Loading
Loading
Incentivized Compute Marketplace powered by the Targon Virtual Machine (TVM).
Neuron registration trends
Organizations deploying AI workloads on shared infrastructure face a fundamental trust problem: they cannot verify that their models and data remain confidential during execution, nor can they prove the integrity of the hardware running their code. Current centralized cloud providers rely on contractual assurances rather than cryptographic proof, leaving sensitive AI applications vulnerable to insider attacks, side-channel leaks, and unauthorized access to proprietary models.
A decentralized network of hardware-verified compute nodes, where each provider's system attestation is cryptographically validated before it accepts work, shifts the burden of trust from contracts to mathematics. Remote attestation—a hardware-level proof that a machine runs authentic firmware and encrypted execution environments—eliminates the need for trust in any single operator. Validators perform independent attestation checks and only route work to miners whose hardware meets security requirements, while the network's open membership lets new providers join by meeting the same standards.
Targon implements this on Bittensor as a confidential compute subnet. Miners run TargonOS, a signed operating system that boots into a Trusted Execution Environment on UEFI machines with Intel TDX (or AMD SEV-SNP for validators), GPU TEE support, and TPM 2.0. Every boot, the node attests its hardware state to the network; validators verify attestation claims and only stake on miners that pass. Emissions flow through auctions—each GPU auction sets a target price and card count, and live capacity that hasn't reached the target receives full payment up to a maximum cap per card per hour. The network burns any emissions not allocated to active auctions.
Use cases were auto-generated from the GitHub repository and may not reflect the latest changes.
Targon is a confidential decentralized AI cloud platform that leverages Trusted Execution Environments (TEE) and Confidential Compute to provide secure, verifiable AI infrastructure. Miners run TargonOS on TEE-enabled hardware with GPUs to offer confidential compute capacity through encrypted inference, while validators attest and verify miner integrity using remote attestation and hardware-level security guarantees.
Targon is a three-tier system: (1) Miners run TargonOS on TEE-enabled hardware with GPUs, providing confidential compute capacity via isolated execution environments with NVIDIA Confidential Compute or PPCIE. (2) Validators run on AMD SEV-SNP CPU servers in VMs, performing remote attestation to verify miner hardware integrity and compute claims. (3) The network layer connects miners, validators, and the Bittensor blockchain via sn4 CLI tools and services. Emissions flow from auctions to miners based on card availability and pricing; validators gate eligibility through attestation. Communication occurs over DHCP-enabled networks with HTTPS to release and tower servers.
The auction system defines target_cards (maximum cards for auction), target_price (USD/hour/card), max_price (payment cap per card, never exceeded), and min_cluster_size (minimum cards required to participate). Prices are in USD cents per hour. Query auctions via https://tower.targon.com/api/v2/auctions.
Incentivized Compute Marketplace powered by the Targon Virtual Machine (TVM).
This information was auto-extracted from the GitHub repository and may be incomplete or out of date. Always refer to the subnet's GitHub repository for the latest instructions.
Any emissions not allocated to auctions are burned. Validators validate miner attestations and hardware claims.
The sections above were auto-generated from the GitHub repository and may not reflect the latest changes. View repository
FAQs were auto-generated from the GitHub repository and may not reflect the latest changes.
Holder & total-staked metrics as of 2h ago (hourly snapshot)
Requires Intel TDX enabled, TEE (Trusted Execution Environment) support mandatory. Installer wipes all eligible disks on the machine. Hardware must be dedicated to TargonOS. Specific GPU models (H200, B200) are referenced in auction targets with minimum cluster sizes (e.g., 8 cards minimum).
Basic compute resources required. Validator runs in a VM on the CPU server. AMD SEV-SNP hardware recommended but setup docs focus on CPU-only requirements.
Enable Intel TME, TME-MT, TDX, SEAM loader, and SGX in BIOS CPU Configuration
Configure BIOS with Intel TDX settings following the guide in installing-targonos.md
Obtain either the ISO (targonos-0.1.0-amd64.iso) for interactive install or chainload the iPXE netboot installer
Download TargonOS installer from https://releases.targon.com/0.1.0/
For iPXE: pass targon.install.unattended=1 and targon.install.hotkey=word1-word2-...-wordN. For ISO: boot from virtual media and follow interactive prompts including network setup, disk selection, and hotkey entry
Run TargonOS installer with hotkey seed phrase via iPXE or ISO boot
Optionally verify integrity of downloaded files before installation(optional)
Verify installer artifacts: sha256sum -c SHA256SUMS
Add TargonOS node IPs, external miner IP, port, hotkey phrase, chain endpoint, netuid, and minimum stake settings
Create or update config.json with node IP addresses and miner configuration
Start the miner service using Docker Compose
docker compose -f deploy/docker-compose-miner.yml up -d --build
Install sn4 CLI tool from Targon APT repository for management and monitoring(optional)
sudo apt install sn4
Install sn4 CLI from Targon APT repository on the validator CPU server
sudo curl -fsSL https://releases.targon.com/targon.gpg -o /usr/share/keyrings/targon.gpg && echo 'deb [signed-by=/usr/share/keyrings/targon.gpg] https://releases.targon.com/apt stable main' | sudo tee /etc/apt/sources.list.d/targon.list && sudo apt update && sudo apt install sn4
Set up environment configuration. Generate random credentials for Mongo using tr -dc A-Za-z0-9 </dev/urandom | head -c 24; echo. Optional variables include TIMEOUT_MULT, DISCORD_URL, NETUID, CHAIN_ENDPOINT
Create .env file in project directory with required variables: HOTKEY_PHRASE, MONGO_USERNAME, MONGO_PASSWORD, VALIDATOR_IP
Download and prepare the validator VM image on the AMD SEV-SNP enabled server
Run tvm/install binary with arguments: --hotkey-phrase '<phrase>' --node-type vali-cpu --submit --service-url http://tvm.targon.com --vm-download-dir /vms/manifold
Start the validator VM. To stop: ps -aux | grep qemu and sudo kill -9 [pid]
cd /vms/manifold && sudo launch_vm.sh
Initialize the validator with environment variables. Must be run on the server hosting the VM or use --ip flag. Confirm environment file when prompted
sn4 vali init [path to .env file]
Monitor validator logs after initialization(optional)
sn4 vali containers && sn4 vali logs --container [container name]
Update validator when new versions available(optional)
sn4 vali update