Launch Your AI Project Without Infrastructure Costs
Deploy your project on a decentralized AI network and earn AET. The network provides the compute, you build the product.
What neuronets are
Neuronets are the decentralized AI networks that AETRON is built from. Each neuronet runs as an independent AI marketplace tuned to one job: LLM inference, model training, image and video processing.
Each neuronet has three roles. The owner creates the neuronet and sets the rules. Miners serve requests and check each other through peer-replay. Stakers delegate AET into the neuronet - their stake both sets the throughput and acts as a vote on the service's usefulness. AI startups get to operate without owning infrastructure, while miners and stakers earn AET for real contribution.
Structure: Miners, stakers, and the neuronet owner - three roles with no cap on participants
Governance: Emissions are distributed by miners' confirmed work
Rewards: All three roles receive a share of the AET emission for their contribution
Neuronet Capabilities
A separate neuronet is built for each scenario. A few examples:
LLM inference
Serving language models with verifiable output
Chatbots, assistants, text generation and analysis, AI agents
Training and fine-tuning
Distributed model training with bit-exact step-by-step replay
Pre-training from scratch, fine-tuning, RLHF, dataset preparation
Content generation
Diffusion and flow-matching models for images and video
Stable Diffusion, FLUX, SDXL, Sora- and Wan-class video generators
Audio workloads
Recognition, synthesis, and transformation of speech and sound
Whisper for STT, TTS, voice conversion, music generation
Embeddings and computer vision
Vector embeddings and CV tasks with bit-exact reproducibility on CNNs
CLIP, DINOv2, ResNet, ViT, embeddings for search and recommendations
Private inference
Optional Confidential Computing mode with TEE attestation for sensitive data
Healthcare models, financial analytics, enterprise assistants
Monetization for Neuronet Creators
A startup doesn't pay for compute for its own AI product: that work is done by the network's participants, while the startup earns AET through the emission.
AET Emission Share
The neuronet owner receives a fixed 10% of the neuronet's emission, independent of any outside sales
Passive income from participationAPI Access to the Network
Run an AI service on top of the neuronet and sell access to it in fiat or crypto, while the miners do the actual compute
Monetization without infrastructure costsPower Rental
Rent out the neuronet's capacity to outside customers for their inference or model training
Flexible business modelWhy Choose AETRON
Zero Infrastructure Costs
No own GPUs or servers needed - the resources are provided by the network's miners.
Earn Tokens
For every confirmed contribution to the network, the owner receives their share of the AET emission.
The network is operational
Miners are already live on the network. Create a neuronet and connect them to your task.
How Neuronet Economy Works
Emissions are distributed by confirmed work. Inside each neuronet the share is split between three roles:
Neuronet owner
Creates the neuronet, defines task parameters, and grows the downstream product
Rewards · 10% of the neuronet's emission for creating and maintaining it
Stakers
Delegate AET into a neuronet: the stake sets the throughput and acts as a vote on the service's usefulness
Rewards · A share of the emission for backing the neuronet with their stake
Miners
Serve requests on their GPUs and verify each other through peer-replay
Rewards · 80% of the emission for serving requests and cross-checking each other
Use Cases
A few typical scenarios for AI startups
LLM Services
Chatbots, assistants, text generation
Example: A ChatGPT-class service without your own server fleet
Content Generation
Images, video, audio
Example: An image or video generator at the level of MidJourney or Sora
AI Agents and Automation
Autonomous agents for business processes
Example: Autonomous assistants that run tasks inside enterprise systems
Model Training and Fine-tuning
Retraining for specific tasks
Example: Adapting open models to a customer's industry-specific data
Technical Architecture
Quality metrics
A miner is scored on confirmed work, the Pulse benchmark result, uptime, and the share of re-checks they passed.
Network stability
Stake defines a neuronet's capacity, so in-demand neuronets attract more compute through the market itself. New miners enter at a reduced reputation weight and raise it by passing checks.
How to Launch Your Neuronet
Six steps from an empty wallet to a working neuronet. There are two ways to get there:
AETRON builder
Assemble the neuronet in a visual wizard: pick a model from the registry, the task type, the allowed hardware, and the economic parameters. No code required. The right path if your technical skills are light or you just want to ship fast.
Your own code
Define the task, the re-check rules, and the routing logic yourself. Full control over every step - for teams that need non-standard economics or specific models.
Setup AETRON Wallet
Download the AETRON wallet client. Network fees and neuronet registration on the chain both go through it.
Requirements · Wallet client, secure your keys
Register Your Neuronet
Register the neuronet on the AETRON chain by staking AET to activate it. Set the specialization and the task parameters. The neuronet owner receives 10% of its emission.
Requirements · AET for the activation stake, neuronet configuration
Choose the model
Pick a model from the AETRON registry (LLM, diffusion, audio, computer vision) or import your own from Hugging Face. Miners connect to the neuronet through the single AETRON client and download the weights automatically - there is no separate miner software to write.
Requirements · A model from the registry, Hugging Face, or your own
Configure the re-check rules
In the builder, you just pick a ready-made profile (standard, strict, custom). If you go the code route, you describe yourself which signals are compared in peer-replay, what variance is allowed, and how the miner's final score is computed.
Requirements · A ready-made builder profile, or your own spec
Deploy Your AI Model
Describe how requests are distributed across miners, what SLA you require, and the rules for splitting the emission inside the neuronet. In the builder this is a set of sliders; in code, a separate config.
Requirements · AI model, SDK/API integration
Launch & Earn
Once miners running the AETRON client connect to the neuronet and stakers provide the throughput, start sending requests. Miners serve them, and the owner earns their share of the emission from a live downstream service.
Requirements · Active participants, monitoring
Important Details
You Define the Rules
Whether you use the builder or your own code, you still decide which tasks miners perform and which signals they use to check each other. Quality metrics, re-check thresholds, and the logic for splitting emissions are all tuned to your specific AI scenario.
Flexible Miner Architecture
A miner can be the owner of a single GPU or an entire data center bundled into one node. Large providers run clusters that look like a single miner to the network; smaller participants connect alongside them. Together they make up a living compute market.
Create your own neuronet
Assemble a Neuronet in the Constructor without writing a line of code: you pick the model, the tasks and how you want to earn, and everything that touches the chain is handled for you. The network provides the compute, so you need no server fleet of your own.