DeFi's $13B Primitive Lands on Stellar: Building Hatom Yield Vaults on Soroban

Yield vaults are the most validated primitive in DeFi. The pattern Yearn pioneered in 2020 became a standard, ERC-4626, and that standard now holds more than $13B in TVL across chains. In 2026 it also became the institutional on-ramp: asset managers curate vaults, exchanges embed them as earn products, and stablecoin yield ships wrapped in vault shares.

Stellar had no production yield vault and until recently no vault standard. So when Hatom, the DeFi protocol that pulled $50M in TVL within an hour of launching on MultiversX, decided to expand to Stellar, they brought us in as their engineering partner to build it with them: the Stellar Capital Efficiency Layer (Hatom Vaults, to users), a risk-tiered yield vault protocol on Soroban, Stellar's smart contract platform. It is live on testnet today, targeting mainnet in Q4 2026.

Why Stellar, and why now

Stellar spent a decade as a payments rail. Smart contracts only arrived in 2024 with Soroban, and the DeFi stack has been assembling since. Stellar's DeFi TVL roughly quintupled in a year, approaching $200M by April 2026 and sitting around $230M as of August 2026. Blend, the chain's core lending protocol, crossed $100M on its own.

In addition, tokenized real-world assets on Stellar passed $3B, roughly quadruple their value in 2025. That includes Franklin Templeton's BENJI, the first US-registered tokenized money market fund, which has been on Stellar for five years, and PayPal's PYUSD, live since September 2025. Most chains add real-world assets on top of an established DeFi stack. Stellar did the reverse: institutions and payment rails first, DeFi second.

That sequence created a specific gap: there is capital and yield on the chain, but no product that connects them the way vaults do everywhere else. Ecosystem research commissioned by the Stellar Development Foundation listed yield aggregation among the DeFi categories it wanted filled. And the ecosystem is converging on the answer: SEP-56, a draft tokenized vault standard co-authored by OpenZeppelin, closely follows ERC-4626. The $13B primitive is being formally ported to Stellar, and that is the window in which we are shipping.

What we built

In a nutshell: users deposit USDC and receive vault shares, each worth a proportional slice of the vault's assets. That USDC is then put to work through the strategy matching the vault's risk profile, at its own expected yield.

All the technical complexities of loops, swaps and rebalancing are managed directly by the protocol, allowing the user a simple set-and-forget interaction.

The protocol is a vault and adapter system, written in Rust on Soroban, which consists of:

• Risk-tiered vaults. Users pick their strategy based on three risk tiers, currently spread across four vaults, each an isolated contract whose losses cannot touch another vault's capital:
  • Conservative — lends USDC on Blend.
  • Balanced — adds Blend's backstop pool.
  • Aggressive — a long/short pair of vaults running leveraged XLM strategies.
• Strategy adapters. Each strategy lives in its own contract that only its vault can move funds through. Adapters can be added and removed through dynamic configuration, without migrating user positions or redeploying the vault.
• Leveraged strategies. YieldAmp loops USDC supply and borrow on Blend for 2x to 3x exposure to the lending rate. Long and Short XLM strategies build leveraged directional positions through oracle-bounded DEX swaps.
• Keeper bot. An off-chain keeper deploys buffer deposits, rebalances allocation drift, compounds rewards, and, importantly, deleverages positions when risk thresholds trip.

Soroban vs. EVM

Soroban is not EVM, and the differences are not cosmetic. Reentrancy, the exploit class behind some of DeFi's largest hacks, is disallowed at the platform level. There is no msg.sender: the caller is passed explicitly and verified with require_auth(), so an adapter can demand that only its own vault ever moves its funds and have the platform enforce it, instead of inferring identity from the call stack. Two classic mistake classes are gone before you write a line of Rust.

Another core difference between Soroban and EVM is that state has rent. Contract storage carries a time-to-live: persistent entries that nobody pays to keep alive get archived, temporary entries are deleted outright. On EVM a user's balance sits in storage forever, for free. On Soroban, keeping user accounting alive is an operational responsibility, and picking the wrong storage class for it is a way to lose user state with no attacker involved. So the storage type of every entry is a decision we make explicitly: persistent for shares and balances, bumped on every touch; instance for vault config; temporary only for data that is supposed to die.

We backed the design with roughly 1,700 automated tests and 20 end-to-end integration suites, run in-host against faithful reproductions of Blend's liquidation gate and failure modes.

Live on testnet, end to end

The protocol is running on Stellar testnet today: four vaults deployed, deposits and withdrawals flowing through the full stack, and the keeper operating around the clock. It is acting on live positions, sweeping idle deposits into strategies and correcting allocation drift, and it has already deleveraged a leveraged position the moment its risk threshold tripped. Hatom Vaults charts share price and TVL for every vault, sampled on-chain every 15 minutes.

Hatom was awarded $135,000 in SCF #41, the community-voted Stellar Community Fund round, to bring the protocol to Stellar. The SCF awarded $14.4M to 154 projects in 2025 alone and gates payouts on shipped milestones. SCF projects can also tap Stellar's Audit Bank, which provides structured security audits and follow-ups as protocols grow: the ecosystem itself invests in the security of its projects.

What's next

After a full security review, mainnet is just around the corner. If the vault thesis holds on Stellar the way it held everywhere else, the chain's $3B in real-world assets and 10M+ accounts are about to get a savings product that feels less like DeFi and more like finance.

We will publish a technical deep dive on the vault mechanics next: share accounting, allocation as a risk mandate, why a naive rebalancer will happily pour money into a failing strategy, and how the protocol splits authority so that the action which adds risk needs a key and the one that removes it does not.

Building on Stellar, or bringing an existing protocol to a new chain? That is exactly the kind of blockchain engineering we do. Talk to us.