Jeff Yan On Hyperliquid’s Plan To Bring All Finance Onchain
Summary
Jeff Yan’s overarching strategy is to build a small set of composable primitives that can ultimately “house all of finance.” Portfolio Margin exemplifies the model: borrowing, lending, spot, and perps should reinforce one another, while Hyperliquid increasingly resembles neutral public infrastructure—“the internet or Linux”—that institutions can use rather than a venue designed to displace them.
The host cited real-world-asset open interest rising from $300 million in January to $3.9 billion, a 13x increase, while HIP-3—on testnet—could extend the same infrastructure into regulated markets.* HIP-3* adds opt-in controls such as trader whitelists; Payward, Kraken’s parent, has said that pending regulatory approval it wants to use the system to serve US users. Yan described the design as adding deployer-selected controls while preserving HIP-3’s transparency, self-custody, resiliency, and existing on-chain functionality.
Yan now sees options as Hyperliquid’s clearest unmet product opportunity and HIP-4 as the likely foundation. He cautioned that timing is hard to predict and some primitives may never take off, but a year ago he wondered whether options traders might simply prefer perps; his answer today is “no,” because some views require convex payouts. An options venue could inherit spot order books and Portfolio Margin while using perps to hedge delta—the composability that might help the category finally “reach escape velocity in crypto.”
Priority fees aim to replace the capital-intensive latency arms race with open, protocol-level competition. The host described read priority as streaming uncommitted order flow from the mempool and write priority as attaching HYPE to orders for queue priority, and noted that almost 180,000 HYPE had already been burned. Yan emphasized the market-design benefit: microwave towers and transatlantic cables create “very obviously zero-sum” technological alpha, whereas an evolved priority-fee system could let anyone express genuine market insight with minimal infrastructure investment and “trade on equal footing.”
Yan argues that greater transparency should generally improve liquidity and large-order execution. The host cited a study of hidden versus publicly telegraphed “T-webs” that found the public flow had better execution because market makers could absorb it. Hyperliquid exposes an L4 order book down to the addresses behind orders, extending electronic markets’ move beyond the pits where the “tallest person, loudest person gets a better fill.” Leveraged ETFs’ formulaic closing-auction rebalances illustrate why predictable flow can attract competitive liquidity rather than worsen execution.
The next stack is spot, perps, and options, but bringing assets onchain is harder than tokenizing their prices. Portfolio Margin becomes more useful as users can pledge the things they already hold; because holding assets is not itself an economic activity, that network effect is harder to bootstrap. Yan therefore sees a high-quality issuance layer as essential to the all-finance thesis. Perpetuals reduce real-world assets to a comparatively elegant mathematical price problem; issuing the underlying assets requires a slower network effect.
Yan measures success less by headline metrics than by “zero to one qualitative shifts” that users voluntarily adopt. Scaling will remain unfinished—moving from one million to 10 million users simply creates a new engineering problem—and the primitive set cannot be frozen if it is meant to accommodate global finance. His cultural warning is equally explicit: remain humble, serve end users, and avoid turning ecosystem growth into complacency, beating competitors, or proving a point. He offered no sweeping community critique because the community is diverse.
Deep dive
1. Hyperliquid is being designed as public infrastructure, not a bigger exchange
Yan’s organizing principle is usefulness: if builders cannot create a financial product on Hyperliquid, “they’re gonna build elsewhere.” The protocol should therefore respond to recurring builder constraints rather than accumulate features for their own sake.
The architectural preference is “fewer things, but they work together well.” Each primitive should be “surgically precise,” broadly useful, and composable—Portfolio Margin shows how native borrowing and lending can combine with spot and perps instead of becoming another isolated application.
Yan resisted every competitive comparison offered, from Aevo and dYdX through major crypto exchanges to CME and Nasdaq. His preferred analogy is “public infrastructure, something like the internet or Linux”: one neutral system that existing institutions could use if it fits their needs, rather than an ecosystem built around displacing them.
2. HIP-3* makes regulated access an opt-in extension of permissionless markets
The host’s evidence for HIP-3’s breakout was real-world-asset open interest rising from $300 million in January to $3.9 billion, or 13x. The host also pointed to a lag in public understanding: TradeXYZ and Shoku’s execution helped demonstrate what HLP3 could enable. Markets such as 24/7 trading and pre-IPOs have helped shift perceptions of Hyperliquid from a crypto-perps protocol toward broader financial infrastructure.
HIP-3* is on testnet and retains HIP-3’s market-deployment model while adding controls such as trader whitelists. Payward, Kraken’s parent company, has publicly said that, pending regulatory approval, it wants to use this structure to offer Hyperliquid to US citizens.
Yan framed those permissions as distilled builder feedback, not a reinvention of the base system. Deployers can opt into the controls their regulatory regime requires while retaining HIP-3’s transparency, self-custody, redundancy, resiliency, and synergy with the protocol’s existing on-chain functions.
3. Options complete the trading stack; spot completes the collateral stack
Yan’s change of mind is important: a year ago, he was unsure whether options users would simply prefer perps, but “I think the answer is no.” Perps satisfy many traders, yet cannot express every market view that requires an option-like or convex payout. He cautioned that some primitives may never take off and that the timing of adoption is hard to predict.
If forced to build an application rather than infrastructure, he would investigate options. HIP-4 offers an obvious route: builders can inherit spot order books and Portfolio Margin, then hedge an options book’s delta on the existing perps venue.
The unresolved product question is how to make that form factor broadly appealing. Convex products have not, in Yan’s view, “reached escape velocity in crypto,” despite their success in traditional finance; builders must execute the product and tell infrastructure developers where HIP-4 still falls short.
Spot creates a different bottleneck. Portfolio Margin is only as useful as its supported collateral, so Hyperliquid must get assets—not merely their prices—onchain. Holding assets is not itself an economic activity, making this network effect harder to bootstrap. Yan called a high-quality issuance layer unavoidable, while acknowledging that tokenizing a real-world asset is substantially harder than creating a perp that tokenizes its price.
4. Better market structure means internalizing latency and exposing information
The host described read priority as allowing traders to stream uncommitted order flow from the mempool and write priority as allowing HYPE to be attached for queue priority; he also noted that nearly 180,000 HYPE had already been burned through the mechanism. Yan rejected describing write priority as a “bribe.” He sees its purpose as internalizing negative externalities and reducing the importance of incidental technological advantages.
The target is incidental technological alpha. Microwave towers and transatlantic fiber let traders react across venues faster, but shaving a microsecond from a ticker’s BBO update is “just not valuable to the world”; in a “super-equilibrium” state, Yan argued, the people laying cables or building microwaves ultimately collect the HFT profits.
Yan distinguished that latency advantage from unique alpha, which markets need participants to bring and combine into the crowd’s collective signal. Without the baseline latency spend in traditional finance, a participant may be unable to express a useful signal at all, depriving the market of information. His desired end state lets traders bring that alpha with minimal infrastructure investment and compete “on equal footing.”
His transparency argument starts with trading pits, where the “tallest person, loudest person gets a better fill.” He argues that when everyone can see the order book, they can incorporate it into their algorithms and make the book more liquid; there is no a priori reason that today’s transparency level is optimal.
The host cited a study comparing hidden and publicly telegraphed “T-webs” that found the public flow had better execution because market makers had time to absorb it. Leveraged ETFs offer another example: their predictable daily rebalancing is largely conducted in closing auctions, where liquidity providers can account for the non-toxic flow. Hyperliquid extends the transparency principle further with an address-level L4 order book.
5. Usage-driven scaling and community humility remain the operating tests
Performance will never be finished. Yan called overwhelming organic block-space demand “the best problem to have”: discovering that infrastructure is inadequate because usage rose from one million people to 10 million is precisely the engineering challenge Hyperliquid wants.
Nor should the primitive list ever become fixed. Infrastructure intended for the entire global financial system cannot credibly declare itself complete while its users, institutions, and builders continue developing new requirements.
Yan’s personal scorecard is qualitative: are “net new cool things that haven’t been ever built before” appearing, and are users choosing them? He said the team does look broadly at metrics, but these zero-to-one shifts are the greatest source of fulfillment.
His request to the community was to remember the end user, “stay humble,” and avoid complacency, beating competitors, or proving a point. Three years earlier, Hyperliquid was only an idea sustained by a small group working day and night, and he offered no broad critique because the community is too diverse for sweeping feedback.