Trading Infrastructure On-Chain: What Practitioners Are Actually Navigating

On 25 August, P2P.org hosted Trading Infrastructure On-Chain: What Institutional Firms Actually Need, a practitioner roundtable on the market structure, latency, data, and execution questions shaping institutional trading on-chain. 

One thread ran through nearly every answer: the technology solving data, privacy, and execution problems is moving faster than the operational layer institutions actually need to trade at scale, settlement speed, best-execution reporting, and risk tooling among them.

Moderated by Max Mironov, General Manager of New Bets at P2P.org, the discussion featured Duncan Moir (President, 21shares), Oleksandr Proskurin (Co-Founder & CPO, Arkis), Genevieve Doo (Senior Account Manager, Talos), and Kyle O'Brien (VP of Capital Markets & Investor Relations, Zama).

LEARNINGS FOR BUSY READERS

• Institutional capital already trades on-chain, mostly through market makers, authorized participants, and liquidity providers acting on a fund's behalf, not through direct execution.
• Tokenized assets settle in milliseconds. The cash side of a trade is still bound to banking rails and compliance checks that don't move at that speed, and that mismatch, not chain throughput, is what keeps flow on centralized venues.
• Hyperliquid's on-chain order book has closed the gap with centralized exchanges on depth and speed, but on-chain markets still lack a consolidated tape, so proving best execution across venues remains genuinely harder.
• MEV is treated less as a bug to fix and more as a cost every panelist expects to manage indefinitely, with real disagreement over whether privacy tooling changes that.
• Risk management tooling, not regulation, was named as the clearest remaining infrastructure gap for institutional trading on-chain.

Who is actually trading, and through what

Duncan Moir opened by correcting an assumption in the question itself: institutional capital moving through an ETP still reaches on-chain execution, just through market makers and liquidity providers trading on the fund's behalf, increasingly on decentralized venues where liquidity depth requires it.

I think it's maybe a misconception that institutional investors don't understand this world. A lot of them just are not restricted from operating in it, and that's why they come to us. Something like half of our assets are hedge funds, prop desks, market makers, so they understand it very well.

Duncan Moir, 21shares

Oleksandr Proskurin pushed the timeline back further, arguing institutions have quietly driven most crypto trading volume for years, Aave and Uniswap included, and that what changed recently is visibility, not underlying participation. Genevieve Doo pointed to the actual constraint: settlement, not sentiment. Institutional interest is accelerating, but flow still defaults to centralized venues and OTC desks because the cash side of a trade hasn't caught up with how fast the assets themselves move.

Tokenized assets can move in literally milliseconds, but the settlement piece is often the more complicated part of the trade, and it cannot really move with that level of speed due to banking and compliance reasons.

Genevieve Doo, Talos

Asked what changed over the last year, the panel agreed on the direction: tokenization and real-world assets accelerated faster than expected, with Hyperliquid repeatedly cited as the venue that forced the pace, and clearer policy signals giving institutions more room to act on interest that already existed.

Where the data and latency gap actually sits

Genevieve gave the sharpest read on where on-chain data has closed the gap with centralized venues, and where it hasn't. Hyperliquid's central limit order book now runs at a scale comparable to major centralized exchanges and leads specifically in real-world asset perpetuals. What's still missing is consolidated depth: a centralized venue offers one order book and one tape, and on-chain markets have no real equivalent, which makes proving best execution across venues meaningfully harder.

Kyle O'Brien framed the structural issue underneath that gap. Public blockchains were built on the idea that verifiability requires public data, workable in crypto's early years, increasingly at odds with what institutions need now.

Many of us would agree that the original sin of crypto was that public verifiability required public data. As more institutions move on-chain, privacy has become somewhat of a prerequisite.

Kyle O'Brien, Zama

Zama's approach, built on fully homomorphic encryption, keeps data encrypted on-chain while remaining publicly verifiable, letting specific parties, an auditor or regulator, decrypt what they need without broadcasting trading activity to the rest of the network. Oleksandr described the mirror-image version of this problem from the operations side: Arkis computes margin off-chain against on-chain oracles while collateral custody and liquidation rules stay enforced by smart contracts, and treats redundant data providers as non-negotiable, since a single RPC node isn't an acceptable point of failure for a prime brokerage. Duncan, only half-joking, argued the opposite case: on-chain data staying hard to clean and extract currently gives an edge to firms willing to do that work themselves.

MEV: a cost to manage, not a bug to fix

Oleksandr set the tone early, with Arkis's own exposure coming mostly from its own smart contracts rather than cross-venue execution.

MEV is a curse, or a blessing. It's a blessing for the MEV bots, and the curse for anyone who is building.

Oleksandr Proskurin, Arkis

Duncan pushed back on the idea that institutions don't understand MEV, and drew a distinction that shaped the rest of the conversation.

You probably need to distinguish between predatory MEV and beneficial MEV. It also helps with price discovery, so it keeps spreads tight, which is good for the ETPs as well.

Duncan Moir, 21shares

Asked for a five-year outlook, the panel split. Kyle expects MEV in its current form to disappear if privacy tooling works as intended. Oleksandr expects it to persist regardless, noting that competing funds already track each other's on-chain positions closely. Genevieve suggested a middle path: MEV eventually settling into something closer to exchange fees, priced in and rarely discussed. Hyperliquid came up again as the clearest example of mitigation working in practice, through centralized transaction submission and binary node distribution that makes sandwich attacks structurally difficult to run.

The gap that isn't regulation

Duncan moved past regulation quickly, calling it broadly workable across most major jurisdictions today, with pace and cost varying by region rather than any hard blocker. The gap he actually named was risk management tooling: live portfolio analytics, backtesting, and scenario stress-testing at the level institutional equity desks already expect, which firms currently have to build in-house.

Kyle argued confidentiality and compliance, the two standard objections from a couple of years ago, are largely handled on the technology side now. What's left is distribution, getting confidential token support into the wallets, custodians, and exchanges institutions already use. Genevieve closed the point by noting institutional adoption doesn't move on one curve: some firms are still evaluating the technology, others have broad conviction and are waiting on a specific custody solution, and both groups are ultimately judging on-chain venues against the same two standards traditional finance already runs on, best execution and counterparty risk discipline.

Closing round

The sharpest exchange came between Duncan and Oleksandr on vault structures. Duncan raised an idea he says he regularly debates with his own legal team: a curator allocating through a smart contract may not legally be managing a collective investment scheme at all, since investors allocate directly and the contract executes. He noted regulators have already pushed back on that reading. Oleksandr agreed vaults expose real inefficiency in traditional asset management, but flagged the unresolved tension underneath their growth, onboarding a curator running traditional strategies still triggers standard proof-of-funds checks that anonymous on-chain deposit addresses can't easily satisfy. His bet was that vaults, like Bitcoin before them, eventually find a regulatory middle ground.

Kyle's closing prediction, offered as talking his own book, was that 95% of blockchain traffic gets encrypted through Zama's protocol within four years. Genevieve's was structural: crypto forced fragmented liquidity, 24/7 markets, and custody problems into the open earlier than other asset classes had to face them, and the infrastructure built to solve those problems now is what eventually gets reused as other asset classes move on-chain.

KEY TAKEAWAY

Across all four sections, the pattern held: data quality, MEV mitigation, and privacy tooling are improving quickly, Hyperliquid's rise is the clearest evidence of that. Settlement speed, consolidated best-execution reporting, and institutional-grade risk tooling have not kept pace. The firms building that missing operational layer themselves are the ones actually moving faster than the rest of the market right now.

You can watch the webinar recording here.

WORK WITH P2P.ORG ON TRADING INFRASTRUCTURE

Disclaimer: The views and opinions shared during this discussion are those of the individual speakers and do not necessarily reflect the views of P2P.org. This recap is intended to summarize the key themes discussed and should not be considered investment, legal, or financial advice. Latency and performance figures referenced are estimates based on internal benchmarks and may vary depending on network conditions, geography, and client infrastructure. Past performance is not indicative of future results.