Moonshift logo

Moonshift

Small teams rent real GPUs by the task. Owners hold GPU shares that earn from every job, and borrow against them at rates set by GPU revenue and on-chain credit, from a native XRPL vault.

Demo Video

Project Information

At a Glance

Small teams rent real GPUs by the task. Owners hold GPU shares that earn from every job, and borrow against them at rates set by GPU revenue and on-chain credit, from a native XRPL vault.

Description

Moonshift lets people own pieces of real GPUs and borrow against them.

GPUs are some of the most valuable machines in the world right now. They earn money every hour, from people renting compute time and from AI work they run. But if you own a share of one, that value is stuck. You can't borrow against it without selling.

How Moonshift works:

Own. Each GPU is split into 1,000 share tokens on the XRP Ledger. When customers book the GPU or send it AI jobs, the income flows to the shareholders.

Rent. Customers pick what matters to them at checkout: cheapest, balanced or fastest. Moonshift prices every GPU against live AWS market data (spot prices for cheap, interruptible capacity, on-demand prices for guaranteed capacity, and how often each machine type gets interrupted), then routes the job to the right GPU automatically. If a GPU is taken away mid-job, the work continues on another one instead of starting over.

Borrow. Lock your GPU shares as collateral and borrow USDC from a lending vault. Your shares keep earning while they are locked, and they come back the moment you repay.

Lend. Lenders supply the USDC that borrowers draw. They deposit into a native XRPL lending vault and earn interest that is backed by real GPU income, with Moonshift's first-loss cover sitting in front of them if a loan goes bad.

What makes the loans smart:

Most crypto lending only looks at the price of your collateral. Moonshift also looks at two things nobody else uses. First, how much the GPU actually earns, because a GPU that pays steady income can safely support a bigger loan. Second, how you have handled money on chain. We read your real wallet history and turn it into a credit score with the reasons shown via Quicknode, so a borrower with a good track record gets a better rate. Connect a wallet and you can watch your score update live as you make transactions.

Every loan decision comes with a short written memo that explains the terms in plain English, written by AI but checked so it can never invent a number.

Staying safe:

Every loan has a margin call line and a liquidation line. A watcher checks every loan against the GPU share price. If the price drops too far, the borrower gets a margin call, and if they don't fix it, the collateral is claimed for the lenders. Moonshift also puts up first-loss cover in the vault, so lenders are protected before anyone else.

Why the XRP Ledger:

The whole money path uses features built into the ledger itself: lending vaults, native loans, escrow, atomic batch transactions and price oracles. There are no custom smart contracts holding user funds, and settlement is fast and cheap.

What is live in this build:

Vault deposits, loans, payments, margin calls, liquidations and collateral release are real transactions on XRPL devnet, because the new lending features are not on mainnet yet. Credit memos are written live by Amazon Bedrock, and routing prices come from live AWS EC2 spot and on-demand pricing.

Technical Details

Lending runs on XRPL devnet with native XLS-65 vaults and XLS-66 loans: lenders fund the vault, Moonshift is the loan broker with first-loss cover and signs LoanSet, the borrower countersigns. GPU shares are issued tokens locked as collateral with token escrow; repayment releases them in one atomic Batch, default seizes them via LoanManage. An on-ledger price oracle and a keeper drive margin calls and liquidations. Credit scores use a points scorecard over account history read through Quicknode (mainnet and testnet), rescored live when a connected Crossmark wallet transacts. The order router solves cheapest, balanced (Pareto knee) or fastest using live AWS EC2 spot and on-demand prices and spot interruption rates, and resumes interrupted jobs on another GPU. Amazon Bedrock (Claude Haiku) writes each credit memo, guarded against invented numbers. Stack: Next.js, Express, xrpl.js, Python FastAPI.

Team

4
AS

Alexander Schuessler

KB

Kristian Bridge

TE

tharun Ekambaram

LZ

Lucy Zhang

Hackathon

AI & Fintech XRP Ledger Hackathon

Duration

Oct 24, 8:30 AM - Oct 25, 8:00 PM EDT

View Hackathon Details

Prizes Won

1

Prizes Won

1 prize awarded

Quicknode 1

Sponsored by Quicknode
$1,500
USD
Total Prize Money
$1,500