Targon is built by Manifold Labs, a small team based in Austin, Texas, founded in 2023 by Robert Myers, an early Bittensor contributor and former Opentensor engineer, and James Woodman, formerly COO of the Opentensor Foundation. The company raised a $10.5 million Series A in July 2025, led by OSS Capital with participation from Digital Currency Group and angel investors from Google, Shopify, and the Bittensor ecosystem. The project is fully open-source and actively maintained.
What is Targon?
Targon is Bittensor subnet 4, a decentralized cloud for confidential compute built by Manifold Labs. It rents out GPUs and CPUs that run inside trusted execution environments, so a workload can be processed on hardware the customer does not own or trust.
The subnet is built on Intel TDX and AMD SEV-SNP for the CPU side and NVIDIA Confidential Compute or Protected PCIe (PPCIE) for the GPU side, with hardware-enforced memory encryption, secure boot and remote attestation — a cryptographic proof that the machine really is configured the way it claims.
In practice users get compute rentals over SSH, deployable virtual machines, and an inventory of available GPUs and CPUs, billed only for compute used rather than idle time. It is aimed at teams running AI training and inference on sensitive data, plus hardware operators who want to supply capacity to the network.
How Targon works?
Miners supply physical machines that meet the supported platform specs — AMD EPYC 9xx4 with SEV-SNP, or 5th-gen Intel Xeon with TDX plus NVIDIA H100, H200 or B200 GPUs.
Before a machine can earn, it runs the TVM installer (Targon Virtual Machine, the confidential VM image that miners boot), which attests the hardware configuration against Targon's attestation service and verifies that the confidential compute environment is genuine.
Each miner registers its nodes in a config file with an IP and a bid, quoted in cents per hour per GPU, plus a minimum alpha stake a validator must hold to be assigned nodes. Emissions are then distributed through an auction system: miner emissions are split proportionally to public demand for each hardware category, with per-category caps on maximum bid and emission share published through the subnet's auctions API.
Anything not allocated to an auction is burned, so unwanted or misconfigured capacity earns nothing.
Frequently asked questions
What is Targon?
Targon (SN4) is a decentralized cloud computing platform focused on confidential AI compute. It provides secure GPU and CPU rentals and serverless deployments where all workloads run inside hardware-encrypted Confidential Virtual Machines, ensuring that even hardware operators cannot access the data or models running on their machines.
How does Targon work?
Targon (SN4) operates a network of over 1,000 GPUs and CPUs where hardware providers contribute bare-metal servers running the Targon Virtual Machine (TVM). Each machine passes remote attestation through Intel Trust Authority before activation, and performs continuous re-attestation every 72 minutes. Users access compute through Rentals or Serverless deployments via a Python SDK, CLI, or API.
What is the main strength of Targon?
The main strength of Targon (SN4) is its confidential computing architecture. Every workload runs inside encrypted virtual machines using Intel TDX, AMD SEV-SNP, and NVIDIA Protected PCIe, where neither the hardware operator nor any external party can inspect the data being processed. The platform claims 99% uptime and sub-50ms latency.
