Subnet 4: Targon
Targon is Bittensor Subnet 4, a subnet focused on confidential AI compute. Its public materials frame the project as a confidential decentralized AI cloud built around Confidential Compute and Protected PCIe technology (Subnet 4 on TaoStats, Targon repository).
What the Subnet Provides
Targon’s project materials describe secure AI infrastructure rather than a model leaderboard or prompt marketplace. The core theme is protected execution for AI workloads: hardware-supported isolation, memory protection, secure boot, and remote attestation around compute used for model inference or related AI infrastructure (Targon repository).
That framing matters because the subnet is not just about finding raw GPU or CPU capacity. The reader-facing concept is protected compute capacity: AI work should be associated with a more verifiable execution environment than an ordinary remote machine.
Confidential Compute Context
Confidential compute is the security context behind Targon’s positioning. The repository emphasizes Confidential Compute, Protected PCIe, GPU trusted execution environments, memory isolation, and remote attestation as parts of the infrastructure stack (Targon repository).
This separates Targon from a generic compute-marketplace description. A generic marketplace asks whether compute is available. Targon’s public framing asks whether the compute environment can be protected and reasoned about when sensitive AI workloads run outside the requester’s own hardware.
Attestation and Protected Execution
Remote attestation gives the subnet’s compute story a verification angle. In Targon’s materials, attestation appears alongside hardware-level security guarantees, protected model execution, and verifiable computation. The point is not only that a machine returned a result, but that the environment around that result is part of the value being offered (Targon repository).
Protected execution also keeps the article from overstating the service as general AI hosting. The distinctive claim is narrower: Targon centers secure infrastructure for AI workloads that benefit from isolation and attestation.
On-Chain Identity
Live SN4 data is available on TaoStats. The Finney identity reports the subnet name as Targon. It lists the GitHub repository as manifold-inc/targon, which documents the confidential-compute behavior described above (btcli reference).
Concept Boundaries
Subnet 4 should be read as a Targon overview, not as a hardware compatibility matrix, or service availability dashboard. The stable concept is the combination of Bittensor subnet incentives with confidential-compute infrastructure for protected AI workloads (Targon repository, Subnet 4 on TaoStats).
Exact hardware availability, deployment details, pricing, and project-specific operating details belong in the project’s own materials. The stable distinction is that Targon is a confidential-compute subnet, not a general AI application article.
Miner and Validator Roles
Subnet 4 operates under the standard Bittensor two-role structure. Miners supply the subnet’s capability and validators evaluate those contributions and set weights. Reward distribution follows Yuma Consensus.
Protected Compute, Not Raw Capacity Listings
The same public materials emphasize Confidential Compute, Protected PCIe, memory isolation, and remote attestation as infrastructure guarantees. Availability of ordinary compute hours alone does not map onto Targon’s stated value proposition without those protection and verification features.
Attestation Is Part of the Offered Environment
The repository describes remote attestation alongside hardware-supported isolation and verifiable computation. A returned inference result should not be treated as equivalent evidence without the attestation and protected-execution context the project associates with its compute stack.
Distinction from Subnet Protocol
Targon is Subnet 4’s market on netuid 4: miners supply confidential compute with attestation and protected execution, and validators score those contributions for Yuma Consensus emissions. A subnet protocol is the separate rule set that defines how validators request work and miners return responses for evaluation on any netuid (Understanding Subnets). This article describes Targon’s market identity and confidential-compute focus; the protocol layer is the exchange mechanism validators and miners follow when moving attested compute tasks on SN4.
Distinction from Dynamic TAO
Targon on netuid 4 participates in Bittensor emissions as one subnet market among many. Dynamic TAO names the broader tokenomics model where each subnet has its own alpha context alongside network TAO; this article covers Targon’s confidential-compute positioning, not the full Dynamic TAO mechanics (Emission).
Distinction from Yuma Consensus
Targon on netuid 4 is a confidential-compute subnet where miners supply protected AI infrastructure with hardware isolation, secure boot, and remote attestation for sensitive workloads. That confidential-execution vocabulary names how Targon measures verifiable protected compute on netuid 4, not the on-chain step that turns validator weight submissions into emission shares each tempo (manifold-inc/targon, Yuma Consensus).
Yuma Consensus is the on-chain mechanism that runs at the epoch boundary on the selected netuid. It reads the weight matrix from eligible validators, applies clipping and bonding, and converts the result into miner incentives and validator dividend shares (Yuma Consensus, Emission).
Earlier sections separate protected PCIe and GPU trusted execution environments from ordinary remote GPU rental. Validators score whether compute capacity can be attested and reasoned about when models run outside the requester’s hardware. Yuma aggregates the weight signals that flow from that infrastructure verification path rather than provisioning confidential VMs itself (Subnet 4 on TaoStats, Understanding Subnets).
Remote attestation and memory isolation therefore sit upstream of settlement. Yuma still credits emissions from included validator weights on netuid 4 once the tempo closes (manifold-inc/targon, Emission).
Readers should treat Targon as the confidential AI compute protocol on netuid 4, and Yuma Consensus as the recurring mechanism that allocates rewards from validator weights each tempo.
- Subnet 4 (Targon) — confidential-compute AI infrastructure subnet with attestation.
- Yuma Consensus — epoch-boundary settlement from validator weights to emission shares on a task subnet (Yuma Consensus: Validator emissions).
Reader Boundary
Subnet 4 Targon should not be read as a generic GPU marketplace or prompt-leaderboard subnet. The Targon repository frames SN4 around confidential decentralized AI compute with protected execution for sensitive workloads rather than undifferentiated remote capacity.