Roadmap
Every idle GPU on Earth, one click away.
AI is the most compute-hungry technology ever built, and most of the world's GPUs sit idle most of the day. Kiln turns that idle capacity into an open, global cloud that anyone can rent and anyone can supply, where every minute is proven and every rental burns $KILN.
- 01 · BuiltIgnition
- 02 · NowLaunch
- 03 · NextTrust
- 04 · ThenScale
- 05 · VisionThe open GPU grid
The flywheel
Usage feeds the network. The network feeds the token.
Kiln is built so that real usage, people paying for GPU time, drives everything else. Each turn of the wheel makes the next one easier.
- 01
More providers
Idle GPUs come online because they earn.
- 02
Lower prices
Competition pushes the cost of compute down.
- 03
More renters
Cheap, proven compute brings in demand.
- 04
More fees
Every rented hour pays the protocol.
- 05
More $KILN burned
Fees buy back and burn $KILN.
…and back to the start: a busier network attracts more providers.
Five phases
From marketplace to global grid.
- Phase 01Built
Ignition
The core works: rent a GPU, prove every minute, pay only for what ran.
- Escrow, leases and settlement on SolanaDeposits held by the program, providers paid per proven second, refunds automatic.
- Proof of rentalSigned heartbeats every minute, only while the renter's workload runs.
- Provider nodeTurns any NVIDIA machine into a cloud node with wallet-signed access.
- One-click workloadsJupyter, SSH, LLM APIs, or any Docker image.
- Dollar prices, pay in $KILNEvery GPU priced in USD; pay in USDC or $KILN in one transaction.
- Live network explorerEvery listing and lease readable on the site and on Solana.
- Phase 02You are here
Launch
Kiln goes live on mainnet and the burn starts.
- $KILN on pump.funFair launch with a public bonding curve. No presale.
- Security auditProgram and provider node reviewed externally, report published.
- Mainnet deploymentProgram upgrades controlled by a multisig, never a single key.
- Buyback & burn engineGPU rental fees and token trading fees buy $KILN and burn it, with every burn public.
- Provider downloadsOne-file installer for Linux. Plug in a GPU, start earning.
- Top-upsExtend a running lease in one click without restarting anything.
- Phase 03Next
Trust
Proof so strong that a renter never has to trust a provider at all.
- Renter attestationsYour workload co-signs every heartbeat, so uptime can't be faked.
- GPU challengesRandom verifiable computations catch fake or underpowered hardware.
- Provider stakingProviders stake $KILN to list. Proven cheating costs them the stake.
- Reputation scoresUptime, reliability and history for every provider, computed from on-chain data.
- Phase 04Then
Scale
From single cards to whole data centres, and from hobbyists to production teams.
- Clusters and multi-GPURent 8× H100 nodes and multi-machine clusters under one lease.
- Persistent storageVolumes and checkpoints that survive between leases.
- Template marketplaceOne-click Stable Diffusion, fine-tuning, rendering and more, published by the community.
- Developer API and CLIRent, deploy and scale GPUs from code or a terminal.
- Pay with anythingSOL and any major Solana token, converted at checkout.
- Phase 05Vision
The open GPU grid
Every idle GPU on Earth, one API call away.
- Serverless inferenceCall a model endpoint; Kiln routes it to the best proven GPU in the world, priced per second.
- AI agents as customersAutonomous agents with their own wallets rent compute for themselves, around the clock, with no humans involved.
- Distributed trainingTraining jobs split across many providers, with every shard proven and paid.
- Community governance$KILN holders steer fees, burn rate and network parameters.
Be early to the grid.
Put your GPU to work, rent compute by the hour, or follow along as Kiln goes live.