Targon

Targon is a decentralized AI inference marketplace running as Subnet 4 on Bittensor, built by Manifold Labs. It rents GPU/CPU compute backed by trusted execution environments (TEEs), specifically using Intel TDX and NVIDIA confidential computing to create Confidential Virtual Machines (CVMs). This makes it physically impossible for hardware providers to inspect users' data or models, addressing the key pain point of cloud services: users no longer need to trust the provider, but can rely on mathematically verifiable privacy instead of mere promises.

Over the past six months, Targon gained notable institutional recognition. In March 2026, Intel engineers co-authored a technical whitepaper with Manifold Labs on decentralized compute using Intel TDX and encrypted CVMs. Operationally, the platform reports over 20 billion paid inference tokens processed daily, backed by more than 1,500 H200 GPUs. Dippy AI, a major AI companion app, now routes roughly half of its text inference traffic through Targon under a production contract. Targon also claims 3.5x month-over-month revenue growth and completed a buyback of over 550 TAO using organic revenue. Meanwhile, skeptics question whether decentralized coordination can scale and maintain quality control against centralized cloud providers.