Targon (SN4)

Overview

A compute network and a subnet token

Targon is a cloud platform for computing work, built around a network of independent hardware providers. Developers can use it to run artificial intelligence (AI) models, rent graphics processing units (GPUs), create virtual machines, and set up short-lived development environments. The network operates as Bittensor’s Subnet 4, often shortened to SN4. On Bittensor, each subnet has its own token, called an alpha token. Targon’s SN4 token is the alpha token linked to this compute network. (targon.com)

The platform and the token have different roles. Customers use Targon’s cloud services to get computing power. Hardware providers, known on Bittensor as miners, make machines available. Validators check those machines and help determine how subnet rewards are assigned. The SN4 token connects these participants through Bittensor’s reward and staking system. Understanding that link is the starting point for understanding Targon as a crypto project. (targon.com)

History & Team

From AI responses to a wider cloud platform

Manifold Labs, an Austin, Texas-based team founded in 2023, develops Targon. Early versions of Subnet 4 focused on serving responses from large language models. In 2024, the team released updates to improve how the network checked model responses and connected with applications. By 2025, its work had expanded toward confidential computing: running workloads on outside machines while using hardware-based methods to protect customer data. (targon.com)

A major step was the 2025 introduction of the Targon Virtual Machine, or TVM. Manifold later described a broader compute marketplace built around secure machines, rather than only an interface for AI model responses. Its March 2026 research paper with Intel set out how protected virtual machines, hardware checks, and subnet rewards work together. (manifold.inc)

In July 2025, Manifold announced a $10.5 million Series A funding round led by OSS Capital, with Digital Currency Group and several individual investors participating. In the same announcement, the company reported that Targon was serving more than 20 billion paid AI inference tokens per day. An inference token is a unit of text processed by an AI model; it is separate from the SN4 crypto token. The figure records a milestone from that announcement, rather than a measure of token activity. (manifold.inc)

Technology & How It Works

Connecting machines, customers, and validators

Targon combines a cloud-style user experience with hardware supplied by independent operators. A customer chooses a computing resource and sends work through Targon’s platform. A miner supplies the machine. Validators check whether that machine meets the network’s requirements and report their assessments to Bittensor, where those assessments help determine rewards. Targon also handles tasks such as preparing machines, assigning work, and giving customers ways to manage it. (targon.com)

Confidential computing is central to its design. In this approach, a workload runs inside a protected environment called a trusted execution environment. Targon’s TVM architecture uses Intel Trust Domain Extensions, or TDX, to isolate virtual machines. It also uses NVIDIA confidential-computing features for supported GPUs. Before an approved machine receives work, an attestation check uses cryptographic evidence to confirm its hardware and software state. The system repeats checks while the machine operates, rather than relying only on a check at startup. (targon.com)

The platform offers several ways to use compute. Its virtual machines provide a Linux environment with GPU options and SSH access, a standard method for connecting to a remote computer. Sandboxes are shorter-lived environments suited to testing code, creating previews, or running AI-agent tasks. Targon provides an application programming interface (API), software development kit (SDK), and command-line tools so developers can manage workloads through software as well as a dashboard. (docs.targon.com)

For providers, TargonOS is a purpose-built operating system that prepares a machine to join Subnet 4. Targon’s provider guide describes a process in which the machine checks its configuration when it starts, brings up encrypted storage, and launches network services. This helps turn separately owned hardware into a more consistent pool of computing resources. (docs.targon.com)

Tokenomics & Utility

How SN4 enters the network

Bittensor introduced subnet alpha tokens in February 2025. SN4 is Targon’s subnet token, while TAO is Bittensor’s main network token. Bittensor’s rules set a maximum supply of 21 million alpha tokens for each subnet. New SN4 tokens enter circulation through the network’s issuance process; part of that issuance is set aside for participants, and part can enter the subnet’s TAO–alpha pool. (bittensor.com)

Participant rewards follow roles. The subnet owner receives a share, miners can earn rewards for providing eligible compute, and validators and their stakers can receive rewards through Bittensor’s consensus process. Validators submit assessments of miners, which the chain combines to determine reward shares. Targon’s provider documentation also describes hardware-based reward targets: the amount available to a provider depends on the type of machine and the compute available on the network. Emissions that are not assigned to its auctions are burned. (bittensor.com)

SN4 also has a staking role. Staking TAO into Subnet 4 exchanges it through Bittensor’s pool for SN4 alpha and records the result as a subnet stake position. That position can receive a share of a validator’s staking rewards. This gives the token a place in the network’s incentive system alongside the work done by miners and validators. Targon’s customer billing, meanwhile, uses platform-level credits, payment methods, and invoices for organizations. (bittensor.com)

Ecosystem & Use Cases

What people build and run

Targon’s most direct use case is access to computing hardware. AI developers can use GPUs to serve models, run experiments, or prepare training workloads. Teams can launch a virtual machine when they need a familiar Linux server with direct access. They can use a sandbox when they want a temporary space to test software or give an AI agent its own working environment. These are different ways to use the same broader compute platform. (targon.com)

The provider side gives owners of suitable machines a way to contribute capacity to the network. Targon’s design aims to make that capacity available to customers without requiring each customer to arrange a separate contract with each hardware owner. Validators form the checking layer between supply and subnet rewards, while Bittensor records the incentive results. (targon.com)

Targon also supports team use. An August 2026 update introduced organizations that can own shared workloads, hold billing details, and grant members different access roles. Teams can use service tokens to let automated software connect to their resources. These features extend Targon beyond one-person experiments toward shared development work. (manifold.inc)

Advantages & Challenges

What the design makes possible

Targon brings together a pool of independent compute providers and tools that feel familiar to cloud users. Developers get virtual machines, sandboxes, and software interfaces without having to manage the subnet’s reward system themselves. Hardware-backed isolation and repeated attestation give the network a defined way to check machines that it does not own. The subnet incentive model gives providers a reason to supply compute and validators a role in checking it. (targon.com)

That design also creates demanding technical work. Protected GPU workloads depend on compatible hardware and several checks working together. Operators must bring machines onto the network, and validators must keep assessing them. Targon coordinates those independent parts while offering customers a usable cloud service. Its platform also includes choices between longer-running virtual machines and short-lived sandboxes, each suited to a different kind of task. (targon.com)

Where to Buy & Wallets

Acquiring SN4 through Bittensor

SN4 can be acquired through Bittensor’s Subnet 4 staking pool. Staking TAO into the subnet swaps TAO for Targon alpha and places the resulting tokens in a stake position. Unstaking reverses the process through the pool. This on-chain route is distinct from paying a Targon invoice or adding credits to a cloud account. (bittensor.com)

The Bittensor CLI can create a wallet and manage on-chain staking. Bittensor also lists wallet options such as Crucible Wallet, TAO.com Wallet, Talisman, and Nova Wallet for network activities. A Bittensor wallet uses a coldkey for holdings and financial actions; a hotkey serves as an operating identity for subnet participants such as miners and validators. Those two key types explain why running a Targon machine involves more setup than simply holding an SN4 stake position. (bittensor.com)

Regulatory & Compliance

Crypto rules and religious assessment

Targon’s token and cloud business touch different sets of rules. In the United States, the Securities and Exchange Commission’s March 2026 interpretation explains how federal securities laws apply to types of crypto assets and activities, including protocol mining and staking. The Internal Revenue Service also treats digital-asset transactions as reportable in relevant circumstances, including income from staking rewards. These are broad frameworks under which SN4 activity can be assessed; they are not a project-specific legal classification. (sec.gov)

In the European Union, the Markets in Crypto-Assets regulation, known as MiCA, establishes rules for crypto-asset offerings and service providers. It began applying broadly on December 30, 2024. Which obligations apply depends on the activity involved, such as offering an asset or providing a crypto-asset service. (esma.europa.eu)

A halal or Shariah-compliance assessment asks a different question. Malaysia’s Securities Commission Shariah Advisory Council, in guidance limited to assets within its jurisdiction, considers matters including how token issuance proceeds are used and what rights or benefits a token provides. Targon’s public description of SN4 explains its staking and reward functions but does not establish a project-specific Shariah classification. Its halal status therefore remains undetermined from the published project information. (sc.com.my)

Future Outlook

Linking network growth to useful compute

Targon’s direction is toward a broader computing platform: one that can bring outside hardware into a shared network and make it useful through ordinary cloud tools. Its move from AI model responses to confidential virtual machines, sandboxes, TargonOS, and team workspaces shows that progression. Future development can build on those pieces by improving how hardware joins, how customers deploy workloads, and how the subnet rewards useful supply. (manifold.inc)

For SN4, the important connection is between real compute work and Bittensor’s incentives. Miners contribute machines, validators check them, customers use the services, and alpha-token rewards help coordinate participation. How well these parts work together will shape Targon’s place in decentralized AI infrastructure. (targon.com)

Summary

Targon is Bittensor’s Subnet 4 and a cloud platform for AI and other computing workloads. Its SN4 alpha token supports staking and network rewards, while Manifold Labs builds the tools that connect customers with independent hardware providers. By pairing confidential-computing methods with Bittensor’s incentive system, Targon offers one model for turning a distributed supply of machines into a usable compute service. (targon.com)