out. The mechanism that decides how TAO emissions are distributed across subnets, arguably Bittensor’s most important function, was overhauled. Two changes stand out: a return to price-weighted emissions, and concentrating the majority of rewards among the top 32 subnets. Sami argues this has made allocation considerably more capital efficient. If you stake into subnets, the practical takeaway is that rewards are now weighted heavily toward the top of the table.

2) The protocol as an investor. Emissions are no longer used only to add liquidity to subnet pools. The protocol now also uses them to buy subnet tokens. Since these protocol investments began in June, the network has bought subnet tokens with roughly 384,000 TAO, worth about $90.2 million at the time of purchase. In effect, Bittensor has built its own investment arm, giving founders another source of financing and making the network a more attractive place to build.

Yes, it’s been a lot to keep up with

If the pace of change this year has left you dizzy, you’re in good company. Sami says the criticism he hears most often is that the protocol changes too frequently, and he openly agrees: it’s hard to build a business when the rules keep moving, and founders need some certainty about the environment they’re operating in.

His counterpoint is that the churn has been necessary. Nobody has attempted what Bittensor is attempting; getting the economic incentives right takes trial and error, and locking in a flawed design too early would be worse than experimenting now.

The telling detail is that builders haven’t walked away. Existing founders have stayed through the volatility, and new ones keep arriving. In one instance, four new subnets launched this last week. That’s a strong vote of confidence in Bittensor as a place to build, despite the rough edges and all.

What’s still coming: Gamma and shorting

Two major upgrades remain on the near-term roadmap: Gamma and shorting. These changes could be the last big swings of this era.

After they ship, the core team at Opentensor Foundation would focus on decentralisation, and the pace of protocol development should slow to something builders can plan around. We’ll cover both upgrades in depth as they approach launch.

So what should you take from all this?

Bittensor is undoubtedly going into 2027 with a more stable protocol, a mature infrastructure layer, and a new generation of subnets starting to produce valuable intelligence.

After years of experimentation, the network is getting close to leaving “beta” and opening its doors to a new wave of founders.

Whether that plays out is something you can track yourself. Here’s our short watchlist:

1) Subnet revenue. Does the ~$54M figure keep increasing from here?

2) Emission offsets. Does the 10.9% keep rising toward full coverage of new issuance?

3) Gamma and shorting. Do they eventually go live, and does the protocol work with these upgrades?

4) The intelligence subnets. Do Affine, Iota, Ingot, Reliquary, Teutonic, and others turn strong models into paying users?

The price may not have changed much this year. But if those four lines keep heading the right way, the chart won’t be the most important thing to watch for much longer.