On-chain perpetual futures are expanding quickly, and open interest has reached an all-time high.
That is the central finding in a market study by Castle Labs, translated by TechFlow, which compares the design choices, go-to-market strategies, and market microstructure of perpetual trading venues. The report says perpetual futures now account for more than 82% of total crypto trading volume over the past 30 days, and on most major venues, perpetual volume is 4.7 times larger than spot volume.
Perpetuals remain the dominant crypto trading product
Castle Labs says perpetuals became one of crypto’s most important products because they let users maintain leveraged price exposure without holding the underlying asset.
For liquid assets such as BTC and ETH, that model has already proved powerful. As more asset classes begin trading on-chain, the value of perpetuals has widened with them.
Trading is still concentrated on centralized exchanges. Over the past 30 days, centralized venues processed more than $4.56 trillion in perpetual volume, while decentralized exchanges handled $614.5 billion. The report says the gap is even wider than in spot markets because DEX market structure still leans more heavily toward spot trading.
Early market structure was shaped by dYdX and GMX
The report frames the early on-chain perpetual market around two design paths, represented by dYdX and GMX.
dYdX, built on the Cosmos SDK, was one of the earliest protocols to see the value of vertical integration and move toward an appchain built around a central limit order book. It migrated from StarkEx to its own Cosmos appchain. Strategically, Castle Labs says that gave the team more control, more decentralization, and infrastructure it could manage itself.
For traders, though, the move added friction. The report points to bridging, more complex fund management, and a harder onboarding process. Cross-chain migration may look routine now, but it was not at the time. The transition also came with technical issues, including outages and validator delays. Decentralized governance slowed product iteration and led to decisions that hurt community trust, including shutting down the token bridge while token holders still had liquidity and assets in place.
GMX V1 took a different route with the GLP model, a dynamically balanced pool of assets such as ETH, BTC, and stablecoins that acted as the counterparty to all trades on the platform.
Castle Labs says GLP offered a workable answer to the early liquidity cold-start problem, but not a clean long-term market structure. The model concentrated assets in one pool, made liquidity providers the counterparty to traders, and paid them fees in return. The tradeoff was that liquidity providers absorbed trader PnL, asset exposure, and inventory imbalance risk.
In the report’s view, GLP was hard to scale and bundled too much risk into one product. Every new market added risk to the same liquidity pool, so open interest caps, fee design, oracle structure, and pool composition all limited long-tail expansion. Castle Labs says GLP was never designed for long-tail assets and worked better for majors and large-cap tokens. Weak capital efficiency also constrained demand as competition increased.
The report adds that GLP V1 suffered a $40 million hack in 2025 due to a reentrancy vulnerability, which further damaged depositor confidence. GMX later moved away from GLP and shifted to V2 and the GLV model to improve flexibility and capital efficiency. Castle Labs says asset-pool-based designs are inherently less flexible than order book models when it comes to launching new markets and scaling horizontally.
That said, the approach did not disappear. Hyperliquid uses Hyperliquidity Provider, or HLP, while Variational uses Omni Liquidity Provider, or OLP, both of which adapt parts of the model.
DEX share is rising, but centralized venues still dominate
The report says the addressable market for perpetuals has always been large, but early models could not scale liquidity well enough to compete with centralized exchanges. Most users therefore stayed on centralized venues.
That has not fully changed, but DEX share in perpetual trading has risen from 4% in 2024 to above 13%. The high point came in December 2025, when decentralized perpetual venues briefly accounted for more than 14.25% of total perpetual volume.
Castle Labs says the top-line market share view only tells part of the story. Under the surface, the sector kept growing, especially in 2025. The report says volume began climbing in late 2024, lining up with the launch and success of Hyperliquid.
The latest leg of growth in 2026 came from several factors at once. One was the rise of real-world assets, which could now be traded through perpetuals. Another was the arrival of more venues, including Lighter, Variational, Extended, and Entropy, along with others mentioned in the report. Castle Labs says more than 261 decentralized perpetual exchanges have launched since Hyperliquid appeared.
Hyperliquid holds more than 56% of open interest
At present, Hyperliquid remains far ahead of the decentralized perpetual market, with more than 56% of total open interest. Aster follows with 9.7%, Variational with 6.2%, and Lighter with 5.7%.
The report says the market was first led by dYdX and then by GMX. Since Hyperliquid launched, it has held the lead almost continuously, except for a brief stretch in November 2025 when Lighter temporarily took the top spot.
Castle Labs also notes that monthly perpetual volume reached its historical peak in late 2025, then cooled somewhat in 2026.
RWA perpetuals are emerging as a new growth lever
The shift in market leadership reflects a broader change in the sector, especially around listing and supporting real-world assets, according to the report.
Castle Labs says one of Hyperliquid’s clearest differentiators, and one reason it drew attention beyond crypto-native circles, was the role of Tradexyz in bringing new assets on-chain. Tradexyz made assets that were previously unavailable on-chain tradable around the clock. At its peak, that segment accounted for 50% of the platform’s total trading volume.
The report highlights crude oil, or CL, as a case study. During the Iran crisis, the U.S. attacked Iran over a weekend while traditional venues were closed. Tradexyz became the only place where crude could be traded. Castle Labs says it produced a full market microstructure analysis comparing Tradexyz with the CL market on CME. Within three weeks, CL evolved from a novel market into a mature trading venue where traditional finance participants could hedge exposure.
For Castle Labs, that was one of the clearest demonstrations of the utility of on-chain RWA trading and its ability to drive adoption. The report says funds such as Abraxas have since become committed users of RWA trading on Tradexyz.
Part of Tradexyz’s success came from a boundary-discovery mechanism that allowed RWA assets to trade 24/7. During off-hours, price movement was constrained within a reference band of ±(1 / max leverage). In the first weeks of testing, the CL market hit the 5% boundary and trading stopped over the weekend. The report says the method has since been battle-tested, with Tradexyz using re-anchoring to maintain continuous price discovery. Once an asset approaches a trigger condition, such as nearing the boundary, the market can be re-anchored and a new band can be set from that price.
RWA products are not only opening previously inaccessible assets to retail traders. They are also offering an alternative venue for traditional financial institutions. Castle Labs says that as of Sept. 10, 2026, seven of the top 10 perpetual assets by trading volume were RWAs. The Nasdaq 100 and gold led the group, combining for more than $1.4 billion in 24-hour volume and more than $850 million in open interest.
That shift has changed the role of perpetuals, the report argues. Instead of being mainly a place to gain leveraged exposure to base crypto assets, they are becoming a venue for price exposure to assets that were previously unavailable.
How Hyperliquid changed the category
Castle Labs describes Hyperliquid as one of the main catalysts for the current phase of on-chain perpetual growth. Its airdrop-style token generation event, combined with its move from a perpetual exchange into a broader ecosystem aiming to become the “AWS of liquidity,” reignited demand for building these primitives and created a model others are now trying to follow.
The report says the key point is not just Hyperliquid’s volume growth. It changed how traders think about on-chain perpetual venues. Before Hyperliquid, most perpetual DEXs were seen as on-chain versions of centralized perpetual exchanges: lower custody risk, but weaker execution, worse user experience, and thinner liquidity. After Hyperliquid, the category started to look more like a full trading platform.
Castle Labs lists several modules that now sit inside that stack:
- HyperCore, a high-performance trading layer for spot and perpetual markets.
- HyperEVM, a native builder environment around the trading venue.
- HIP-3, which lets builders deploy new perpetual markets and pushes listings toward a permissionless model.
- HIP-4, which introduces outcome markets for prediction products and option-like tools.
- Builder codes, which let external front ends route order flow to Hyperliquid and monetize distribution.
- HLP, Hyperliquid’s liquidity vault for passive market-making capital.
- RWA front ends such as TradeXYZ, which use Hyperliquid infrastructure to bring stocks and other off-chain exposures into the same trading environment.
The report says each of these pieces contributed to Hyperliquid’s success, and together they put HYPE at the center of multiple growth vectors. HyperCore delivers the core exchange experience with order book trading, fast execution, deep liquidity, and a clean interface. HLP provides a native liquidity primitive. HyperEVM gives builders an ecosystem to work in. HIP-3 turns market creation into both a product and a revenue source. HIP-4 adds outcome markets. Builder codes turn wallets and apps into distribution partners. TradeXYZ shows how the same stack can bring traditional financial assets into one interface.
Those modules also allow users to keep portfolio margin across the platform and run more complex strategies, including spot and perpetual exposure across both crypto and RWA assets, liquidity provision, and event hedging. The report adds that last week Hyperliquid launched manual lending, allowing anyone to supply HYPE or BTC to borrow USDC and USDT.
From airdrop growth to a revenue flywheel
In its early growth phase, Hyperliquid followed a familiar playbook for perpetual venues, using points campaigns and a large airdrop to reward users with direct economic interest and build an initial base of loyal traders.
Castle Labs says the token generation event was only the start. What mattered more was the alignment across the protocol and the broader ecosystem. The report points to more than $1.24 billion in cumulative revenue as evidence.
It also pushes back on the idea that all of that revenue came from perpetuals. Spot markets, auctions, priority burns, and HyperEVM gas fees all contributed, while HYPE remained central to the growth model.
HLP played a strategic role in the early cold start of HyperCore. Users could deposit funds into the market-making vault and run passive strategies. That helped seed liquidity in the early days and created another retention loop for users who did not want to trade actively. It also meant Hyperliquid did not need to rely on outside market makers or external capital.
The report says HLP differs from GMX’s GLP because it is more active than a static pooled liquidity basket and reinforces Hyperliquid’s identity as a front end where users can run different strategies. Over time, HyperCore and market-making activity grew enough that HLP was no longer foundational and was gradually deprioritized. Still, the report says HLP appears to have picked up again after mid-August 2026, with plans to move idle HLP USDC into lending so capital does not sit unused.
There were challenges as well. Castle Labs points to the JELLY incident in March 2025, when a trader tried to squeeze the Hyperliquid market by pushing up the spot token price while holding both short and long positions in the asset. The report says that episode is discussed in more detail later in the full study.
Builder codes, HIP-3, and HIP-4 widened distribution
In HIP-3 markets, Hyperliquid is no longer the only team deciding which markets matter. Builders can buy trading codes through an auction system and deploy perpetual markets within the protocol framework, with fees flowing back to the token.
Builder codes also mean Hyperliquid does not have to focus only on direct user acquisition. It can expand horizontally by offering its trading technology as a white-label solution, letting others launch perpetual trading without building an engine from scratch.
Castle Labs says builder codes have been highly successful and show how much emphasis Hyperliquid places on backend infrastructure. Over the past 24 hours, they contributed more than $40 billion in trading volume, or 7.4% of the total. Wallets such as Phantom and MetaMask did not build perpetual trading infrastructure themselves. Instead, they integrated builder codes, generating more than $25 million in revenue for Phantom and more than $10.5 million for MetaMask.
More recently, Hyperliquid announced HIP-4 markets, its binary outcome product, which helps the platform expand into event markets, prediction markets, and options-related use cases.
The report says perpetual venues now have to move and iterate faster than before. Shipping features is not the hard part. User acquisition is getting harder because the pool of active traders is limited and increasingly mercenary. A venue needs distribution, liquidity, product breadth, and a reason for traders to keep coming back.
Hyperliquid’s answer, according to Castle Labs, is to make the platform useful from several angles at once: trading on the core venue, building on the EVM layer, importing flow through builder integrations, deploying markets through HIP-3, trading outcomes through HIP-4, or bringing in new asset classes through external front ends.
HYPE sits at the center of the alignment model
One less obvious but important point in the report is the alignment between platform success and the HYPE token. That shows up in the protocol’s programmatic buyback structure: 95% of all fees collected by the protocol are used to buy back HYPE.
So far, the protocol has bought more than $1.26 billion worth of HYPE on the open market. Whether through HIP-3, builder codes, or HIP-4, Hyperliquid has stayed focused on expanding distribution.
Castle Labs says an exchange can generate activity before a token launch, but it is much harder to sustain that activity after obvious airdrop farming ends and mercenary capital leaves. For most exchanges, activity peaks before the airdrop because the product’s value is still unproven. Hyperliquid, in the report’s view, shows how token alignment, execution, and distribution can create a deeper flywheel that carries the story well beyond the token launch.
Even so, the report says many protocols are trying to copy the model, and it remains unclear whether they can match Hyperliquid’s trajectory.
The sector is still early, but the growth is hard to ignore
In its conclusion, Castle Labs says the on-chain perpetual ecosystem has already gone through a major transition. Much of the activity has shifted away from earlier leaders such as dYdX and GMX toward Hyperliquid, Variational, and Lighter. That transition brought different architectural choices and introduced a new growth lever: RWA perpetual trading.
Platforms such as TradeXYZ are not only bringing traditional financial assets on-chain. They are also enabling 24/7 trading and leading active price discovery when traditional markets are closed, as seen earlier this year during the Iran attack period cited in the report.
Castle Labs says much of the sector’s recent growth can be traced to Hyperliquid, which controls more than 56% of on-chain open interest and has reshaped the perpetual market by turning venue usage into broader token alignment and a platform flywheel.
The report closes by saying on-chain perpetuals are still in an early stage, but their growth is becoming difficult to dismiss. Hyperliquid has shown that decentralized venues can compete with centralized exchanges on nearly every metric. At the same time, on-chain venues are increasingly focused on distribution, whether through consumer-facing apps or partnerships, and that push is continuing.
Castle Labs adds that it has also covered more protocols and design approaches in the perpetual sector and produced a separate market microstructure analysis comparing execution quality on CEXs and DEXs. The full report is available separately.
