Sovereign GPU infrastructure, unified into a single network.
Five independently controlled layers, from raw energy to production applications. Reserve accelerator capacity across 18 datacenters in 8+ countries, metered per second and settled onchain in $GPU.
0xfBaac7666A33757E23cBf46c5a51A042673e7777Sovereign compute infrastructure
Five independently controlled layers, each operated on its own terms. Consume one of them, or the whole stack, through a single interface.
APPLICATION
Your production applications
Build on top through a single API. Capacity is metered per second and settled onchain, so spend tracks usage instead of a reserved instance you forgot to shut down.
Runs the tooling you already use
Rent GPU capacity
From a network of 18 datacenters across 8+ countries. Reserve sovereign compute on the geography that matches your data residency, latency and compliance requirements.
- 90%Liquidity poolLocked at launch
- 5%Ecosystem rewardsOperators and integrations
- 3%Exchange liquidityMultisig, locked
- 2%TreasuryOperations and development
Liquidity locked
Liquidity provider tokens are locked at launch, so the pool backing the token cannot be withdrawn.
Contract renounced
Mint authority is revoked. Total supply is fixed at one billion and cannot be increased later.
1% buy and sell tax
A flat 1% on each side, and nothing beyond it. The fee funds network operations and ongoing development.
Choose the perfect GPU instance for your workload
From GeForce class accelerators up to rack-scale Vera Rubin. Flexible pricing across on-demand and reserved terms, and the hourly rate drops the longer you commit.
Hosting and renting are coming soon
The rates below are indicative. Deployment opens once capacity onboarding completes in phase 02.
NVIDIA Vera Rubin
A rack-scale supercomputer engineered for agentic AI and heavy reasoning workflows. Successor to the Blackwell architecture, built from six tightly integrated custom silicon chips and ramped into full production.
- Memory
- 288 GB HBM4
- Bandwidth
- 22 TB/s
- Scale
- NVL144 rack
- Silicon
- 6 custom chips
Billed per second, no commitment
A100
80 GB VRAM
Billed per second, no commitment
A6000 / A40
48 GB VRAM
Billed per second, no commitment
GeForce 4090
24 GB VRAM
Billed per second, no commitment
GeForce 3090
24 GB VRAM
Billed per second, no commitment
A5000 / A10
24 GB VRAM
Billed per second, no commitment
Need a larger cluster?
Multi-node reservations, dedicated racks and custom terms are arranged directly. Tell us the workload and we will size it.
Rates shown are indicative and exclude storage and egress. Availability counts reflect capacity being onboarded and will change before launch.
What legacy infrastructure charges millions for
Three phases. Traditional providers quote quarters of lead time and seven figure invoices for each one. The network delivers them as self serve infrastructure.
Network launch
Legacy procurement: 15 to 20 days, $0.5M to $1M
GYATT: under an hour, pay per second
- Token deployed
- Liquidity locked
- Contract renounced
- Public documentation
Capacity onboarding
Legacy procurement: 45 to 90 days, $1M to $3M
GYATT: self serve, no minimum commitment
- Datacenter partner onboarding
- On-demand GPU marketplace
- Usage based billing in $GPU
- Exchange listings
Protocol decentralisation
Legacy procurement: 180 to 360 days, $5M to $10M
GYATT: governed by token holders
- Validator set expansion
- Subnet marketplace
- Staking and delegation
- Onchain SLA enforcement
The settlement layer for decentralised compute.
Every subnet is a self-governing marketplace for GPUs, AI, storage or bandwidth. Usage is metered onchain, settled in $GPU, secured by validators and steered by token holders.
Spin up subnets, define the services they offer, set their economics and onboard the hardware that backs them.
Token holders vote on which subnets receive priority, how emissions are distributed and where capacity is expanded next.
Secure the chain, finalise blocks, verify that metered compute was actually delivered, and earn $GYATT emissions for it.
How to buy
Four steps. No sales call, no procurement cycle, no enterprise agreement.
- 01
Set up a wallet
Use any EVM compatible wallet, such as MetaMask or Rabby, and add Robinhood Chain as a network. Store your recovery phrase offline.
- 02
Fund it
Hold the network gas token to cover fees, plus NVDA, which is the token the GPU pair is quoted against.
- 03
Swap on flap.sh
Open flap.sh, paste the contract address shown on this page, check the slippage setting and confirm the swap in your wallet.
- 04
Verify the token
Import the contract address into your wallet so the balance displays, and always check it against the address published here.
Ready to allocate compute?
Verify the contract address, confirm the swap, start building.