Who exactly runs Hyperliquid? governance audit

Places with monthly liquidation exceeding US$200 billion and accounting for approximately 70% of the chain's perpetual contracts are protected by 27 validators. Its foundation ran all validators at launch. Critics and defenders are using outdated data, so here is the audit report: the composition of the current set of validators, the actual power, and the real gaps.

Summary

Hyperliquid's verifier set has grown from 4 at launch to 16, 21, 24, and up to 27 in June. Registration requires no license and the verifiers with the largest pledged amount constitute the active set. This year, key numbers have changed: Foundation-run validators currently hold approximately 49.3% of pledged HYPE, with the remaining 50.7% distributed among 22 other operators, up from 81% in early 2025. The fiercest criticism-that the foundation can imprison verifiers at will-is inconsistent with the document, which states that the incarceration is triggered by peers, targets delays and reliability failures, and that there is no automatic forfeiture mechanism in the system. The real gap lies in size, not malice: compared with the approximately 1800 in Solana and the hundreds of thousands in Ethereum, 27 validators guard a venue with a transaction volume of more than $200 billion a month, and its node software is still closed., and a plan is commissioned to verify the identity of participants.

Singapore regulators added Hyperliquid to their investor warning list in June, transforming the decentralized debate from a philosophical debate into an issue with legal consequences.

Body

The most valuable part of the decentralization debate is often the data it forces to make public, and versions of Hyperliquid have been running on outdated data for eighteen months. In January 2025, a node operator issued a letter stating that among the 16 validators, five foundation validators controlled more than 81% of pledged HYPE, and this number entered the discussion and never left. In June 2026, a well-known investor claimed that the network did not require a license at all, citing reasons such as the concentration of verifiers in a single building, the fact that node software remained closed, and the foundation could jail operators and force upgrades. Both interventions were considered final. But neither reflects the current state of the network-at this time the network has expanded to 27 validators, and the foundation runs nodes holding just under half of the pledge-and neither touches on the description of disputed rights in the agreement's own documentation. At the same time, the object of contention itself continues to grow: a venue that handles more than $200 billion a month, accounting for about 70% of decentralized perpetual contract transactions, has an order book, matching engine and clearing system all running on these 27 machines, which generate approximately $1 billion in fees every year. This article is the audit report that has been missing from the debate: the current set of verifiers, the documented powers, the precedents for these powers actually working, and the gaps that still exist after each amendment.

Verifier set, count one by one

Start with trajectory, because direction is the missing part of existing criticism. Hyperliquid started with only a few validators, all run by the foundation, which is essentially a licensing network with public token names hanging. In January 2025, the validator set expanded to 16, and that was when the initial decentralized letters and 81% of concentration data were generated. In April 2025, the foundation reorganized the registration itself: the collection became 21 nodes, the registration was open to everyone, and the top 21 sorted by pledge amount formed an active collection, which transformed the identity of the verifier from appointment to auction. The growth continues, reaching 24 by June 2026 and eventually 27. The pledge amount at the entry threshold has exceeded 1 million HYPE. This figure itself constitutes the real entry price of the network. Concentration data also changed accordingly. After a round of re-commissions from foundation validators in June, the nodes run by the foundation held approximately 49.3% of the pledged HYPE, and approximately 50.7% was distributed among 22 independent operators. The foundation runs 5 of the 27 validators. This is fundamentally different from the network described by the 81% concentration still circulating in the criticism, and any honest audit must first acknowledge the improvement before listing the remaining problems.

On April 7, 2026, Bitget CEO Gracy Chen announced that Hyperliquid is not a real DEX, but a centralized exchange, citing the JELLY incident.

The underlying mechanisms deserve precision because they define who can participate. The consensus mechanism is the certificate of trust rights: the verifier needs to lock in at least 10,000 HYPEs for a year as self-trust. The consignor faces a one-day lock-up period and a seven-day de-pledge queuing period. The rewards continue to accumulate and are automatically re-invested. There is no automatic forfeiture mechanism in the system, which is unusual and has both advantages and disadvantages: operators will not lose pledges due to mistakes, nor will they lose pledges due to misconduct, so the cancellation of the pledge queue period and social consequences constitute the execution layer. Governance runs based on delegated pledge weights, verifiers state their positions, and the results are determined by the tokens behind them, not the number of verifiers, which means that concentration data is governance data, not trivial details.

Review the three powers one by one.

Now let's look at the specific accusations and analyze them one by one according to the document, because two of them are true and one is not true.

Imprisonment. The most widely circulated theory is that foundations can imprison verifiers and remove them from active collections for any reason. The protocol document describes different scenarios: Verifiers can be jailed for delays and reliability failures through peer voting, and imprisoned verifiers stop generating rewards for their clients until they are released from imprisonment and no forfeiture is involved. Peer-triggered removals due to performance issues are standard practice for the Proof-of-Stake Network and not the discretion of the foundation. The remaining real concern is narrower but more realistic than the accusation: when nodes associated with the foundation hold nearly half of the pledge, peer voting weighted by pledge weights is not completely independent of the foundation, so the fairness of the mechanism depends on its underlying distribution. This is a debate about concentration, an argument repeatedly mentioned in this article, rather than a debate about arbitrary power.

Force upgrade. The statement that verifiers must use protocol upgrades is largely accurate and largely unsurprising. Every chain running a single client implementation faces the same reality: nodes that refuse upgrades will break away from consensus, which is a coordination fact, not a governance power. Here, the single binary architecture makes the problem even more acute. Hyperliquid runs a single implementation, and the foundation's defense of this by pointing out that Solana has been running the same way for years. This defense is honest but incomplete: a single client network concentrates risk-mistakes or decisions in one codebase can become mistakes or decisions for the entire network, which is why Ethereum views client diversity as a security attribute rather than an efficiency issue.

Closed source. This accusation is true and is the most critical of the three. Node software remains closed, and the foundation's stance since early 2025 has been to open source when code is stable, citing reasons including development speed and security. Eighteen months later, after significant growth, this promise has not been fulfilled, which was the core of the June criticism: that verifiers running unreadable binaries are trusting the author, and no degree of pledge distribution can make up for this. Users can verify status on the chain, but no one outside the team can independently verify software behavior before it produces status. For a venue that liquidates $200 billion a month, this is the widest gap between online claims and external verifiability.

You may also be interested in: What are ancillary assets? Words that determine the fate of cryptocurrencies.

Precedents: When power comes into play

Governance debates are always abstract until events materialize them, and the Hyperliquid incident arrived in March 2025 with a memin called JELLY. A trader opened a large position and manipulated the weak spot market underneath it, causing losses to fall into the agreement liquidity vault, a pool of funds that absorbs liquidation positions on behalf of depositors. When the treasury faced eight-figure losses, the validators voted to remove the market and settle at a price favorable to the agreement, and the losses were controlled. The intervention worked, the user was protected, and the incident ended within hours.

On June 18, 2026, CZ claimed that Hyperliquid's invention was excellent and filled the gap in Binance. He emphasized his non-KYC model while questioning his decentralization statement.

This also answers governance questions empirically. When the network's own capital is at risk, a trading market on the network can be closed through pledge weighted voting, and the voting at that time is conducted through a collection of validators, and the foundation holds the decisive share. This is why the event was described in the industry media as a "validator put option": an implicit guarantee that the platform will intervene when the platform itself loses money. There are two interpretations, and both are reasonable. A tolerant interpretation is that any exchange, decentralized or not, must be able to prevent manipulation, and a place where coffers are depleted by apparent attacks deserves criticism. The interpretation of intolerance is that decentralization is tested only when the intervention becomes attractive, and Hyperliquid intervenes. What this incident determines is not the quality of the network, but its nature: a place with functional emergency brakes and a small number of operators. Traders should price both aspects accordingly, because the same brake that protects vault depositors in March 2025 could also close a market where traders are making profits.

Comparisons that persist after each correction

After removing outdated data and exaggerated claims, a gap remains that no redelegation can fix: the set of validators is very small relative to the value of its protection. 27 validators, compared to about 1800 in Solana, hundreds in Cosmos Hub and hundreds of thousands in Ethereum. The technical objection is reasonable and deserves to be accurately stated: the Byzantine fault-tolerant consensus does not require thousands of participants to ensure security, it requires an honest super majority within existing collections, and a high-performance small collection is the reason why the network achieves sub-second finality and makes on-chain orderbooks feasible. Hyperliquid's entire product advantage-speed and finality enough to compete with centralized venues-comes at the expense of the size of the validator pool. This was a deliberate trade-off, not an oversight. The question is whether the price is right at this scale, and the calculations are troubling. Twenty-seven validators protect venues with more than $200 billion in monthly trading volume, as well as open interest contracts, treasury deposits, and markets now based on equity and other builder deployments. The important attack surface is not cryptography, but society and regulation: 27 operators means 27 phone calls, 27 subpoena jurisdictions, 27 pressuring relationships, and nearly half of the foundation's pledge volume means discussing a few people can determine most outcomes. Enlarged collections commissioned plans check the identity of participants, which improves accountability and also means that in practice, extensions are screened rather than open. Each of these facts is reasonable in its own right. But together they describe a network whose decentralization is best described as a managed trajectory: real, measurable, improving, but still a long way from the attributes implied by its marketing language.

Regulators are here

This is where the debate is no longer just philosophical. On June 26, the Monetary Authority of Singapore added Hyperliquid to its investor warning list, a register of entities that consumers may mistakenly believe are licensed. Inclusion on the list is not a ban, enforcement action or identification of misconduct, and Hyperliquid's response is accurate in every regard: it never claims regulatory authorization, there has been no change to the network, users retain self-custody rights, and settlements remain on-chain. Bybit also joined the same list nine days ago, and KuCoin February, Binance has been on the list since 2021, which puts Hyperliquid among familiar companies and suggests that regulators are handling it by list rather than specific agreements.

New News: The Singapore Monetary Authority has marked Hyperliquid HYPE as unlicensed and added it to the investor warning list.

The importance lies in the impact that inclusion has on vocabulary. "No license required" is a technical description internally of cryptocurrencies and is becoming a legal position externally, as agreements that claim to be the infrastructure rather than the operator make an argument as to who, if any, is responsible for the venue. The criticism that emerged on the same day-that a network with closed-source software, a screened collection of validators, and fund-weighted governance did not fit the description-was therefore more than just a pure argument. It claims the legal position is based on facts that the network has not yet been fully proven, and regulators reading the same debate will draw their own conclusions as to which entity, if any, is operating the exchange. This is the real stake in governance issues in 2026, and why numbers transcend ideology in this article: the distance between 49.3% and smaller, and the distance between closed source and open source, and the distance between credible infrastructure claims and controversial claims.

Listing rights and the funds behind them

One dimension of governance issues is completely independent of the validator debate and may be more important to traders: who decides what to trade here. The Internet's newer listing mechanism-which opens up the authority to create perpetual contracts and produces the developer deployment market for equity-related contracts independently audited here-is controlled by the amount of pledge rather than approval. Deploying the perpetual contract market requires the pledge of a large amount of HYPE for at least a period of time, while builder deployments on the EVM side are carried out through regular auction of seats. From one perspective, this is the most truly permission-free part of the system: there is no committee that determines which markets exist, only capital, which is why the venue is able to list synthetic equity exposures faster than any regulated exchange can convene a meeting. From another perspective, it replaces gatekeepers with wealth qualifications, meaning that the ever-expanding product surface of the venue, including markets that touch regulated asset classes, is determined by those who can provide collateral.

Economic mechanisms link the two parts of governance issues. Transaction fees flow into the repurchase mechanism of tokens, which this article has reported as the clearest example of the network in cryptocurrencies routing real income to its assets, while pledged HYPE is also a security margin, governance weight and listing key. This triple responsibility is both elegant design and centralized mechanism: the same token that protects the chain determines its rules and controls its listing content, so any accumulation of HYPE is an accumulation of all three powers. On a chain where about half of the pledged amount belongs to an affiliated group and more than one million tokens of tickets entering the verifier collection, the actual question is not whether the system does not require permission in principle, but how much capital is needed to have an impact. The answer rises as token prices rise.

This is a framework worth bringing out in this audit. Hyperliquid's governance is not a story about foundations refusing to let go; the trajectory shows the opposite, steadily and measurably. This is a story about design, in which influence tracks capital in an unusually direct way, and its software remains unreadable from the outside in a venue that is now larger than most regulated exchanges. Whether this is acceptable or not is up to each user to judge. The exact circumstances have now been recorded.

June 19, News: Hyperliquid will expand the set of active validators from 24 to 27 in about a month.

Points of Interest

Pledge distribution, not number of verifiers. Quantity is the indicator that is easy to grow and the least informative. Whether the amount of pledges run by foundations continues to fall below 49.3%, and whether any independent operators accumulate obstructive positions, are the measures that determine who really determines the outcome.

Open source commitment. The promise to release node software has not been fulfilled since early 2025, and this is the single change that most changes the audit results. Its persistent lack is itself information, and the longer it goes, the weaker the reasons for stability.

Next intervention. The JELLY incident shows that the network will take action to protect its coffers. The next such event, and whether decisions are made through a pledge distribution that no longer has majority support from the foundation, is a test of whether governance changes or just arithmetic changes.

Regulatory follow-up. Singapore's list currently has no operational impact. Whether other jurisdictions follow up, and whether any jurisdiction treats the foundation as an operator of an unlicensed exchange, will be scenarios where every fact in this audit ceases to be a point of debate and becomes evidence.

Disclaimer

This article is for informational and educational purposes only and does not constitute financial or investment advice. The number of verifiers, pledge distribution and protocol parameters continue to change, and the data reflects the report status at the time of writing. This document does not constitute a recommendation to buy, sell, hold or trade any asset or venue. Please always research on your own. The information is accurate as of July 26, 2026.

Frequently Asked Questions

How many validators does Hyperliquid have? As of June 2026, there are 27, from four or five at launch to 16 at the beginning of 2025, 21 in April 2025, and 24 later in the year. Registration is open to anyone, the verifier with the largest amount of pledges constitutes an active collection, and the entry threshold requires a pledge of about 1 million HYPE or more. Verifiers must self-commission at least 10,000 HYPEs and lock them in for one year.

Who controls the amount of pledge? After re-commissioning in June, the validators run by the foundation held approximately 49.3% of the pledged HYPE, with approximately 50.7% distributed among 22 independent operators. The foundation operates 5 of the 27 validators. This is a substantial change from early 2025, when a widely cited analysis showed that foundations controlled more than 81% of pledges among 16 validators.

Can the foundation remove verifiers at will? According to the documentation, no. Imprisonment is described as being triggered by peers. In response to delays and reliability failures, imprisoned validators stop earning rewards until incarceration is lifted. There is no automatic forfeiture mechanism in the system. The legitimate concern is indirect: because peer voting is weighted by the amount of pledge, and foundation related nodes hold nearly half of the amount, the independence of the mechanism is limited by the same concentration issues generally as governance.

Is Hyperliquid's code open source? Node software remains closed, and the foundation has said since early 2025 that it will open the code once development is stable, citing reasons including security and release speed. That promise remains unfulfilled, which is the most substantial of the existing criticisms: the validator is running binary files that cannot be audited, and no pledge distribution can make up for this.

What was the JELLY Incident? In March 2025, a trader manipulated a thinly traded mein market, pushing losses into the agreement liquidity vault. The validators voted to remove the market and settle it at a price to protect the treasury, controlling the eight-figure loss. The intervention worked and was also seen as evidence of "verifier puts", that is, networks would act when their own capital was at risk, and this was done through the fund-dominated pledge distribution at the time.

How does the number of verifiers compare to other chains? It is much smaller: Solana has about 1800 validators, Cosmos Hub has hundreds, Ethereum has hundreds of thousands, and Hyperliquid has 27. The Byzantine fault-tolerant consensus does not require large collections to ensure security. Small collections provide the sub-second finality required for on-chain order books, but also concentrate social, regulatory and coordination risks for a venue that handles more than $200 billion a month.

What does the Singapore list mean? On June 26, the Monetary Authority of Singapore added Hyperliquid to its investor warning list, a register of entities that consumers may mistakenly believe are licensed. This is not a ban or enforcement action, Bybit, KuCoin and Binance are also on the same list. Its importance is that it moves the "no license requirement" issue from the technical debate into the legal realm, because the assertion that it is infrastructure rather than operators depends on whether the governance facts are as stated in the agreement.

What should traders learn from this? The network has a functional emergency brake and a limited number of operators, an attribute that requires pricing rather than a scandal that requires condemnation. Decentralization here is a managed trajectory: pledge distribution improves measurably, source code issues remain unresolved, and small relative to the value at risk. Position sizing in any location should reflect governance realities rather than marketing vocabulary. This is educational analysis, not investment advice.

Read more: Are cryptocurrency tokens overpriced when equity holds real profits?