The word "subnet" appears constantly in Avalanche documentation and gets conflated with several different things: custom blockchains, validator sets, Avalanche9000, and independent L1s. These are related but distinct concepts, and conflating them obscures what the mechanism actually does.
A subnet in Avalanche is a set of validators. Specifically, it is a dynamic set of validators working together to achieve consensus on the state of one or more blockchains. The subnet is not the blockchain itself — it is the validator group that secures it. One subnet can validate multiple blockchains. Every blockchain in the Avalanche ecosystem is validated by exactly one subnet.
Understanding this distinction is the foundation for understanding the system.
How Subnets Are Structured
The Avalanche network has a special subnet called the Primary Network. This subnet validates three built-in blockchains:
- X-Chain (Exchange Chain): Handles asset creation and transfers. Uses a UTXO model similar to Bitcoin. Runs the Avalanche consensus protocol.
- P-Chain (Platform Chain): Manages validators and subnets. All subnet registrations and validator sets are tracked here. This is where subnet metadata lives.
- C-Chain (Contract Chain): An EVM-compatible smart contract environment. AVAX transactions, DeFi, and dApps run here. This is the primary destination for Ethereum developers porting to Avalanche.
Every validator on the Avalanche network must participate in the Primary Network. Historically, this also meant holding at least 2,000 AVAX as stake — the minimum validator requirement for the Primary Network.
Custom subnets are created on top of this. A subnet creator submits a transaction to the P-Chain that defines the subnet. Validators then opt into that subnet independently, subject to whatever conditions the subnet creator specifies. A subnet can require validators to hold a specific token, pass identity verification, meet hardware benchmarks, or accept custom staking terms. The P-Chain records which validators belong to which subnets.
Each subnet runs its own blockchain with its own virtual machine. The most common choice is an EVM fork — functionally identical to Ethereum's execution environment — allowing Solidity contracts to run without modification. But Avalanche supports custom virtual machines. A subnet can run a completely different execution environment if it chooses, using Avalanche's consensus layer while defining its own transaction format and state logic.
Consensus Within a Subnet
Subnets use Avalanche's consensus protocol, not the Nakamoto PoW mechanism. The core mechanism is probabilistic subsampling: validators repeatedly sample a random subset of peers, checking what they believe the correct state is. If a sufficiently large fraction of sampled peers agree on a value, the validator shifts its own view toward that value. This repeats until the network converges. Convergence time is logarithmic in network size rather than linear — the protocol achieves finality in under two seconds under normal conditions.
This consensus model requires validators to be able to communicate with each other at low latency. Subnet validators communicate directly. A subnet with 20 validators can reach consensus far faster than a subnet with 2,000, because communication overhead scales with validator count. This is the fundamental trade-off: smaller validator sets mean faster finality but a smaller economic security base.
Security in a subnet is proportional to the economic value committed by its validators. A subnet with 10 validators controlling moderate stake is easier to attack than the Primary Network with thousands of validators. Subnets do not inherit Primary Network security — they establish their own. This is different from Polkadot's parachain model, where all chains share the security of the Relay Chain's validator pool.
Avalanche9000 and the Shift to Independent L1s
Before late 2024, every custom subnet validator was also required to validate the Primary Network. This meant meeting the 2,000 AVAX staking requirement regardless of the subnet's purpose. The practical effect: any project building a subnet was implicitly subsidizing Primary Network security through their validators' capital allocation.
ACP-77 (Avalanche9000), which went live on mainnet in December 2024, changed this. Under the new model, a subnet can register as an independent L1 on the P-Chain and declare that its validators do not need to co-validate the Primary Network. Validators for these independent L1s pay a continuous fee to the P-Chain for the service of being tracked — rather than locking 2,000 AVAX as Primary Network stake.
This changes the capital structure meaningfully. A project building an institutional settlement chain no longer needs validators who can each post 2,000 AVAX. They can define their own staking token, their own minimum stake, their own validator requirements — and operate entirely outside the Primary Network validator economy.
The trade-off is that independent L1 validators lose Primary Network validation revenue. Co-validating both the Primary Network and a subnet previously offered a combined reward stream. Independent L1 validators give this up.
Cross-Subnet Communication
Communication between subnets is handled by Avalanche Warp Messaging (AWM). The mechanism: each validator has a BLS (Boneh-Lynn-Shacham) key pair registered on the P-Chain. When a cross-subnet message is sent, validators on the source subnet sign the message with their BLS keys. The source chain aggregates these signatures into a single BLS multi-signature. The destination subnet verifies this multi-signature against the public keys registered on the P-Chain.
No external bridge contract is required. No third-party oracle. The trust assumption is: if enough validators on the source subnet signed the message, it's valid. The security of the cross-subnet message inherits the security of the source subnet's validator set.
AWM is a low-level primitive. Applications build on top of it — teleporter contracts abstract the BLS signature mechanics into a developer-friendly message-passing interface.
Where Constraints Live
The subnet model has three main constraint categories.
Economic constraints: The P-Chain continuous fee for independent L1s represents an ongoing cost. Fee levels affect the economics of small or low-activity subnets. As of early 2026, these fees are modest, but they introduce a running operational cost that co-validation with the Primary Network did not.
Security isolation: Subnet security is not shared. A subnet with a small validator set and low economic stake can be attacked at lower cost than the Primary Network. Projects that need high security must recruit sufficient validators and stake — this requires external trust-building, not just technical deployment.
Cross-subnet latency: AWM introduces messaging latency between subnets. Each cross-subnet transaction requires source-chain finality, signature aggregation, and destination-chain verification. This is functionally slower than single-chain transactions — similar to the IBC lifecycle on Cosmos.
What's Changing
Elastic scaling is on the Avalanche roadmap. The goal is to allow a single subnet to dynamically allocate blockspace across multiple cores as load increases — rather than being constrained by a fixed block cadence. This is conceptually similar to rollup scaling but at the subnet layer. As of early 2026, this is in development and not yet deployed to mainnet.
HyperSDK — Avalanche's framework for high-performance custom virtual machines — is gaining adoption among projects building non-EVM subnets. The framework enables significantly higher throughput than EVM-compatible chains by eliminating EVM overhead.
The practical trajectory of Avalanche9000 adoption is the most consequential near-term question. Whether independent L1s attract new projects (versus existing subnet teams migrating) will determine whether the model expands the ecosystem or primarily reduces existing validator capital requirements.
What Would Confirm This Direction
- On-demand coretime and independent L1 registrations growing on the P-Chain beyond existing subnet migrations
- Avalanche elastic scaling delivering demonstrable throughput improvements in testnet and mainnet environments
- AWM transaction volume growing across subnets, showing cross-subnet activity at scale
- New non-EVM HyperSDK deployments with sustained usage
What Would Break or Invalidate It
- Cross-subnet AWM exploit, particularly an attack that forges or replays BLS multi-signatures, would fundamentally undermine the cross-chain trust model
- P-Chain bottleneck at scale — if the P-Chain becomes a constraint on subnet registration or cross-subnet communication volume, the architecture faces a systemic limit
- Independent L1 fees increasing to the point where the model is no longer economically preferable to co-validation
- Competing frameworks (OP Stack, Cosmos SDK, Polkadot Coretime) capturing the customizable blockchain market at comparable cost and speed
Timing Perspective
Now: The Primary Network is operational, C-Chain has active DeFi and institutional activity, and Avalanche9000 is live. Existing subnets are evaluating migration to the independent L1 model.
Next: Elastic scaling delivery and HyperSDK adoption over 2026 will test whether Avalanche can serve use cases that require throughput beyond EVM limits. The independent L1 adoption curve is the near-term signal to track.
Later: Broader viability depends on whether the validator-customization model wins a meaningful share of the customizable blockchain market, or whether rollup frameworks and shared-security models capture most of that demand.
Boundary Statement
This explanation covers the subnet mechanism and how Avalanche9000 changed the architecture. It does not cover AVAX tokenomics, staking yields, or the economics of any specific subnet deployment. It does not constitute a recommendation regarding any Avalanche ecosystem project.
The mechanism works as described. Whether any specific subnet represents a useful system depends on factors outside this scope.
