Collateral Selection is Risk Management

In the aftermath of recent events, I want to take a moment to reflect on the decisions that led to this moment. Events like these are painful, but they are also the greatest teachers. If we don’t analyze the "why," we are doomed to repeat the "how."

First, I want to say that DeFi is not broken; the incentives are. While brands can serve as a partial proxy for safety, they should not be relied on as a substitute for risk management. There are competitive pressures for everyone to risk-on to stay competitive in yield. This pressure exists across the market.

The point of this post is not to assign blame. I think there are a number of issues here we need to learn from as an industry and improve. Some improvements are the use of independent rating agencies, a resolution framework, and internal junior capital to hold risk decisions accountable.

Why? Because risk and revenue are naturally at odds.

In traditional finance, it is not normal for those trying to optimize revenue to also be the ones underwriting risk. It creates all kinds of perverse incentives. There needs to be a genuine separation of these two roles to prevent the disasters we keep seeing.

Given this, our expectation was always that DeFi would eventually mature to institutional standards, so we intentionally minimized our risk-on behavior, even at the expense of protocol revenue.

It’s a trade-off we’ve made since Day 1: Spark (and Sky) prioritizes longevity over short-term TVL growth.

A good example of this is SparkLend, which positions itself as a more conservative lending market. Pooled lending works quite well when the collateral set is restricted to a highly scalable, minimal set. Complexity is the enemy of security.

This extends beyond collateral selection to core infrastructure. Price formation, for example, should not rely on a single source of truth. Redundant oracle design reduces dependency on any one input and limits the risk of cascading failures.

The goal of SparkLend is to address the common use case of borrowing stablecoins (USDT, USDC, USDS, etc.) with cross-margin volatile crypto collateral (BTC, ETH, LSTs, LRTs). This is basically the CDP model from Maker, but with some additional features such as cross-margin.

Collateral rehypothecation (or “looping”) on SparkLend is limited to ETH and Lido wstETH, reflecting a deliberate constraint on recursive risk rather than yield optimization. When you broaden the "looping" gates too far, you introduce hidden correlations and liquidity dependencies that increase systemic fragility.

The same applies to capital velocity. Without constraints on how quickly capital can enter or exit positions, systems become more vulnerable to large, coordinated moves. Rate limits on supply and borrow are a simple but effective way to reduce this risk.

The decision was made early on to isolate “looping” in markets outside SparkLend to avoid such cascading liquidity risks. Isolated markets offer several benefits, including the ability to price risk correctly and unwind the position when the opportunity ends.

Institutional-grade risk management is what Spark is all about. We understand the infrastructure layer deeply, and aim to use the best tools available rather than trying to fit a square peg in a round hole, while continuing to refine and improve this approach over time.

The industry is growing up. It’s time our risk frameworks did the same. I hope this event serves as a wake-up call for the industry. If risk and revenue remain tightly coupled, these outcomes won’t change; they'll just scale.