Hyperliquid: Inside the $4 Trillion Onchain Market Machine
Hyperliquid began with a contradiction crypto had never properly solved.
The industry promised markets without custodial gatekeepers, opaque balance sheets or trusted intermediaries. Yet the venues where serious traders actually went for liquidity, leverage and execution were still centralised exchanges. They offered speed and depth, but required users to give up custody and trust the operator.
Hyperliquid attacked that contradiction directly. It did not try to make DeFi more ideological. It tried to make it good enough for traders who normally do not care about ideology at all. The result was an onchain derivatives venue that felt close enough to a serious centralised exchange to attract real volume, while preserving the self-custody, transparency and programmability that DeFi had long promised.
By mid-2026, Hyperliquid was no longer just a DEX in the old sense. It had become the reference case for onchain order-book trading, expanded beyond crypto perpetuals into commodities, equity-linked and pre-IPO synthetic markets, and begun moving into outcome and options-style instruments. What started as a better trading venue had become a purpose-built chain with an embedded exchange and a growing market-deployment layer.
The founder story matters because Hyperliquid looks like Jeffrey Yan’s biography translated into market structure. He was raised in Redwood Shores by a single mother who taught him to think “beyond the sky”; discovered elite mathematics late, failed, redirected himself into physics and won International Physics Olympiad gold; studied mathematics and computer science at Harvard; then joined Hudson River Trading, where he found markets intellectually pure but personally limiting. Yan did not leave because he disliked quantitative finance. He left because he wanted to build something where his own leverage over outcomes was larger.
That question led first to a failed 2018 prediction-market project, then to Chameleon Trading, and finally to Hyperliquid. It was not born from a venture deck or token launch. It was born from an operator’s frustration with the gap between what crypto claimed to be and where real liquidity actually lived.
This paper argues that Hyperliquid matters not only because it became one of crypto’s most successful trading venues, but because it is one of the clearest attempts to rebuild market structure from inside crypto’s own constraints. The central question is no longer whether onchain trading can work. Hyperliquid has answered that. The harder question is whether an exchange-funded network can become neutral infrastructure for a broader share of global finance.
Before Hyperliquid
Jeffrey Yan’s background is not interesting because it is conventionally inspirational. It is interesting because the pattern of his decisions before Hyperliquid helps explain the pattern of the system after it.
He grew up in Redwood Shores, surrounded by the Bay Area’s technology wealth but not born into its affluent core. After his parents divorced, his mother, an accountant, raised Yan and his younger sister largely on her own. The family situation matters less as biography than as operating psychology. Yan’s later insistence on self-direction, discomfort with performative status and indifference to conventional corporate markers all fit someone who grew up near Silicon Valley’s promise without inheriting its security.
His mother’s phrase, “beyond the sky”, appears to have given him both ambition and permanent anti-complacency. That duality, extreme competitiveness paired with scepticism toward celebration, reappears throughout Hyperliquid’s culture.
The most predictive detail from Yan’s adolescence is not simply that he was gifted. It is that he entered elite competition late and then tried to compress years of preparation into months of self-study. He discovered serious mathematics only in eighth grade, after following a friend from a private school to a competition, then began waking at five in the morning to work through old olympiad papers alone. Within a year he had reached the U.S. Math Olympiad training camp, but failed to make the national team. Instead of treating that failure as final, he redirected himself into physics, taught himself from upperclassmen’s textbooks and Feynman lectures, and within roughly a year rose to the top tier of U.S. high-school physics, eventually winning International Physics Olympiad gold.
The relevant point for Hyperliquid is not merely that Yan was smart. It is that he appears to specialise in identifying the distance between himself and an elite frontier, then closing it through concentrated, self-imposed work. Hyperliquid’s rapid movement from testnet concept to custom chain and institutional-scale venue looks like the adult version of the same pattern.
Harvard reinforced that orientation. Yan studied mathematics and computer science, took on difficult technical work early, and finished first in a notoriously difficult algorithms course as a freshman. More important than the credential was the peer group. He formed close ties with other olympiad-calibre students, including Scott Wu, later a founder of Cognition AI. Their conversations centred on what it means to be the best in a field and what constitutes the “essence” of high performance. That line matters because Hyperliquid does not read like a product built by people optimising for messaging or fundraising. It reads like a product built by people obsessed with the purest structural version of a problem. Yan’s later aversion to half-measures, especially his reluctance to accept AMM or hybrid-order-book compromises, makes more sense in that context.
His internships also foreshadowed the synthesis. At Google X and Nuro, he worked around autonomous systems. At Tower and then Hudson River Trading, he moved closer to a world where latency, routing and execution quality define economic reality. He joined HRT full-time in late 2017, liked trading intensely, and saw markets as a pure real-world game in which smart people compete while producing useful liquidity and efficient prices. Yet he left after only about eight months. The issue was not that the work felt trivial. It was that he felt too replaceable inside a machine that would already run well without him.
That dissatisfaction is conceptually important. Hyperliquid is partly a response to market-structure flaws, but it is also a response to Yan’s desire to build a system in which his own leverage over outcomes would be larger than it had been at HRT.
That desire produced his first major crypto project. In 2018, after becoming convinced by Ethereum’s broader implications, Yan left HRT with his Harvard roommate Brian Wong and built Deaux, a prediction-market startup. Deaux pursued offchain matching with onchain settlement because Ethereum itself was too slow to run a serious exchange. Yan later said the team had many things right on the infrastructure side, but was not yet ready as builders and had launched into the wrong market backdrop.
Deaux appears to have taught four lessons that later shaped Hyperliquid. First, architecture does not rescue a product nobody wants to use. Second, neutral rails do not remove regulatory reality, especially in anything resembling event markets. Third, user experience matters as much as idea purity. Fourth, some financial primitives only become viable once surrounding market infrastructure is already strong enough to support them. When Yan later returned to outcome markets, he was not repeating 2018. He was returning with a chain, a user base, a liquidity pool, deployer tooling and real fee flow already in place.
After Deaux failed, Yan returned more than half of the roughly $450,000 raised, waited out his HRT non-compete and eventually moved to Puerto Rico in late 2019. This became the Chameleon Trading period, and it is hard to understand Hyperliquid without it. Chameleon began with about $10,000 of Yan’s savings and developed into a substantial anonymous trading operation. It was initially run from a one-bedroom apartment near the beach, with Yan writing Python scripts, monitoring exchange APIs and iterating relentlessly.
The key point is not the exact return profile. It is that Chameleon trained Yan and his collaborators directly on the weaknesses of centralised crypto exchange infrastructure. They learned from the inside how fragmented, inefficient and strategically gameable those markets were. That is founder-market fit in the strictest sense.
Chameleon also explains why Hyperliquid could be bootstrapped. The trading operation appears to have created enough personal capital for Yan to fund Hyperliquid’s incubation without relying on venture financing. That mattered because it removed an ordinary startup constraint. Hyperliquid did not have to accept venture money to survive, and it did not have to sell token inventory to subsidise early market makers. Those absences, especially in crypto, shaped everything downstream: ownership, go-to-market, credibility and the cultural force of the eventual HYPE launch.
Why shut down Chameleon if it was profitable? The answer appears philosophical as much as economic. By 2022, Yan had become frustrated that crypto had recreated many of the same centralised trust structures it had promised to eliminate. Chameleon let him extract value from market inefficiencies. Hyperliquid offered the possibility of redesigning the infrastructure that created those inefficiencies in the first place.
In that sense, the move from Chameleon to Hyperliquid was not a shift from finance to ideology. It was a shift from trading against bad rails to trying to build better ones. That decision, more than any token launch or valuation mark, is the actual founding act of Hyperliquid.
Why Hyperliquid had to exist
To understand why Yan believed Hyperliquid had to be built, one has to start with the structure of crypto derivatives before Hyperliquid. In crypto, perpetual futures became the dominant speculative instrument because they are easier to hold than dated futures and more capital-efficient for traders who want continuous exposure. The broader product idea traces back to Robert Shiller in the 1990s, but in crypto it became practical through BitMEX and then spread across major centralised venues. By 2025, the global perpetuals market had become enormous: Reuters reported $61.7 trillion of annual trading volume, while offshore venues had normalised leveraged, always-open directional speculation for a global customer base.
Centralised exchanges such as Binance and FTX therefore solved one problem while exacerbating another. They gave traders tight spreads, familiar order books, high leverage and reliable execution. But they required trust: trust in custody, trust in solvency, trust in liquidation handling, trust in internal conflict management, and trust that the infrastructure was not being run for insiders. Yan’s team had direct exposure to that world first through bot-running and market making, then through DeFi. Their conclusion was that crypto’s most liquid venues had reintroduced the intermediation problem that Bitcoin and Ethereum were supposed to remove. FTX’s collapse did not create that view, but it made it impossible to treat it as theoretical. In Yan’s own framing, FTX was the call to arms, not the original idea.
The obvious alternative, early DeFi, was not good enough. Yan’s early explanations of the team’s reasoning were unusually clear in market-structure terms. They were not committed to order books for ideological reasons. They started from what traders actually wanted: low latency, visible liquidity, low slippage and a clean user experience. On that standard, the first-generation DEX stack failed in several ways. Constant-product AMMs were clever bootstrapping tools, but from a market maker’s perspective they often meant being arbitraged back and forth, especially once liquidity mining faded. Yan’s critique of AMMs was therefore structural, not tribal. They distributed execution against passive liquidity providers in ways that could be unsustainable for serious market making unless heavily subsidised.
Oracle-based and hybrid designs fixed some of those problems but introduced others. Yan’s comments on GMX are telling. He treated oracle-dependent systems as ingenious band-aid solutions: more capital-efficient than simple AMMs, but still dependent on trusted oracle logic, arbitrary rules and exploit-sensitive boundaries. A design that required constant fine-tuning around oracle manipulation was not, in his view, a final answer for the core market venue. He applied an even sharper criticism to dYdX v3. Its then-live order book, in his view, was effectively run offchain in a centralised manner, with settlement onchain but order placement and cancellation dependent on a privileged server. In economic terms, that meant the most latency-sensitive part of the venue, the part that most determines market quality and fairness, remained outside transparent public consensus. Hyperliquid’s original product thesis was built around eliminating precisely that split.
That helps explain why Hyperliquid decided not merely to build a new interface, but to build a new chain. The team initially experimented on Arbitrum and considered more modest approaches, including hybrids and batch auctions. The conclusion after the testnet period was that existing infrastructure could not support the experience they wanted. Order placement, order cancellation, matching, margining and liquidation all had to live close enough together, and move fast enough, that the user would not feel they were trading on a compromised imitation of a centralised venue. That meant sacrificing the convenience of deploying on a general-purpose chain and instead building a purpose-built execution environment. Yan has said plainly that if a performant L2 had existed that could handle the required order throughput, they would have used it. The custom chain was therefore not recursive crypto empire-building. It was a technical necessity created by the product specification.
This is the central strategic point. Hyperliquid did not build its own chain because having an L1 was fashionable, or because a token needed a home. It built its own chain because Yan and the team believed the only way to make order-book trading, liquidations, margin and user experience work together under transparent, non-custodial constraints was to treat the whole system as a unified state machine. The chain was originally framed as a means to support the exchange, not as a universal-purpose platform. That narrowness was a strength early on. It let the team optimise for one high-value use case instead of pretending to solve all blockchain problems at once.
The timing matters too. FTX was the emotional catalyst because it proved that trusted intermediaries could vaporise customer confidence overnight. But Hyperliquid’s founding frustration predated the collapse. The team had already been building for months. Their regret, as Yan later put it, was that they had not started earlier. That distinction is analytically important. If Hyperliquid had been merely an anti-FTX product, it might have remained a niche custody-safe exchange. Instead, it was built from the beginning around a stronger claim: that DeFi had the right values but had not yet produced the right execution environment. FTX sharpened the mission. It did not invent it.
The order-book decision then had second-order implications. Putting the book onchain improves auditability, data availability and the ability to reconstruct every order, cancellation and liquidation. That creates a transparency advantage, but it also imposes harsh performance requirements and exposes the venue to a different kind of scrutiny. The later debates around October 2025 liquidations and the JELLY incident show both sides of that trade-off. Hyperliquid can be audited in ways centralised exchanges cannot, yet its visible stress episodes also become headline material because the data are actually there. The platform’s broader strategic bet is that over time, the transparency dividend will outweigh the optics penalty. The June 2026 academic work on Hyperliquid’s microstructure points in that direction, finding that visible TWAP orders can reduce execution costs relative to comparable hidden metaorders and attract liquidity provision while shifting adverse-selection costs toward those who choose not to preannounce. In market-structure terms, Hyperliquid’s openness is not just moral theatre. It changes execution.
That is why Hyperliquid had to exist in Yan’s eyes. Not because crypto lacked another exchange, but because it lacked a non-custodial venue that treated order-book performance, transparent execution and self-custody as parts of one problem rather than separate modules. In that sense, Hyperliquid’s intellectual origin is simpler than the mythology around it. Yan and his colleagues believed traders wanted what institutional traders had always wanted, and believed crypto’s existing stack either could not or would not provide it without putting trust back in the middle. Hyperliquid began as the attempt to remove that contradiction.
From a perpetual DEX to a financial network
Hyperliquid’s first important strategic move after launch was not the token. It was liquidity bootstrapping. The exchange launched publicly in late February 2023 and, for a time, attracted mainly experimental users, including NFT traders testing perpetuals for the first time. The professional liquidity the venue needed did not appear automatically, and Yan refused to solve that with the standard crypto toolkit of paying market makers in cash, equity or token entitlements. Instead, in May 2023, he pushed Chameleon-style strategies into the Hyperliquidity Provider vault, or HLP. Users could deposit capital into a transparent onchain market-making and backstop-liquidity vehicle and receive the economic results directly, with no management fee or carry. This was not only a clever bootstrap. It was also a political statement: one of the most lucrative activities in crypto microstructure would be opened, in principle, to ordinary users rather than reserved for a closed set of firms.
HLP also solved a reputational problem. Hyperliquid needed liquidity, but did not want to depend permanently on a house market maker analogous to the way Alameda had been intertwined with FTX. Yan explicitly wanted HLP not to become essential to Hyperliquid’s functioning. The vault therefore acted as a bridge between a zero-liquidity startup and an exchange that could attract independent market makers without paying them directly. This is one of the clearest examples of the Hyperliquid pattern: use a tightly integrated internal primitive to solve a launch problem, then try to decentralise the economically important function outward over time.
The next strategic turn was spot trading, and its significance is often underappreciated. Perpetuals can be settled economically without anyone holding the underlying asset. Spot markets cannot. When Hyperliquid moved into spot, it collided with the hardest custodial question in finance: who actually owns and settles the asset? This appears to be the moment when Yan stopped thinking of Hyperliquid merely as an exchange on a chain and began thinking of it as a chain with an exchange built into it. That conceptual inversion is crucial. It opened the path to HyperEVM, external builders and the broader ambition to house more of finance on the network. Spot was not an adjacent product. It was the forcing function that pushed Hyperliquid from application to platform.
The HYPE genesis in November 2024 then did several jobs at once. Roughly 31% of total supply was airdropped to about 94,000 early users, while the team’s own allocation was 23.8% and vested over years. The public record supports the broad direction of travel: no venture investors received a private allocation, no insiders received a preferential public-market entry, and the airdrop handed meaningful ownership to the user base. Strategically, that was at least four things simultaneously. It was a decentralisation event, because it widened token ownership. It was a go-to-market event, because it rewarded use rather than marketing reach. It was a cultural legitimacy event, because crypto participants interpreted it as unusually fair. And it was a loyalty mechanism, because users who had previously been customers suddenly owned part of the network’s future economics. It was not pure altruism, and it did not need to be. It was the most efficient way Hyperliquid could buy long-term legitimacy with an asset it controlled.
The airdrop also changed the chain’s governance and security trajectory. Yan has described the token generation event less as an operating milestone than as a decentralisation milestone that enabled native proof-of-stake consensus and a more distributed validator set. The implication is that HYPE was never meant to be only a speculative asset. It was security collateral, deployment collateral, alignment collateral and, increasingly, a meta-asset around which other network functions could be organised. That matters for valuation debates. If HYPE were only a fee-abstracting exchange token, Hyperliquid would look like a high-margin trading venue with a popular loyalty asset. Because HYPE also sits in staking, deployment and potentially application-layer composability, the token’s economic meaning is broader, though still not broad enough by mid-2026 to sever its dependence on trading activity.
This is where HyperEVM enters. Yan has described HyperEVM not as another EVM chain, but as a programmable portal into Hyperliquid’s native primitives. That framing is important. HyperEVM’s strategic utility is not simply that developers can port Solidity contracts. It is that those contracts can access balances, liquidity, staking and exchange-state primitives already native to Hyperliquid. If that architecture works, then applications built in the EVM environment do not need to recreate a separate liquidity universe. They can plug into the one that already exists. This is the operating-system analogy in its most credible form. Hyperliquid is not aspiring to be a better app store full of random dApps. It is trying to be a financial substrate with an immediately useful liquidity core.
Builder codes were the first step in exporting distribution. They allowed third-party developers to build user-facing trading apps on top of Hyperliquid’s liquidity and keep a cut of the fees their users generated. By 2026, those builders had reportedly earned more than $70 million since October 2024. Matt Huang’s description of this as franchising out the user experience is analytically useful because it captures the organisational difference from Binance or Coinbase. Hyperliquid is trying to separate infrastructure ownership from interface ownership. If that model scales, the core protocol becomes harder to displace because it no longer needs to win every end-user relationship directly. The risk, of course, is fragmentation and lower quality control. But the strategic direction is clear: Hyperliquid wants outside entrepreneurs to own businesses on top of the network.
HIP-3 extended that principle from interfaces to markets themselves. It created a mechanism through which anyone staking sufficient HYPE could deploy perpetual markets, choose parameters and retain half the trading fees. Yan’s comments on the model make the philosophy explicit: finance is too large and too specialised for one team to list, source and operate everything internally. A centralised listing desk may move faster at first, but a permissionless deployer network can be more robust and globally scalable if it works. This is one of Hyperliquid’s most consequential design choices. HIP-3 is not merely a new market category. It is a decentralised listing system, a fee-sharing model and a distribution model for domain expertise. It turns the exchange from a venue with a catalogue into a platform that lets specialists create the catalogue.
By early 2026, that thesis had clearly moved beyond crypto. Trade[XYZ], the leading deployer, had launched markets in silver, crude oil, stock indices and foreign exchange, while independent deployer markets had reportedly grown to a significant share of total volume. Yan also said HIP-3 silver markets had reached about 2% of global silver price-discovery volume. Even if that figure is treated as a founder statement rather than a fully audited exchange comparison, it captures the ambition correctly. Hyperliquid is no longer trying only to put crypto derivatives onchain. It is trying to make perpetuals the generic market wrapper for anything liquid enough to support them.
The March and June 2026 Wall Street Journal reporting makes clear why this matters. During the Iran-related oil shock, Hyperliquid’s 24/7 oil perps gave global traders a venue for continuous positioning while mainstream futures markets were closed, and cumulative oil-futures volume on the exchange reportedly jumped from roughly $339 million to about $7.3 billion within days. Shortly after, S&P Dow Jones Indices licensed the S&P 500 to Trade[XYZ] for what the WSJ described as the only licensed S&P 500 perpetual contract. By June, Hyperliquid’s SpaceX-linked perpetuals had become one of the platform’s most actively traded instruments, and pre-IPO contracts were being treated as real-time speculative valuation signals for one of the world’s most anticipated flotations. The significance is not that Hyperliquid has solved equity ownership or commodities clearing. It has not. The significance is that it has already created a credible market-creation layer for synthetic exposure where legacy market hours, access rules and listing processes are slower or narrower.
HIP-4 represents the next extension. Yan and the Colossus profile both frame it as the route back into options and prediction-style, or more accurately outcome-linked, markets. The distinction is useful: spot and perps express linear exposure, while outcome markets let users express bounded or nonlinear beliefs. By mid-2026, the WSJ was already describing Hyperliquid as having recently introduced outcome prediction markets and options trading. That is enough to say the product line had moved from proposal into public-facing rollout. It is not enough to treat HIP-4 as a mature or deeply liquid business line. The current public record supports viewing CPI, Fed or sports-linked markets as plausible early use cases rather than proven growth pillars.
The collateral story is strategically important, but still less fully documented than the exchange and HIP-3 businesses. Yan has described an aligned stablecoin model associated with Native Markets, lower fees for users of the aligned stablecoin and protocol-level revenue-sharing logic. Later June 2026 materials suggest a transition from USDH toward USDC as the aligned quote asset, involving Coinbase and Native Markets, though the final legal and operational structure remains unclear. The narrower conclusion is still meaningful: Hyperliquid clearly views quote-asset design as strategic infrastructure, not as a neutral afterthought. Whether the final form proves to be a cleaner institutional on-ramp or a new dependency on a regulated U.S.-linked issuer remains open.
So what category does Hyperliquid belong to by June 21, 2026? The answer is that it spans several, but not equally. Economically, it is still primarily an exchange whose revenues and attention are driven by leveraged trading. Architecturally, it is a blockchain with a native exchange and a programmable layer. Organisationally, it is moving toward a market-creation network in which external deployers and interfaces own more of the product surface. The most useful label today is probably not “exchange” or “L1”, but “financial network with an embedded flagship exchange”. The strongest version of the financial-operating-system claim remains aspirational. The network has the right architectural shape; it does not yet have enough non-perps economic weight to make the thesis fully earned.
The economics of Hyperliquid
Hyperliquid’s economics require careful vocabulary. The available dataset records 1.20 million total users, $4.36 trillion in cumulative volume, $372.19 billion in cumulative deposits, $368.49 billion in cumulative withdrawals, and a quarterly revenue table running from 2025 Q3 through 2026 Q2 quarter-to-date. The Q2 QTD structure aligns with an 82-day window from April 1 through mid-2026, which makes the quarterly table best understood as a mid-2026 quarter-to-date snapshot. The 24-hour market data should be treated separately as a late-June point-in-time snapshot rather than live market data.
The first analytical discipline is terminology. In this dataset, reported cumulative “revenue” includes HyperCore buybacks, HyperEVM burns and auction fees. The “costs” line consists entirely of HLP costs. On that basis, the cleanest label is protocol revenue and value accrual, not company revenue. Hyperliquid Labs is not shown as directly receiving trading fees in the visible protocol-level accounting, and Yan has stated that Hyperliquid does not operate a discretionary buyback programme and that Hyperliquid Labs is not the same thing as the protocol. Colossus separately reports that none of the protocol fees flowed to the team and that Yan still personally funded many team costs. The implication is straightforward: the visible onchain accounting is a protocol-level ledger, not an income statement for a conventional software company.
This is an excerpt from the full insights4vc research report.
The full version includes Hyperliquid’s protocol economics, HIP-3 market expansion, competitive positioning, and more.
Sources
https://colossus.com/article/beyond-the-sky-jeffrey-yan-hyperliquid/
https://www.ft.com/content/7aeef922-6d26-4a65-97b1-751c36440bf0
Partner Highlight
At insights4vc, we pay attention to teams that are not just shipping products, but shaping categories.
Moto is one of them. The company is building a premium card experience positioned as an alternative to traditional incumbents such as Amex, combining a high-end user interface with more modern financial infrastructure.
Moto is currently expanding access to a limited testing group. If you would like to test the product please contact us at: office@insights4.vc
Cover Artwork
Farmers and their herd near a fountain in front of a panoramic landscape
Francesco Zuccarelli, R.A. c. 1702–1788
Risk Disclaimer:
insights4.vc and its newsletter provide research and information for educational purposes only and should not be taken as any form of professional advice. We do not advocate for any investment actions, including buying, selling, or holding digital assets.
The content reflects only the writer’s views and not financial advice. Please conduct your own due diligence before engaging with digital assets or related technologies, as they carry high risks and values can fluctuate significantly.