Introducing the Subnet Risk Index (SRI) Methodology

Bittensor is beginning to attract institutional capital. Funds, family offices, and active allocators are entering the ecosystem, and yet there is no standardised infrastructure for normalising risk across subnets.

Every allocator runs due diligence independently. There is no shared benchmark, no common vocabulary, and no way to compare risk profiles across the 100+ active subnets in a consistent framework. This is the gap that existed in traditional credit markets before rating agencies, and in DeFi before protocol risk frameworks emerged. Bittensor is at that inflection point now.

TAO Institute is building the risk infrastructure layer, starting with the Subnet Risk Index.

What SRI Is (and What It Is Not)

SRI is a risk framework. It measures structural risk exposure across quantifiable dimensions. It gives allocators a common language for the risk side of the equation.

SRI is not a subnet investment framework. It does not tell you which subnets to buy. A low-risk subnet can be a poor investment. A high-risk subnet can be a great one, if you are compensated for the risk. SRI is one input into an investment decision, not the decision itself.

Subnet teams will naturally want the metrics that make them look strongest. That is not what a risk index does. SRI measures what it measures, and publishes why.

Design Principles: Why Open Source and Trustless

The Subnet Risk Index is fully open source: code, weights, inputs, methodology.

Coverage is commercially supported; scoring is methodologically independent. Subnet teams that pay for extended coverage cannot influence their score. Anyone can run the same methodology and get the same output.

If TAO Institute is going to be the benchmark, the benchmark has to be trustless. This is the core underpinning for institutional trust.

The v1 Framework

Four Dimensions

Why This Weighting

Emission Viability (40%): In the current dTAO regime, a subnet's ability to attract and retain capital against its emission schedule is the most immediate determinant of structural survival. This is the most verifiable, most consequential signal in the near term.

Governance & Operations (15%): Not because governance is unimportant, but because the quantitative signals available today (validator concentration, freshness) capture a narrow slice of the full governance picture. As qualitative scoring matures, this dimension's contribution will grow.

Weighting is a design choice, not an objective truth. We are publishing our reasoning so it can be debated on its merits.

Scoring Mechanics

Composite: 0 to 100 scale, familiar to institutional allocators and trackable over time.

Percentile normalisation: After computing each dimension's raw average, scores are percentile-normalised across the scored cohort before weighting. This means SRI is relative: a score of 40 does not mean a subnet is "bad" in absolute terms. It means it sits in the lower portion of the current cohort on risk metrics. Every subnet cannot be above average. This is by design.

What SRI Does Not Capture

We are publishing the methodology's current limitations because naming them is stronger than having them named for us.

Revenue generation. This requires reporting directly from subnet owners, and/or revenue designed to be routed back via token buybacks. Until that data is available and verifiable, it sits outside the framework.

Team quality. There is no team assessment currently. The quality, experience, and track record of a subnet's team is a material risk factor that the current framework does not score.

Competitive moat. Whether a subnet has durable differentiation or is vulnerable to substitution is not captured quantitatively.

Demand-source classification. SRI does not yet distinguish between structural demand (agents staking TAO, cross-subnet composability, real switching costs) and convenience demand (cheapest inference today, easily substitutable). This distinction matters enormously for allocators.

Inter-subnet dependency. Subnets increasingly purchase services from each other. SRI scores each subnet in isolation and does not yet capture contagion risk from inter-subnet economic linkages.

Lagging indicators. Every metric in this framework is a lagging indicator. SRI captures where a subnet has been, not where it is going. Leading indicators (roadmap execution, partnership pipeline, demand trajectory) require qualitative assessment that is in development.

This Is the First Public Release

The methodology behind SRI has been through multiple internal iterations. This is the first public release.

We made specific choices and we are publishing those choices for scrutiny. The current framework is purely quantitative: every scored metric is objectively measurable from on-chain data.

We are publishing the full methodology, weights, and inputs because we expect scrutiny and believe the framework is stronger for it. Where the community identifies genuine gaps, we will incorporate them into future versions with full version control and a public changelog.

SRI Scores Are Free, Always

The SRI composite and dimension-level scores for every assessed subnet are publicly accessible at no cost.

If the framework has gaps we have not identified, we want to hear it. Feedback and methodology discussion are welcome in the [methodology repository].