My Personal Highest-Conviction Bet

Today's piece is a personal one.

I'm writing about the largest position in my portfolio, so read everything below with that in mind. I could have laid it out as a valuation model with a price target, and those numbers are in here toward the end, but leading with them would skip the part that actually convinced me.

This is a company most people outside crypto have never heard of. Eleven employees, no investors, and a founder who has turned down more money than most startups ever raise. What made me hold it has as much to do with how it was built as with what it earns, and those two things turn out to be the same story.

So I want to tell it properly, from the beginning, in a way that makes sense whether or not you have ever opened a crypto exchange in your life.

An exchange never has to be right about anything.

It takes no position and carries no inventory. It does not need to know whether oil goes to $40 or $140, only that enough people disagree strongly enough to trade on it. Every disagreement pays a toll, and the exchange collects from both sides of it.

The most important thing about an exchange is its liquidity. Traders go where other traders already are, because that is where orders fill at good prices. More traders make the venue better, which attracts more traders. Liquidity is the rare product that improves purely by being used, and once an exchange wins a market it tends to hold it for a very long time.

The proof is in how old these companies are. The New York Stock Exchange traces back to an agreement signed under a buttonwood tree in 1792. The CME was founded in 1898 to trade butter and eggs and now sets the price of most of the world's commodity futures. Very few businesses of any kind survive that long. Almost none of them keep their monopoly the whole way.

Crypto proved the model works even faster without the paperwork. Binance processed $25.1 trillion in derivatives volume in 2025, roughly 30% of the entire global crypto derivatives market.

Every one of these businesses shares the same requirement. Your money sits on their balance sheet. You see the prices they decide to show you. Their books are private, their liquidations are unverifiable, and when something goes wrong you learn about it afterward.

For two hundred years that was simply the cost of having a market at all.

He grew up in Redwood Shores, between San Francisco and Palo Alto, in the shadow of Oracle's mirrored headquarters. His parents divorced when he was in third grade, his father left, and his mother raised him and his younger sister alone on an accountant's salary, working overtime through every tax season.

She had a saying, a Chinese idiom:

He taught himself everything. There was no tutor and no summer program his family could afford. At thirteen he discovered competitive mathematics, realized there was a race he had never known about and that he was already years behind, and started waking at five in the morning to work through old competition papers alone in his room. Within a year he was among the top fifty high school mathematicians in the country. He taught himself physics next, out of textbooks meant for students three years older, and at eighteen he won a gold medal at the International Physics Olympiad, ranked twenty-fourth in the world.

He came home with his mother's idiom finally measured. Twenty-three people stood beyond him.

Harvard covered nearly all of his tuition. After his studies he joined Hudson River Trading, one of the most sophisticated trading firms on earth, on its algorithms team.

His manager noticed something odd that took a while to place. New hires hit walls in the code, you work through it together, then they hit the next wall. Yan never hit walls. He arrived every week with ideas. Everything about his work was excellent and none of it appeared to matter to him at all.

He lasted eight months.

What he could not stop thinking about was a question with no comfortable answer. He had spent eight months improving a system that was already exceptional, inside a firm that would remain exceptional without him. So what value was he actually adding to the world?

The answer arrived in a document. Reading Ethereum's yellow paper, he found a description of a computer that no single party could switch off, running calculations the whole world agreed on. A way to replace trust with code.

He decided to leave and pursue it on his own.

It took him a while to work out what that meant. In late 2019 he moved to Puerto Rico with $10,000 and a suspicion that something large was coming, and started trading crypto himself. He named the operation Chameleon, after his gaming handle from middle school. He shared a one-bedroom apartment near the beach for under $2,000 a month, never bought a monitor, worked off the living room television, and put in fourteen hours a day.

The opportunity he had spotted was structural. Trading equities professionally required connections to thirteen exchanges, three colocation sites in New Jersey, compliance with a thicket of federal regulation, microwave links to Chicago, and tens of millions of dollars of setup cost. In crypto, everyone connected to the same improvised web infrastructure. A man with an Amazon server could compete with Jane Street.

For two and a half years, Chameleon compounded at thousands of percent a year. It grew from one man and $10,000 into six employees and one of the most substantial anonymous trading firms in crypto, and it made him very rich before he turned twenty-eight. To every bartender and surfer in San Juan he was just another kid in board shorts.

By 2022 he had been in crypto four years and had started to care about it beyond his own profit and loss, which is where the trouble began.

The logic bothered him. Bitcoin had given the world a way to hold and move money without trusting an intermediary. Ethereum had given it a computer nobody could switch off. Between them they contained most of what you would need to rebuild the financial system from scratch, and almost nothing had been built with either. The two largest exchanges in crypto were private companies holding customer funds on balance sheets nobody could inspect. Crypto kept quietly reintroducing the exact thing it had been created to remove.

In November, FTX collapsed.

The third largest exchange in crypto, valued at $32 billion, had been lending customer deposits to a trading firm run by the founder's girlfriend.

The result: Yan shut Chameleon down.

Not because the business was struggling. Not because there was no more room to grow.

He told his six employees they were finished trading. Several of them did not understand the decision and left, and it changed nothing, because there was no board to persuade and no investor to answer to. It was his money and his decision.

Put the decision in plain terms. He closed a business printing extraordinary money to go and build a decentralized exchange, in a category that had never once worked, weeks after the industry's credibility had been set on fire, in the worst bear market in crypto history, funded entirely out of his own savings.

Almost nobody thought it was a good idea.

Hyperliquid is a decentralized exchange for perpetual futures, running on a purpose-built Layer 1 blockchain of its own. The pitch then is the pitch now: the speed and feel of a centralized exchange with the transparency of DeFi. A fully onchain order book handling up to 100,000 orders per second, sub-second finality, gasless orders, and no company anywhere holding your funds. In plain terms, it looks and trades like Binance, except nobody can freeze your account or lose your money, because there is no company in the middle to do either.

It launched at the end of February 2023. For the first two months there were no serious traders on it at all. The users were mostly small fish who had never touched a futures contract, placing ten-dollar trades and learning what leverage meant through paper trading competitions.

The problem he faced is the oldest one in the exchange business, and it is worth understanding because everything that follows is a consequence of how he solved it.

An exchange is worthless without liquidity. When you place a trade, somebody has to be standing on the other side, ready to buy what you are selling at a fair price. Usually that job belongs to market makers, professional firms that quote both sides of a market continuously and earn the spread between them. They are the reason you can sell a hundred thousand dollars of Bitcoin instantly without moving the price against yourself.

Which gives them enormous leverage over any new venue. No liquidity means no traders, and no traders means no liquidity, and the only way any exchange had ever broken that loop was by paying market makers to show up. Every exchange in crypto did it. Rebates, fee waivers, direct payments, token allocations.

So they came to Hyperliquid to collect. One of the market-making firms told them this exchange would never take off unless they were paid.

They did not pay them. They did not pay anyone.

What Yan did instead was the single most important decision in the company's history.

In a casino, the house takes the other side of every bet, and the house always keeps the edge. On an exchange, market makers are the house. Yan had spent two and a half years building trading strategies good enough to make him rich playing that role, so rather than renting the house out to a firm demanding payment for the privilege, he put his own strategies onchain and let anyone in the world become the house.

The vault was open to anyone. Ten dollars or ten million, no management fee, no performance cut, and every dollar of profit flowing back to the depositors. All of it ran on a public ledger, which meant that if you deposited ten dollars you could watch it work in real time, position by position, with nothing hidden anywhere.

Consider what that actually offered. A strategy of that quality inside a hedge fund charges two percent of your money annually plus half of everything it earns, and it will not take your call unless you have millions and an introduction. Here it was free, open to anyone with a wallet, and fully auditable. One early depositor said he would gladly have paid, and that people still do not understand how special it was.

Here is the part that made people fall in love with it.

Market makers do not earn money by being right about direction. They earn the spread, which means their best conditions are choppy, ranging, violent markets where prices move constantly without going anywhere in particular. Every stretch where crypto went sideways, and especially every liquidation cascade that ripped through the market, the vault printed for its users, which brought more depositors, which meant more liquidity, which brought more whales and more volume, which drew even more deposits.

Most of them were ordinary people with a few hundred or a few thousand dollars in, watching an algorithm work their capital in real time on a public ledger. Trending markets were harder on it, and depositors were loud about the flat stretches, which was fair enough. But for the first time in the history of markets, regular people were sitting on the house side of the best business model ever invented.

The oldest problem in the exchange business, solved without handing anyone leverage over the exchange, on an equal playing field.

Eventually the professionals came anyway, because the platform grew to the point where the economics brought them.

Getting paid to quote prices on an exchange with no volume is worth less than quoting for free on an exchange with real flow, and by late 2023 the flow on Hyperliquid was real. Firms that had demanded payment started integrating without it, one at a time, because the spread on genuine volume beat the rebate on an empty book. The kingmakers turned out to be replaceable after all.

What those firms found inside the exchange was a detail only a trader would notice. Yan had built a mechanism that made it harder for the very fastest firms to pick off slower market makers, which let everyone else quote deeper liquidity without needing to win a latency arms race. The industry has since copied it. It also cost Hyperliquid money, because a meaningful share of exchange volume comes from those firms sniping each other, and he had knowingly given up that revenue in exchange for better prices for ordinary users.

Again, users first. You can start to see where this is going.

Then the venture capitalists arrived.

They came bottom-up, which almost never happens. Analysts at the funds had been using the exchange quietly on their own time, and one by one they went to their partners and said this thing is actually good. By January 2024, funds were flying in. Yan and his co-founder had done no outreach and had no pitch deck, so when investors asked to see one, the two of them would talk until it became clear that Jeff had no real interest in outside capital.

He set one condition. He would only look at a term sheet valuing the company at a billion dollars, less than a year after launch, which should have ended the conversation.

An investor met the number. Somewhere around $100 million at a $1 billion valuation, while Jeff was burning hundreds of thousands of dollars a month out of his personal savings.

He took the weekend to think about it. He went to multiple founders who had raised money and to the investors themselves and asked them to explain what the point was. Nobody could give him an answer he found convincing.

On Monday morning he told his co-founder they were not taking it. She had spent two weeks preparing for precisely the opposite outcome while watching his money disappear, and her response was not printable. The rest of the team took it no better. He called the fund and declined, and they did not believe him either, assuming he must have accepted someone else's offer instead.

Four days later, Jeff tweeted:

Those four lines are the whole design, and they are worth slowing down on, because the next decision only makes sense once you understand them.

A normal company issues equity. Equity means early insiders and shareholders, and shareholders mean a board, a jurisdiction, an office that can be raided, and a person who can be pressured into flipping a switch. Every promise about neutrality survives exactly as long as the people holding the shares want it to. FTX had shareholders. So has every exchange that ever froze a withdrawal.

If you want a truly permissionless market that nobody owns and nobody can switch off, you cannot have shareholders. So instead of selling equity to investors, you create a token and give it to the people who actually use the thing. Then you route the platform's revenue into buying that token back on the open market and burning it. Every dollar the exchange earns permanently removes supply. Ownership sits with the users rather than the funds, the upside accrues to whoever holds it, and there is no cap table above them and no preferred shares. Everyone is equal.

The token is not a marketing exercise. It is the mechanism that makes the neutrality real.

On a Friday at the end of November 2024, they launched Hyperliquid's native token and gave away 31% of the entire supply to roughly 94,000 users. No conditions, no vesting, no lockups. If you had used the platform, you woke up that morning with tokens airdropped to your wallet.

It was worth over $1 billion at the opening price and roughly $16 billion at the highs. The largest wealth transfer in the history of crypto, and every dollar of it went to users.

The venture funds received nothing, ever. Any fund that wanted exposure had to buy on the open market at the same price as everyone else, on Hyperliquid, because the token was not listed anywhere else on earth.

People posted screenshots that morning of six and seven figure fully liquid balances that had not existed when they went to sleep. One wrote that it was enough to sort himself out for years, help his family, and invest through the rest of the cycle. Some traders retired simply by having used the product early.

The numbers started compounding in a way that made the earlier decisions look less eccentric.

Perpetual futures are the center of gravity in crypto, and it turns out perps are the most liquid instrument for price discovery. The instrument traces back to Robert Shiller in the 1990s, who pointed out that almost nobody trading a pork belly future actually wants pork bellies, so forcing the contract to expire serves very little purpose. Traditional exchanges saw no reason to change. Crypto built it in 2016, and it went on to account for roughly 75% of all trading in the asset class and more than 90% of derivatives activity, running six to eight times larger than spot markets at around $7 trillion a month.

Nearly all of it went through centralized exchanges. The total crypto derivatives market ran approximately $85.7 trillion in volume in 2025, Binance alone did $25.1 trillion of it, and the top four venues controlled over 60% of everything.

In 2022, decentralized exchanges held essentially none of that market. By mid 2026 they held around 16% of global perps volume, and Hyperliquid held roughly 59% of the open interest across every decentralized venue combined. Its monthly volume runs near 17% of Binance's and its open interest near 21%, numbers that would have been dismissed as impossible for an onchain protocol two years earlier.

The reason shows up in the order book. Blockworks measured Hyperliquid's Bitcoin perpetual depth against Binance's and found it tighter near the middle of the market. At one basis point from mid price, Hyperliquid showed around $3.1 million of depth against Binance's $2.3 million. At two basis points the gap widened to $5.8 million against $4.1 million.

The most liquid derivatives exchange ever built is losing the liquidity comparison to a three year old protocol that never paid a market maker a cent.

Hyperliquid had become the primary venue for cryptocurrency price discovery.

Last year the company generated over $1.022 billion in profit. It had eleven employees.

That is roughly $93 million of profit per person, and it helps to walk up the ladder to see how absurd the number is.

Nvidia, one of the most valuable companies in the world, produces around $2 million of profit per employee. That is the highest figure among the global megacaps, ahead of Apple, Meta and Alphabet.

OnlyFans is the business usually cited as the most efficient on earth, running a $7 billion platform with 46 people and clearing somewhere near $15 million of profit per employee.

Hyperliquid does roughly 6x that, and about 46x Nvidia. Not the most efficient company in crypto. The most efficient company anywhere, by a margin wide enough that the comparison stops being useful.

The team is between 24 and 31 years old, almost all of them extremely capable introverts, and Yan is the only one working under his real name. He recruits off the podiums of international math and science olympiads, the same circuit he came through. One engineer holds a silver medal in informatics. Another trained with the US national team. Only two people on the team, including Yan, had worked in crypto before Hyperliquid, which is partly deliberate, since the people drawn to crypto early were mostly there to get rich quickly and he was hiring for a decade.

They hold exactly one meeting a day. Everyone crowds around a single engineer's screen for a morning standup, with a stuffed dragon sitting on top of the monitor, and the conversation is mostly about what could break. For long stretches it stops being a conversation at all. Yan crosses his arms, lowers his head, and stares at his bare feet, and the engineer beside him does the same. The silences run long and nobody in the room finds them strange.

They eat lunch together every day, family style, around a black wooden table. On Thursdays they eat Chipotle, which does not exist in Singapore, so they gave the recipes to their chef.

The office contains almost no evidence that anyone in it is wealthy. The furniture came with the lease. Their additions were two board games, some NFTs on the walls, and thirty-four stuffed animals, mostly the work of one engineer whose wife will not let him bring any more home. Twelve cats sit on a cabinet, alongside sharks, lizards, koalas, penguins and several dragons draped over the monitors. Three bottles of vodka and whisky have sat untouched in the kitchen since a community event two years ago failed to hit its minimum spend, because the team drinks tea.

Their cleaner believes she works for a merchandising company that makes stuffed cats. Nobody has corrected her.

Yan owns fifteen pairs of the same shorts and ten of the same shirts in three colors, and wears a version of the same outfit every day. He cuts his own hair, because going to a barber costs time.

It is important to understand that Hyperliquid is not just an exchange. It is an infrastructure protocol that lets anyone build on top of its blockchain and launch trading interfaces on top of Hyperliquid markets. That lets the team focus on the technology while the surface area scales without them, compounding liquidity over time.

On October 13th, 2025, they shipped HIP-3, which allows anyone who stakes 500,000 HYPE, worth around $31.5 million by mid 2026, to deploy a perpetual futures market for any asset they choose, with control over the oracle, the leverage limits, and the risk parameters. Fees split evenly between the deployer and the protocol.

Before HIP-3, Hyperliquid listed crypto assets. After it, anyone could list anything with a price. Equities, commodities, indices, currencies.

The first real demonstration came on a Saturday in late February 2026, when the United States and Israel began bombing Iran and oil moved 30% over a weekend. Every traditional venue for trading crude was closed. The CME was dark, ICE was dark, and the price of the most important commodity on earth had just repriced violently with nowhere to express it.

Hyperliquid was open, because a blockchain does not observe weekends.

Crude volume on the platform went from around $21 million a day to billions. For a stretch that weekend, an onchain protocol built by eleven people was the primary venue for global oil price discovery. The Wall Street Journal wrote about it.

This is when TradFi started noticing.

The pattern repeated across asset classes. During the Cerebras listing on Nasdaq, a Morgan Stanley trader was photographed on the floor of the New York Stock Exchange with Hyperliquid's pre-IPO Cerebras chart open on his screen. When SpaceX listed, the pre-IPO market became the second most traded asset on the entire platform and did over a billion dollars in volume, and new wallet creation jumped tenfold in a day. One minute before the $SPCX listing, the perp traded at $171 per share, exactly the IPO's opening price. By January, a third-party silver market was doing roughly 2% of the CME's 24-hour volume, an exchange founded in 1898. In March 2026, S&P Dow Jones Indices formally licensed the S&P 500 to Trade[XYZ], the most active HIP-3 deployer, making it the first officially licensed S&P 500 perpetual contract in existence.

The aggregate numbers moved accordingly. HIP-3 open interest went from zero to over $2.9 billion in six months. Fewer than half of the top thirty markets on Trade[XYZ] are crypto markets now, with the rest split across commodities, equities and indices. On peak days, markets that did not exist a year ago account for more than half of all trading activity on the platform, and they have nothing to do with crypto.

Yan did not build those markets. That is the entire point, and it is the part most people miss when they try to value this.

His answer is that an accessible financial system is the better outcome, one running on public rails that no single company owns. Which means the work is not adding features. It is making it possible for other people to come and build businesses on top of him and keep half the revenue.

The next expansion is launching now. HIP-4 brings permissionless prediction markets and options onto the same engine. Prediction markets currently do around $21 billion a month globally. Crypto options clear over $180 billion in monthly volume at peak, and Coinbase paid $2.9 billion for Deribit to buy a position in that market. Putting both inside Hyperliquid means they cross-margin against existing perpetual positions, so a trader holding long Bitcoin or NVDA can hedge into an interest rate decision without moving a dollar off the platform. No standalone prediction market or options venue can offer that.

The reason any of this matters for valuation is the size of the pool it opens into. Those crypto derivatives that ran $85.7 trillion last year sound enormous until you set them beside over-the-counter foreign exchange and interest rate derivatives, which annualize somewhere near $3.7 quadrillion.

Crypto is about 2% of that market.

Asked directly whether he actually believes Hyperliquid will house all of finance, Yan smiled and conceded the word all was doing some work.

Over the last twelve months the protocol generated approximately $1.022 billion in earnings.

Essentially all of it goes back to holders. Ninety-nine percent of revenue flows into a mechanism called the Assistance Fund, which buys HYPE on the open market and burns it. By June 2026 the fund had bought and burned over 45 million tokens worth roughly $2.8 billion. The bid underneath the token is mechanically tied to usage, so volume produces fees, fees produce buybacks, and none of it requires anyone's discretion.

The revenue curve tells you how young this is. Monthly revenue was around $10 million in December 2024, peaked above $120 million in August 2025, and normalized near $55 million by December. That entire climb happened before HIP-3 contributed anything meaningful.

Two things sit on top of that base and neither is in most models.

The platform currently runs HIP-3 markets in growth mode, subsidizing the take rate by roughly 90% to attract liquidity into new markets. That subsidy comes down as markets mature, which is margin expansion requiring no additional volume at all.

And in May 2026, Coinbase became the official USDC treasury deployer on Hyperliquid, capturing around a 90% revenue share on float. Against approximately $6.13 billion of USDC collateral on the platform at a 3.65% gross treasury yield, that implies somewhere near $200 million of annualized revenue arriving on top of trading fees. A year earlier this had been a community bake-off between smaller stablecoin issuers. Now the terms are negotiated directly with Coinbase.

The institutional wrapper arrived in May 2026, when three HYPE ETFs launched: 21Shares on Nasdaq, Bitwise on the NYSE, and a Grayscale staking product. They absorbed over 1% of total HYPE supply within ten trading days, the strongest debut of any spot crypto ETF on record.

The valuation work, using Multicoin's framework:

Current price: approximately $55.

The base case is deliberately unambitious. It assumes derivatives market growth slower than the 44.8% CAGR the sector actually delivered from 2020 to 2025, zero further market share gains for Hyperliquid inside DeFi, and no contribution whatsoever from HIP-4, HyperEVM gas, or the Coinbase stablecoin arrangement. All three of those are live or in alpha.

On the multiple, Coinbase trades around 24 times 2028 consensus earnings and Robinhood around 32 times. Applying 20 times to a business growing at this rate, with eleven employees and 99% of revenue flowing to token holders, is closer to a floor than a stretch.

If you got something out of this, the most useful thing you can do is send it to one person who would get something out of it too.

And genuinely, reply to me. Hitting reply on this email comes straight to my inbox, I read everything and I answer everything. If you're reading this online, leave a comment instead and I'll be in there.

One thing I actually want to know, and it will shape what gets written next. What do you want more of?

Memory and semis, where most of the recent work has been. Hyperscalers, and the large cap value setups hiding inside the AI buildout. Or more of this, deep dives on things almost nobody is covering properly yet.

Reply with one word and I'll write more of it.

This piece represents personal research and conviction, not financial advice. neym is an independent research publication. The author holds a position in HYPE. Biographical reporting on Jeff Yan draws on Dom Cooke's profile "Beyond the Sky" in Colossus Magazine, April 2026. Valuation figures sourced from Multicoin Capital's Hyperliquid analysis, June 2026, unless otherwise noted. Do your own research. Nothing here constitutes a solicitation to buy or sell any asset.