USDT Strategy

Spark Savings USDT (spUSDT) is a non-custodial, permissionless stablecoin savings vault backed by USDS.

We surface the USDT liquidity strategy in particular because of the more restrictive liquidity profile in comparison to USDC, as well as highlight how well it performed during the rsETH incident of April 18th, 2026.

This document sets out Spark's standing framework for idle liquidity management, how we approach venue selection, rate-setting, and liquidity buffers across market conditions, with the central goal of deploying capital above the Minimum Liquidity Requirement (MLR) efficiently without compromising SparkLend borrower liquidity or depositor redemption capacity.

Core Principle: Tiered Deployment, Automated Liquidity Defense

Spark operates a tiered deployment model. Capital is allocated in priority order from most liquid to least liquid, with an automated planner that withdraws from junior venues first when liquidity tightens. The goal is to keep SparkLend fully liquid at all times while earning a return on every USDT above the MLR buffer.

• SparkLend (Primary Venue). SparkLend is the default and highest-priority deployment venue. USDT lent here is available on demand and earns the benchmark rate. Spark operates the largest on-chain USDT lending market by size, with cross-collateral capacity no other venue matches. Preserving SparkLend liquidity is non-negotiable - it underpins both depositor safety and borrower UX.

• Morpho Blue Chip Vault (Overflow Deployment). Capital beyond SparkLend's absorption capacity is deployed into the Spark Morpho vault across blue-chip collateral markets (WBTC, cbBTC, sUSDS, wstETH).

• Duration Products (Capped, Risk-Adjusted). A limited allocation to fixed-income, duration-bearing products (currently Maple Finance syrupUSDT) captures the yield premium available at the short end of the curve. This tier is intentionally capped. Duration allocation is sized against: (i) redemption lag relative to MLR stress scenarios, (ii) concentration of this allocation as a share of the total product, and (iii) the withdrawal cadence of large depositors. We do not increase this tier unless liquidity buffers and depositor composition clearly support it.

• DEX Liquidity & Stablecoin Inventory Management. A portion of idle USDT is deployed into DEX liquidity pools as a capital-efficient alternative to holding cash. Spark currently maintains a USDT/sUSDS pool on Curve, which serves a dual purpose: it earns fees on stablecoin flows while acting as an on-chain inventory mechanism, allowing the planner to swap between USDT and USDC/sUSDS programmatically as inventory needs shift. Spark will introduce a USDT pool on Uniswap v4, which provides meaningfully stronger price-making and market-making capability relative to Curve's stable AMM design, enabling tighter spreads and more efficient inventory rebalancing at scale.

USDT converted to USDC through these pools is deposited into Sky's sUSDS vault at ~3.65%, the effective risk-free rate in the Sky ecosystem, materially reducing cash drag on idle balances without taking on duration or counterparty risk beyond the Sky stack.

Rate-Setting Philosophy

Spark sets spUSDT savings rates against three inputs: P&L sustainability across the full deposit base, depositor composition, and on-chain borrow demand. Rates are adjusted to reflect what deployment yields can support. When borrow demand softens, rates move accordingly. We price for the full book, not the marginal depositor.

The automated planner enforces a rate-response feedback loop: as liquidity approaches MLR thresholds, rates rise automatically to attract deposits and incentivise borrower repayment. The system replaces excess idle buffers that are not required as insurance.

How We Respond to Market Stress

The rsETH / KelpDAO incident in April 2026, $196M+ in Aave bad debt, $13.2B DeFi TVL drawn down in 48 hours, Spark Savings USDT was the only major lending venue that remained fully liquid throughout, never breaking below $400M in idle USDT. SparkLend TVL grew +50% in the days that followed.

The incident produced several durable changes to how we think about stress management:

Conservative MLR anchoring. MLR is sized to the largest single-depositor exposure. During stress, large depositors run liquidity checks and make partial withdrawals. Spark's approach is to hold enough idle USDT in Spark Savings at all times to absorb these without triggering rate spikes that squeeze borrowers.

Duration risk is asymmetric. In stress, the venues that fail first are duration products with redemption lags. Spark sizes duration allocations to ensure that even a full Morpho exit plus partial Maple redemption leaves SparkLend borrowers unaffected. Duration allocation is sized against the downside, not the yield.

Avoid over-deployment into yield at the cost of liquidity. The implicit lesson from Aave's experience with rsETH is that reaching for yield through concentrated, leveraged, or bridge-backed collateral creates contagion pathways. Spark's collateral profile, blue-chip only, conservative LTV, rate-limited supply caps, is a structural risk differentiator, not just a positioning choice.

Case Study: Rate Management Framework

Spark's planner dynamically manages spUSDT supply rates against the liquidity position. When liquidity falls below a set threshold of MLR, the planner withdraws from Morpho. When it falls below the critical threshold of MLR, rates rise automatically to incentivize borrower repayment and new deposits. Meanwhile, the strategy can also set a benchmark rate target for both SparkLend and Morpho to maintain SparkLend's borrower depth as the primary cross-collateral venue, while using Morpho as an overflow deployment sink.

Current Deployment Snapshot (May 5, 2026)

What Changed and Why

• Spark Savings USDT was the only major lending venue that remained fully liquid. Spark held $400M+ in idle USDT throughout the crisis, never breaking below liquidity requirements. SparkLend TVL grew +50% in the days following the incident as the capital fled Aave to Spark. This validated maintaining a conservative idle buffer and not over-deploying into duration.
• On-chain borrow demand dropped sharply post-incident. Post-incident borrower demand on SparkLend/Morpho collapsed. The on-chain USDT borrow market went from $642M peak borrow (April 26) to ~$488M by early May. Morpho asset spreads were cut post-incident to maximize Morpho deployment of otherwise-idle USDT.
• SLL USDT inventory wound down entirely. The $150-250M USDT SLL inventory buffer (held ready for rapid cross-chain deployment) was wound down to zero. With borrower demand soft, holding large SLL inventory generated pure drag. The team will shift to more capital-efficient allocation strategies as indicated above.
• syrupUSDT capped; duration risk managed carefully. Maple initially wanted $100-175M at 2-week duration and return compression at scale. Spark capped at $76M.