Perp DEX Mid-Season Battle: Fallen Players, Self-Recoverers, and Newcomers

Author: Zhou, ChainCatcher

Last week, Hyperliquid's trading volume reached approximately $15 billion, with commodity-related contracts such as crude oil, gold, and silver being the main drivers.

With oil prices fluctuating wildly, the daily trading volume of crude oil perpetual contracts on Hyperliquid exceeded $2.2 billion, second only to Bitcoin.

As tensions escalate in Iran, the Strait of Hormuz becomes tense, and the CME closes over the weekend, global traders flock to a decentralized on-chain exchange seeking price discovery.

Meanwhile, GMX Labs, which once held nearly a quarter of the decentralized perpetual contract market share, is publicly recruiting a CEO, acknowledging that its early founder-driven model is no longer sustainable and seeking to transition to a traditional leadership structure.

One is taking over the spillover demand from traditional finance, while the other is still rebuilding its foundation.

Why did GMX and dYdX fail?

A closer look at GMX Labs' announcement reveals that the CEO candidate pool spans DeFi, CeFi, traditional finance, and technology industries, with a base salary of $150,000 to $200,000 paid in stablecoins, and performance directly linked to protocol fee growth. The proposal passed the DAO governance vote with 96.42% of the votes in favor.

A decentralized protocol, with overwhelming community consensus, has decided to introduce a professional manager in the traditional sense. This means that the community has realized that the original makeshift management model is no longer sustainable, and the solution they can think of is to move closer to traditional corporate management.

dYdX's situation was even more dire. At the beginning of 2023, dYdX held a 73% share of the decentralized perpetual contract market, almost dominating it; by the end of 2024, this number had fallen to single digits, and its token price had plummeted by more than 90%.

Today, the news about these two protocols in the media isn't about product updates or market share, but rather token buybacks. When a protocol focuses its main efforts on maintaining token value rather than gaining market share, its strategic focus has fundamentally shifted.

The decline of GMX and dYdX is due to a complex set of reasons.

First, there's the issue of the starting point. An OKX Ventures report shows that in 2021, dYdX boosted its daily trading volume to approximately $9 billion through transaction mining, briefly surpassing Coinbase. This figure was inflated by token incentives; users were artificially inflating trading volume for rewards, rather than engaging in genuine transactions.

The more serious consequence is not the falsified data itself, but that the team treated the false user feedback as genuine product signals, causing the iteration direction to go astray from the very beginning.

Secondly, there's the issue of architecture. GMX uses a multi-asset liquidity pool plus oracle pricing model. This design had its merits in 2021, when order books couldn't run on the Ethereum chain, making the AMM model a viable option.

However, this architecture has a quantifiable ceiling. The total open interest that the protocol can support is about 5 times that of TVL. The upper limit of TVL locks in the upper limit of the trading volume.

In this model, LPs are naturally at an informational disadvantage, acting as the collective counterparty of all traders, but they lack the ability to proactively manage risk. Professional market makers are unwilling to enter the market under these conditions, so liquidity depth is always limited.

dYdX saw the potential of order books and decided to migrate to its self-built application chain on Cosmos. While the technical judgment was correct, execution ran into problems. After the migration, users needed to re-adapt to new wallets and cross-chain asset bridging, significantly increasing transaction costs. More importantly, the protocol fees in version 4 went to validators instead of token holders, rendering the community's perceived growth benefits negligible.

The third point lies in judging the decisive factors. GMX bets on the liquidity model, and dYdX bets on self-built chains, but the real decisive factors in this field are only two: performance and the density of the market maker ecosystem.

OKX Ventures points out that most perpetual DEXs merely shift the risks of centralization from the custodian layer to the less visible execution and liquidation layers, treating decentralization as a narrative rather than a real product problem to be solved.

dYdX's shift to synthetic stock perpetual contracts and its opening to US users is a move to trade compliance for survival, circumventing direct competition. GMX's hiring of a CEO is an attempt to compensate for strategic misjudgments through organizational upgrades. These are all correct self-rescue actions, but they are still dealing with results, not causes.

The Logic of Latercomers

When Hyperliquid launched in 2023, GMX and dYdX were still the dominant players in the market. It had no funding, no VC backing, and no major launch events.

Early growth was slow. Without token incentives to inflate trading volume, the number of traders and market makers accumulated during the cold start period was limited, resulting in consistently poor platform data. HLP Vault's profits and losses were verifiable on-chain in real time, attracting those willing to deposit real money, but at the time, this wasn't a significant advantage.

In terms of technical approach, founder Jeff chose from the outset to build his own L1 and create a fully on-chain order book. The underlying logic is that a completely transparent on-chain environment allows market makers to identify different types of transaction flows and thus adjust their pricing strategies accordingly.

This approach dictates that it cannot follow the path of migrating application chains like dYdX, nor can it follow the path of relying on oracle pricing like GMX; it must be built from the ground up. Although this theory is still controversial within the industry, it provides a clear main thread for Hyperliquid's product direction.

In terms of traditional asset allocation, HIP-3 will be launched in October 2025, first using crypto assets to build a market maker ecosystem, and then successively adding gold, silver, and crude oil.

The report points out that when dYdX launched its permissionless traditional asset market in 2024, the daily trading volume of Tesla synthetic stock was $4,000, and that of the Turkish lira was zero. Without market makers present, assets were listed with zero volume.

Hyperliquid's approach was to expand its asset classes only after the market maker ecosystem matured, so when the Iranian crisis broke out, it absorbed the wave of trading volume.

According to CoinGecko data, based on 24-hour open interest as of March 26, Hyperliquid accounted for approximately 54% of the top ten perpetual DEXs, while Aster ranked second with approximately 15%. Hyperliquid's size still exceeds the combined size of the other nine.

Aster, ranked second, and Hyperliquid entered the market around the same time. Why did Hyperliquid eventually surpass Aster?

In an interview, Aster CEO Leonard stated, "When dYdX emerged, we started trying to build our own things on-chain, and the first version of Aster appeared—Apollo X. Since then, perpetual contract DEXs have gone through several cycles, with projects like GMX representing an era. We've always tried to build what the market truly needs, and that's how Aster came about."

His words also reveal that Aster's approach was gradual. Starting with the AMM model, they iterated step by step, adding an order book, and then addressing the limitations of a transparent market by implementing privacy order functionality. Each step responded to market feedback, and each step was a reasonable product decision.

Simply put, it always follows the evolution of the track, rather than defining the evolution of the track.

Don't release your product too early.

In the crypto industry, technological paradigm shifts occur so rapidly that incremental iteration means you're always chasing the winning edge of the previous era.

People are constantly searching for answers in this field, and they still are.

The crypto industry is currently viewed unfavorably, with a significant outflow of talent and capital. However, precisely because people are leaving, the technological window won't be quickly filled, giving builders more time. Each infrastructure iteration—L2 maturity, application chain feasibility, and on-chain order book functionality—opens up new possibilities for product development.

First-mover advantage is far more fragile in this industry than in traditional industries. This presents both a risk for established players and a real opportunity for newcomers. Especially in an era where AI tools are leveling the gap in productivity, homogeneous competition is intensifying, making it increasingly difficult for just-right products to gain a foothold.

When summarizing the lessons learned from the past year of entrepreneurship, the founder of Particle quoted a quote from Google founder Sergey Brin at Stanford: "Don't release your product too early." He meant that once you release signals too early, you're tied to a delivery timeline, leaving no time to actually get things done.

Therefore, the real issue in starting a business is not how fast you can run, but figuring out where the final form of this market is.

Conclusion

GMX hiring a CEO may seem like a small matter, but it might be looked back upon as a significant event at some point.

The entrepreneurial boom of the first generation of perpetual DEXs has ended. The era of makeshift teams, founder-driven growth, and rapid iteration has reached a point where professional management is needed.

New opportunities lie elsewhere. Just as Hyperliquid capitalized on the geopolitical climate with its commodity contracts, decentralized exchanges are moving from internal competition within the crypto industry to becoming a true alternative to traditional financial infrastructure. This trend is only just beginning.