Perpetual trading is where “DeFi UX” and “DeFi infrastructure” collide head‑on. Traders want tight spreads, deep liquidity, and instant execution. Builders need a system that can process constant order updates, margin calculations, funding flows, and liquidation bursts without breaking under load.
That is why derivatives teams keep choosing Starknet technology. It aligns with how perp markets actually work: high throughput, heavy computation, low-cost settlement, next-generation proving, and a real decentralization trajectory.
Why perpetuals push chains to the limit
Perps are not “just another DeFi app category.” A perp venue is a real‑time risk engine that never sleeps:
- High-frequency state changes: order placements, cancels, partial fills, position updates, PnL, funding accruals, and liquidations can all hit at once during volatility.
- Latency sensitivity: in perps, milliseconds matter. If the base layer can’t provide fast confirmations and consistent block production, the matching engine and liquidation logic become harder to run safely.
- Security requirements beyond “no hacks”: traders need credible self‑custody guarantees and correctness guarantees for the exchange logic—especially around liquidations and margin checks.
This is exactly why perps architectures often converge on one of two approaches:
- High-performance order handling (often off‑chain)
- Trust-minimized settlement and correctness (on‑chain)
Starknet’s design (and the broader Starknet stack, including appchains and StarkEx) is unusually aligned with that reality: Heavy computation can run off-chain where needed, while state transitions are proven and settled with STARK validity proofs. The result is strong security and self-custody, with settlement costs that can stay below one cent.
A derivatives track record: dYdX lit the fuse
Starknet didn’t “discover” derivatives recently. Derivatives venues have been choosing StarkWare’s validity-proving
stack for years, because perpetuals are one of the harshest stress tests for scaling: you need high throughput, predictable latency, and correctness you can’t hand-wave away.
A major inflection point came in 2021, when dYdX and StarkWare brought cross-margined perpetuals to production on a Layer 2 system powered by StarkEx. The value proposition was straightforward: scale an exchange-grade perp experience with dramatically lower costs and an architecture built for high-capacity, cross-margined trading.
dYdX was a market leader in the early DeFi perps era. In 2021, it surged to a dominant position among decentralized perpetuals, peaking at roughly 94.8% market share in Q4 as volumes across decentralized perp venues hit their cycle high. In 2022, it maintained that leadership at scale, with users trading about $466.3 billion in derivatives volume and generating $137.8 million in fee revenue over the year — numbers that put it in a different league from most onchain venues at the time.
From there, the pattern repeated. Teams building exchange-grade derivatives kept converging on the same requirements: throughput under load, provable correctness, and user-protective recovery mechanisms. StarkEx was built for that class of workload, and Starknet extends those properties into a permissionless, composable L2 environment.
Proof from the markets: three top perp venues on Starknet tech
In 2026, Starknet’s underlying technology powers three of the top 10 perpetual DEXs by volume. What makes this especially important is that they represent three different deployment models:
- Extended runs directly on Starknet mainnet
- edgeX runs on StarkEx
- Paradex runs as a Starknet appchain
Extended: perps on Starknet mainnet
Extended went live on Starknet mainnet in August 2025, marking a shift from operating on StarkEx to becoming fully embedded in Starknet’s onchain DeFi environment. Its architecture is deliberately hybrid. Extended runs a central limit order book (CLOB) where order processing, matching, position risk checks, and sequencing happen off-chain — while validation and final settlement are executed on Starknet mainnet. For builders, this design reflects how high-performance perps infrastructure works in practice:
- The matching engine remains ultra-fast and adaptable.
- Critical state transitions are validated and settled on Starknet, ensuring provable correctness and self-custody.
This separation allows Extended to deliver CEX-level performance without sacrificing the guarantees of onchain settlement.
What Starknet unlocks for Extended
Extended gets two core benefits from settling on Starknet mainnet: self-custody, with assets held in Starknet smart contracts, and onchain enforcement of trading logic, so invalid actions — including incorrect liquidations — cannot be finalized.
Mainnet also adds composability: a shared DeFi environment where liquidity, collateral strategies, vaults, portfolio tools, and secondary apps can build around the same settlement layer and work through the same wallets and accounts.
With collateralized vaults, Extended also turns collateral management into an onchain primitive: deposits, withdrawals, risk parameters, and accounting enforced at the contract level. That makes strategies easier to package and integrate across the ecosystem without introducing additional custody or validation trust.
The results are visible in the numbers. Since going live on Starknet mainnet in August 2025, Extended has shown rapid, accelerating growth. Total trading volume reached ~$180 billion in just six months, alongside TVL rising to ~$200 million and a steadily growing base of daily active traders.
edgeX: StarkEx-built orderbook perps at massive scale
edgeX shows the StarkEx path: an orderbook perp exchange built for high-performance trading, where trades settle to Ethereum through StarkEx and trading logic is validated on-chain.
The exchange follows an orderbook-based design in which StarkEx serves as the settlement and validity layer, with trades batched to Ethereum through an external committee structure.
That matters because StarkEx is built for high-throughput exchange workloads, not lightweight demo use cases. It supports features such as transaction bundling, forced withdrawals, and recovery mechanisms that are especially important for users who want self-custody guarantees even if the operator fails.
edgeX consistently ranks among the top perp venues by volume (often in the top-3), and it is among the top revenue-generating products across crypto more broadly (top-10). As of March 2026, edgeX reached over $800 billion in cumulative trading volume with about $2.5 billion traded for 24 hours.
Paradex: a Starknet appchain tuned for perps
Paradex represents the appchain branch of the Starknet perp story: a venue built on Starknet technology and optimized for high-performance derivatives trading while still relying on STARK proofs for correctness.
It also shows how quickly exchange infrastructure can improve when core upgrades reach the stack. Paradex integrated Grinta and gained concrete exchange-level improvements, including a new mempool, decentralized sequencing architecture, faster transaction flow, and broader performance gains.
For perps, that has direct consequences. Fair ordering, consistent inclusion, and lower latency shape execution quality during volatile periods and matter even more around liquidations.
S-two integration
Paradex also became one of the first major exchanges to run fully on S-two proving. That upgrade improves exchange economics in a direct way: lower proving costs, lower end-to-end latency, and more room to scale as throughput grows, while preserving STARK-based security.
Product innovation
Paradex also shows a less obvious advantage of the Starknet stack: it expands the design space for derivatives products. That shows up in concrete features and roadmap directions that are difficult to deliver without a high-performance, validity-proven settlement layer:
- Privacy-preserving trading flows
- No-ADL design choices: so profitable traders are not automatically de-levered
- Better-than-CEX execution goals: tight control over execution quality and market behavior
Why Starknet fits perps
Perp venues choose Starknet technology for a mix of trader-facing performance and builder-facing guarantees.
Throughput and execution capacity
For perp builders, the real question is not raw TPS. It is whether the chain can handle large, computation-heavy trading workloads efficiently.
That is where Starknet stands out. Cairo-native execution has materially improved perp performance and helped push sustained throughput above 100 TPS for actual perp trades, not simple transfers.
Starknet also supports unusually large transactions: a single transaction can bundle around a hundred Extended trades, making it possible to settle denser activity and more complex on-chain logic than most environments can comfortably support.
This matters most during liquidation spikes and fast market moves, when the system has to process heavy workloads predictably.
A proving stack built for real applications
STARK validity proofs are not a theoretical extra here. The proving layer sits on the critical path.
- Starknet now uses S-two on mainnet. Stone and S-two are open-sourced under Apache 2.0, along with other core components such as StarkEx and Cairo verifier contracts.
- For perps, one important property is the absence of a trusted setup. That removes one category of ceremony risk that other proving systems may inherit.
Just as importantly, the prover becomes part of product performance. Faster and cheaper proving gives derivatives venues more room to scale without weakening security guarantees.
Decentralization that trades can rely on
Starknet has a clearly defined decentralization roadmap, and several parts of it are already in motion.
Grinta introduced a decentralized sequencing architecture with multiple sequencers participating in consensus and block
production, alongside new mempool and fee-market primitives designed for a broader operator set over time. The longer trajectory includes staged staking responsibilities, decentralized block validation, and eventually decentralized proving.
For perps, this intersects with core exchange trust issues:
- neutral inclusion / ordering is enforcement against censorship and certain MEV pathologies, which become existential
once serious size trades onchain. - a multi-operator future is the difference between an exchange that can be paused by a single entity and a network
that continues to settle and prove even under adversarial conditions.
Account abstraction as trading UX
Perps win on usability as much as on raw performance. Starknet treats account abstraction as part of the protocol, not an add-on.
That opens the door to session-key style interactions, gas abstraction, and paymaster flows. For traders, that means a simpler path from wallet to trade. For builders, it means better onboarding, fee sponsorship, and safer authorization models without stitching together custom infrastructure from scratch.
A battle-tested team and stack
Perps builders choose stacks that de-risk operations.
StarkEx has been running in production on Ethereum mainnet since mid-2020, and it supports exchange-oriented safety and availability features such as forced withdrawals, escape-hatch style recovery paths, and flexible data availability configurations (e.g., rollup vs. validium-style modes).
That production history matters because it compounds into Starknet: the same core proving approach, the same Cairo-based provability model, and the same team that already supported exchange-scale workloads.
Build on Starknet
If you’re building a perp venue (or infrastructure for one), Starknet offers a foundation that has already been validated by leading derivatives teams.
For builders, that means you can design for high-performance trading without giving up on verifiability, self-custody, or long-term scalability. Just as importantly, open-source infrastructure, mature tooling, and established security patterns reduce the cost of building exchange-grade systems on top of Starknet.
Start with the Starknet developer docs to get up to speed on Cairo, account abstraction, tooling, and onchain architecture. And if you’re turning this into a real product, explore Starknet Grants for non-dilutive funding and ecosystem support.
