Targon
SN4Run AI models in a secure cloud where nobody, not even the host, can see your data
The first subnet to co-author a technical whitepaper with Intel. Targon is building a confidential compute marketplace where AI runs in hardware-enforced secure enclaves, and nobody, not even the server operator, can see your data.
// AI compute you can actually trust.
Targon is a marketplace for renting GPU computing power to run AI models. What makes it different is security: every computation runs inside a hardware-protected environment called a Confidential Compute enclave. Your data and models stay encrypted even while they're being processed.
The simple version: It's like renting a safe deposit box at a bank, except instead of storing valuables, you're running AI models inside it. The bank (server operator) can't open the box or see what's inside, but you still get full use of it.
Centralized equivalent: Think AWS Nitro Enclaves or Azure Confidential Computing, but decentralized and priced through competitive bidding. Targon claims access to H200 GPUs for as low as $1.07/hr through this auction mechanism.
How it works:
- Miners provide GPU hardware with Confidential Compute capabilities (NVIDIA CC or PPCIE). They compete in price auctions to win compute jobs, driving costs down.
- Validators verify that the hardware is genuine, check attestation proofs, and route jobs to the most cost-effective providers.
- The problem it solves: Companies want to use AI but can't risk exposing proprietary data or models to cloud providers. Healthcare, finance, and defense all need compute they can verify is secure.
- The opportunity: The confidential computing market is growing rapidly as AI adoption hits regulated industries. Every enterprise running sensitive AI workloads is a potential customer.
- The Bittensor advantage: A decentralized network of verified secure compute providers creates competition on price while maintaining security guarantees. No single provider controls the infrastructure.
- Traction signals: Co-authored whitepaper with Intel on hardware verification and incentivization. 8,104 holders. 12,509 TAO net inflow over 7 days. The team includes Rob Myers (CEO, ex-OTF core contributor) and James Woodman (COO, ex-COO of OTF).
Category: Inference and Compute | Centralized Competitor: AWS Nitro Enclaves, Azure Confidential Computing, Google Confidential VMs
Targon sits at the intersection of two trends: the explosion in AI compute demand and the growing need for verifiable security in cloud infrastructure. Most decentralized compute subnets compete on price alone. Targon competes on trust, using hardware-level encryption to guarantee that data stays private during computation.
Mechanism:
The subnet operates an auction-based marketplace. When a job comes in, miners bid on it. Validators orchestrate the bidding rounds, verify hardware attestation (confirming the GPU actually supports Confidential Compute), and award the job to the lowest qualified bidder. This drives prices down while maintaining a security floor that centralized providers struggle to match at the same cost.
The codebase is written in Go with 1,978 commits across 19 contributors. Development has slowed in recent weeks (0 commits in the last 4 weeks), though this may reflect a transition from active development to operational focus as the network matures.
Financially, Targon is one of the stronger subnets in the ecosystem. With a root proportion of 0.164, the vast majority of its 302,415 TAO market cap comes from organic demand rather than protocol subsidy. It captures 6.3% of total network emissions and has a Gini coefficient of 0.686, indicating moderate holder concentration. The emission buy percentage sits at 31.5% with an acceleration ratio of 1.62x (chain buys outpacing the EMA), suggesting active accumulation.
The team's pedigree is notable. Rob Myers is a former Opentensor Foundation core contributor, and James Woodman served as COO of OTF. This is an insider team that understands Bittensor's architecture deeply, which explains the Intel partnership and the technically sophisticated approach to hardware verification.
- Development stagnation: Zero commits in the last 4 weeks. The codebase may be stable, but continued silence could signal a plateau.
- Competition: AWS, Azure, and Google all offer confidential computing products with massive existing customer bases. Targon needs to prove its pricing advantage is sustainable.
- Concentration: HHI of 0.074 and Gini of 0.686 suggest a moderately concentrated holder base. Large exits could move the price significantly.
- Enterprise adoption unproven: The Intel whitepaper is a credibility signal, but there's no public evidence of enterprise customers using the network at scale yet.
