Last winter, Unitree released a warehouse video I could not stop replaying. A fleet of humanoid robots moved through a logistics center with the choreography of a stage director's obsession: no wires, no handlers, no hesitation. Shoulders and hips built to mirror my own, rotating through tasks that would have consumed a human crew eleven shifts. The caption celebrated efficiency. The subtext was far more unsettling.
Somewhere between those frames and today, those machines were converted into a financial instrument. Unitree Robotics is seeking $904 million in an initial public offering, a listing that would make it the first publicly traded humanoid robot maker in history. The number is impressive. The governance gap it exposes is far larger.
Because I spent 2020 inside a different kind of collective experiment โ co-designing the governance structure for UnityDAO, a $5 million treasury managed by 3,000 members across nineteen time zones โ I developed a habit. I read every organizational formation the way an auditor reads a balance sheet: not for the numbers, but for the unseen promissory notes. This IPO is full of them.
What happens when a human-shaped machine becomes a shareholder's asset? What happens to the data those eyes collect? And who, exactly, gets a vote when a robot begins walking through your neighborhood, your factory, your hospital corridor? Nobody has answered these questions. The prospectus won't either.
This is the story of the space between the ticker symbol and the walking machine โ and why the crypto industry, for all of its flaws, has spent a decade building governance tools for precisely the moment a company like Unitree arrives.
Context: The Rise of the Walking Asset
Unitree began, as these stories often do, with a student and a stubborn belief. Wang Xingxing founded the company in 2016 in Hangzhou, emerging from a Ph.D. program to build machines that could move with an agility the academic literature had only modeled on screens. The Go1, launched in 2021, was a quadruped that anyone could buy for the price of a good drone. The B2 handled industrial inspection. Then came the H1 in 2023, a full humanoid capable of running, and the G1 in 2024, a sleek, articulated figure that undercut the humanoid market on price at sixteen thousand dollars. By late 2025, the company had moved from viral videos to factory floors, deploying robots in logistics and light manufacturing across Zhejiang and Guangdong provinces.
The IPO changes the stakes. At roughly $904 million raised, it would make Unitree the first listed humanoid robot maker on any major exchange. The company filed on Shanghai's STAR Market, the technology board designed to channel national savings into strategic industries. Context matters here. The listing does not land in a vacuum; it lands inside a policy ecosystem. China's industrial strategy has elevated humanoid robotics to a national priority โ a cornerstone of what planners call "new quality productive forces." Government targets for robot density, subsidies for automation adoption, and a coordinated push across supplier networks have created conditions for a company like Unitree to scale faster than almost any Western counterpart.
But here is the context that crypto people should care about. Humanoid robots are not just hardware. They are the most data-dense physical assets ever attached to a corporate balance sheet. Each G1 unit carries depth-sensing cameras, microphones, inertial measurement units, and enough onboard computing to process the world in real time. Deployed across a warehouse, a hospital ward, or a residential complex, one fleet of humanoids produces behavioral data โ movement patterns, human interactions, ergonomic responses โ that rivals the output of an entire social platform. And that data is captured by a company that answers to shareholders before it answers to anyone else.
The market projections are seductive. Forecasters place the humanoid robotics sector in the tens of billions of dollars by the late 2030s. Analysts compare the moment to the early days of autonomous vehicles or, more piquantly for my readers, to the genesis of Ethereum: a technological inflection point where the underlying architecture is still being decided. The difference is that Ethereum made its architecture open, auditable, and community-governed from day one. The humanoid robotics industry is making a different choice, and the IPO is the most visible symptom of that choice.
Core: A Governance Audit
I am going to do something slightly unusual here. Rather than reviewing Unitree's financials โ which, as an outsider, I can only see through the prism of media reports and the company's own disclosures โ I will treat this IPO as a governance audit. I have spent twenty-seven years watching organizations convert claims into power, and more than a decade building tools to make that conversion transparent. From my work auditing AI-generated content in DAO discussions under the Human-First Protocols initiative, I learned that the most dangerous systems are not the ones that lie to you. They are the ones that tell you the truth about everything except the one thing that matters.
The First Thing That Matters: Ownership of Decisions
A humanoid robot is not a toaster. It will carry your elderly parent across a room. It will administer medication in a memory-care facility. It will sort packages, weld chassis, harvest crops, and, eventually, it will perform tasks that require it to make a judgment call between two imperfect options: stop and call a human, or complete the action and accept a risk. That judgment call is a decision. And the architecture of the IPO determines who owns that decision.
In a traditional listing, control concentrates in the hands of the largest shareholders. The founder holds a class of shares with outsized voting rights. Institutional investors hold another class with meaningful influence. Retail investors โ the millions of individuals who will buy Unitree's stock on the STAR Market โ hold what is effectively a lottery ticket on the company's trajectory. They have no say in the safety parameters of the robots. They have no say in the data policies. They have no say in which communities the robots are deployed into. They have a vote on board composition and executive compensation, and in any system where turnout hovers below five percent, even that vote is a formality. The on-chain governance industry has spent years confronting this problem in token systems. The lessons apply verbatim to the shareholder register of a robotics maker.
Consider my experience with UnityDAO. When we designed the governance structure, we did not start with tokenomics. We started with a question: who will be harmed if this system fails? The answer was not just the token holders. It was the communities in which the DAO operated, the contractors who built for it, the users who trusted it. So we implemented quadratic voting to prevent whale dominance, and we ran forty-two monthly community calls to build what I can only describe as social cohesion โ the messy, slow, human process of aligning 3,000 people who disagree on everything except the belief that the project mattered. Industry average participation in governance proposals sat below five percent. Our participation rose to more than triple that, not because we made voting easier, but because we made the stakes legible.
Unitree's IPO makes the stakes anything but legible. The human beings who will live alongside these robots โ who will share sidewalks with them, work in facilities that deploy them, and rely on them in care settings โ are not being offered a seat at any table. They are not being offered a governance token. They are not being offered a vote on safety thresholds. They are being offered, at best, the opportunity to buy shares in the company that controls the machines entering their lives. This is the corporate governance version of a factory farm: the party with the most at stake in the outcome has the least say in how the entity behaves. And I say that as someone who has participated in a wide range of governance experiments, from grassroots education initiatives in Chicago to institutional negotiations with BlackRock's venture arm.
The irony is that the tools for distributed decision-making already exist. Quadratic voting, conviction voting, and delegated representation are not theoretical constructs; they are battle-tested mechanisms that DAOs have refined over half a decade of real treasury management. Could Unitree have structured its offering with a stakeholder council empowered to veto deployment decisions in sensitive environments? Technically, yes. Could it have minted a governance token for the communities that host its robots, giving them a voice in behavioral parameters? Legally difficult, but not impossible. The refusal is not a matter of capability. It is a matter of intent. The prospectus will not include these mechanisms because the entire point of the corporate form is to aggregate control in a single, efficient, accountable-to-capital entity. The humanoid robotics industry is making a different choice, and the IPO is the most visible symptom of that choice.
The Second Thing That Matters: Data Architecture
Every humanoid robot is a surveillance device. The cameras, microphones, and depth sensors that enable the robot to navigate a crowded space are also recording that space. The same vision model that helps a G1 avoid a collision with a child in a corridor can generate a permanent behavioral record of that child's movement pattern. The same microphone array that allows a robot to respond to voice commands can capture ambient conversations in a home. None of this is secret. It is written into the functional requirements of the hardware. And yet, almost none of it is governed by anything resembling a transparent consent framework.
The crypto industry understands this failure mode intimately. For three years, the concept of Soulbound Tokens โ non-transferable credentials bound to a specific identity โ has been a subject of endless white papers and conference panels. The reason the concept remains stalled is not technical. It is that no honest institution actually wants a permanent, verifiable record of human behavior on a shared ledger. People do not want their credit history, their medical interactions, or their employment record inscribed in an immutable format that they cannot contextualize, appeal, or age out of. And the people who would issue those records โ the companies, the state institutions, the platforms โ do not want to surrender the interpretive power that comes from keeping the records scattered and unverifiable.
Now transfer that tension to humanoid robotics. Unitree's robots will generate behavioral data at a scale and intimacy that makes a search engine's clickstream look like a diary. Whose property is that data? The shareholder's, under the current arrangement. The corporation that deploys the fleet. And, eventually, the state that regulates the corporation. The individuals whose homes, bodies, and conversations are being observed have no property right in the record of their own lives. They do not even have a seat at the table to debate what gets recorded, what gets retained, what gets shared with third parties, and what gets sold to insurers, employers, or advertisers. The IPO does not create this problem. But it does something worse: it formalizes the problem, converting a residual ethical concern into an asset class. The data from humanoid robots becomes a growth narrative in the prospectus, described in the polite language of "proprietary data moats" and "network effects," while the human beings generating the data are reduced to environmental inputs.
There is a parallel here to the stablecoin market that I have written about for years. Tether dominates over seventy percent of the stablecoin market, and its reserves have never received a truly independent audit. The entire ecosystem pretends this problem does not exist, because the alternative โ admitting that the largest dollar-pegged asset in crypto rests on unaudited claims โ would be too disruptive to price in. The humanoid robotics industry is building the same house of cards, just in the physical world. The safety claims, the data policies, the behavioral models โ none of these have been independently audited against real-world outcomes. We are being asked to trust Unitree's internal assertions about how its robots will behave in unstructured environments, just as we have been asked to trust Tether's internal assertions about how its reserves behave in stressed markets. In both cases, the incentives reward optimism over verification.
But I want to go further than the stablecoin parallel, because embodied data has a quality that reserves do not: it is generated by the bodies of people who never consented to the observation. A Tether dollar does not record your heartbeat. A robot in your home does. The data governance question becomes a bodily autonomy question, which is why it will not be resolved by a privacy policy buried in an app update. It requires structural accountability.
The Third Thing That Matters: Verification
I have audited AI-generated content in DAO discussions as part of the Human-First Protocols initiative. The experience taught me a specific, uncomfortable lesson: verification is not a technical problem. It is a philosophical one. When you build a manual verification layer for one thousand key governance proposals, you discover that the hardest part is not detecting whether a piece of text was generated by a language model. The hardest part is deciding what level of evidence is sufficient to justify a decision that will affect real people. Ten percent confidence is not enough to block a proposal. Ninety-five percent confidence may be enough to override a human's judgment. Where is the line? The line, in my experience, is a moral choice, not a statistical one.
Embodied AI pushes this problem into dramatic relief. A text-generation model that hallucinates a legal citation can be corrected before the damage is done, or at least before anyone signs a contract on the basis of the hallucination. A humanoid robot that misjudges the weight distribution of a patient during a transfer cannot be corrected in time. The cost of failure is physical. And yet, every indication suggests that the industry is approaching verification the way the crypto market approached reserve audits: as a compliance checkbox rather than a cultural commitment. Red-team testing, staged deployment, and safety documentation are not the same as independent, continuous, adversarial verification with public accountability for the results.
My own background in finance taught me to distrust good news that cannot be tripped up. In 2017, during the ICO boom, I launched a series of educational workshops in Chicago โ the Ethical Ledger โ to teach retail investors how to read smart contract safety and spot the philosophical signs of centralization in projects that waved the word "decentralized" like a talisman. Over a hundred and fifty participants came through those sessions. One project we flagged collapsed a few weeks later, and the participants avoided an estimated two hundred thousand dollars in collective losses. What made the difference was not access to superior information. It was a discipline of verification: asking what evidence would be required to falsify the claims, and refusing to invest in systems that could not produce that evidence on demand.
The humanoid robotics market is asking the world to make an investment of a different kind โ a physical investment in the safety of homes, workplaces, and public spaces โ without a comparable discipline of falsification. What would prove that Unitree's robots are safe in unstructured human environments? A randomized controlled trial with independently audited outcomes. What would prove that the data collection is proportionate? A published data impact assessment, signed by an independent auditor, with a mechanism for public challenge. What would prove that the governance architecture accounts for the people who live with the machines? A stakeholder engagement framework that gives those people something closer to the rights of shareholders. None of these exist. Not because they are impossible, but because the current governance structure makes them optional. And the IPO, by rewarding the company's growth narrative rather than its accountability architecture, tells the next cohort of robotics founders exactly which behavior is being priced.
The Fourth Thing That Matters: The Community Seat
There is a scene I return to often, from the aftermath of the FTX collapse in 2022. Chicago's crypto community was decimated. I spent that winter organizing Rebuild Chicago, a peer-support network for former employees and investors โ career counseling, emotional support, and eventually fifty thousand dollars raised from personal networks to fund legal aid for scam victims. The experience was not about profit. It was about preserving the human element of an industry that had been exposed as performing trust rather than practicing it. And the deepest lesson was this: communities do not forgive harm because compensation was paid. They forgive harm when they are given a genuine seat in the decision-making process that governs their future.
The communities that will host Unitree's robots โ the residential complexes, the hospital networks, the factory towns โ do not have a seat. They will not have a seat under the IPO structure. The shareholder meeting will decide that a fleet is deployed in a particular district. The regulator will approve it, or not. The human residents of that district will be informed, at best. They will not be consulted in a structured way. They will not have a mechanism to challenge the robot's behavioral parameters. They will not have a voting stake in whether the robot's cameras are active during non-operational hours. They will not be able to audit the data retention policies that determine how long their movements remain on a company server.
The crypto governance community has an answer to this problem, and I have spent years implementing it. Quadratic voting, conviction-based voting, and multi-stakeholder councils are not theoretical constructs. They are technologies for distributing legitimate authority among parties with unequal resources but equal stakes. In UnityDAO, quadratic voting prevented a small cluster of whales from dictating treasury policy, and the forty-two monthly community calls created the social infrastructure that made meaningful participation possible. The same toolkit can be applied to embodied AI governance. You can imagine a local robotics council with the power to approve deployment licenses, funded by an allocation from the company's revenues. You can imagine a community compensation mechanism for data generation, structured as a continuous, transparent flow of value rather than a one-time license fee. You can imagine a design where the residents of a district collectively hold bargaining power over the robots' behavioral boundaries โ what time they can operate, what spaces they can enter, what data they can record. None of this is technically difficult. All of it is institutionally radical.
And this is precisely why it will not appear in the prospectus. The IPO is not designed to distribute power. It is designed to concentrate it โ attractively, lawfully, and with the emotional comfort of a name-brand institution. This is the fundamental tension at the heart of the first humanoid robot listing: the more essential the technology becomes to public life, the more its governance architecture needs to be public. And the current architecture is private by design.
The Fifth Thing That Matters: The Human-in-the-Loop Test
During the Human-First Protocols work, I spent months training five hundred new DAO members to distinguish human intent from AI noise. The experience left me with an uncompromising belief: for every proposal that can be automated, there must be a human who can say no. We built a manual verification layer for one thousand key proposals, not because we doubted the efficiency of algorithms, but because we understood that legitimacy flows from the possibility of refusal. A system where no one can say no is not a system; it is a machine.
Humanoid robotics companies face this test in a far more urgent form. If a robot is designed to optimize for task completion, and the task completion conflicts with a human's need for dignity, privacy, or safety, who has the authority to stop the machine? The current answer, under the IPO governance structure, is the deploying corporation. The robot's parameters are set by engineers and approved by managers, all responding to quarterly earnings pressure. The human in the corridor โ the patient, the worker, the child โ has no lever. This is the mirror image of every governance failure I have seen in crypto. When a DAO's decision-making becomes fully automated, the community loses the capacity for judgment. When a robot's behavior becomes fully automated, the human loses the capacity for objection.
The human-in-the-loop is not a feature to be added later; it is the constitutive principle of legitimate embodied AI. An architecture that cannot accommodate human refusal is an architecture that will eventually cause harm that no apology can remedy. The question is whether Unitree's governance model โ and the regulatory framework that will grow up around the first humanoid robot IPO โ will embed this principle from the start, or retrofit it after the first tragedy.
The Sixth Thing That Matters: The Financialization of Risk
Every IPO is a machinery for distributing risk and reward across a new class of holders. The Unitree listing has a peculiar feature, though. The rewards go to shareholders. The risks โ physical, psychological, and social โ are distributed across the public. A child injured by a robot in a public square will not see any share of the upside from that robot's efficiency gains. The liability will be litigated, possibly compensated, but the asymmetry remains structural. This is not a novel observation; it is the classic externality problem. But humanoid robotics makes the externality visceral in a way that software platforms never managed. A social media algorithm harming mental health is diffuse. A robot harming a body is immediate, verifiable, and deeply humanizing in its horror.
The financialization of this risk matters for a practical reason: insurance and capital markets will eventually price it. Actuaries will demand data on failure rates, near-misses, and behavioral edge cases. That demand will collide with the industry's preference for opacity. And when it does, the governance vacuum I have been describing will become a financial crisis. The robotics industry will face the same reckoning the stablecoin industry avoided for years: the gap between claims and verifiable evidence will widen until the market forces a correction. The prudent move โ building the audit infrastructure now โ is also the most difficult move, because it requires the company to accept external oversight before the market demands it.
There is an opportunity here for the crypto ecosystem. The tools for transparent, auditable claims already exist. Decentralized oracle networks can aggregate independent sensor data. Immutable audit logs can record deployment history. Smart contract-based insurance pools can collateralize liability in a transparent way. But these tools will only be adopted if the governance architecture creates demand for them. The IPO, as currently structured, does not create that demand. It creates demand for growth, which rewards the most optimistic claims, not the most verified ones.
The Seventh Thing That Matters: The Standards Race
The first humanoid robot IPO does not just price a company. It sets a precedent for an entire industry that will follow. Every robotics startup seeking capital in the next decade will be measured against the Unitree template. If the template rewards growth over governance, the signal is clear: build fast, disclose little, and let the public absorb the risk. If the template โ through regulatory pressure or market feedback โ rewards transparent verification, stakeholder councils, and auditable data practices, then every serious competitor will adopt those features to compete. This is the lesson I carried into the Values First coalition in 2025, when we succeeded in conditioning a ten million dollar grant from BlackRock's venture arm on the adoption of our transparency protocols. Institutions respond to standards when the standards affect their access to capital.
China's regulatory framework will shape this race in ways that Western observers often misunderstand. The STAR Market listing requirement, the national robotics standards, and the industrial policy appetite for strategic technologies create a dense regulatory environment. It is tempting to dismiss this as simple top-down control. But I have learned, from years inside governance systems, that dense regulation is not the same as effective governance. What matters is whether the regulations create meaningful feedback loops between stakeholders and decision-makers. A deployment license process that consults local communities is governance. A national safety standard that is enforced through opaque certification is paperwork. The IPO gives the world a rare opportunity to see which kind of governance regime actually surrounds humanoid robotics.
The Contrarian Angle: The Case for Centralized Clarity
Now let me hold my own argument up to the light, because if I have learned anything from years of building governance systems, it is that purity is a luxury of the uninvolved. I spent 2025 leading the Values First coalition, uniting fifteen smaller DAOs to establish a unified charter for ethical institutional engagement. We ended up negotiating a ten million dollar grant allocation from BlackRock's venture arm, conditioned on their adoption of our transparency protocols. Some in the community called it capitulation. I called it leverage. The institutions were coming in regardless; the choice was between setting standards from within their orbit and watching standards be set without us.
There is a case to be made that Unitree's IPO is the same kind of moment. Humanoid robotics will be deployed in mass quantities whether or not the governance architecture is ready. The question is whether the governance architecture can be built in time to shape the deployment. A publicly listed company, for all of its flaws, is at least a legally accountable entity. It can be sued. It can be regulated. Its directors can be removed. Its safety record can be discovered through disclosure obligations. A decentralized robotics ecosystem, by contrast, offers no equivalent point of accountability. If a robot in a DAO-controlled network harms a person, who do they sue? There is no persona, no board, no registered address. The very decentralization that protects communities from capture also protects them from accountability.
I believe, genuinely, that the most realistic path to good governance for embodied AI runs through the mess of corporate law, not around it. The IPO creates a target. And a target โ visible, capitalizable, exposed to the reputational damage of a single horrific incident โ is something that accountability movements can actually hit. The market will demand quarterly disclosure. Regulators will impose safety standards as a condition of continued listing. Publicly traded humanoid robotics companies will be forced to answer questions that a thousand DAOs could avoid indefinitely. In that sense, Unitree's offering is less an abdication of decentralized values than a necessary detour through centralized clarity. The revolution will not be decentralized from day one. It will pass through the shareholder meeting first, and the people who understand both worlds will be there to write the rules of engagement.
There is another contrarian angle worth naming. China's industrial strategy, despite its centralized framing, is actually a form of coordinated pluralism. Multiple robotics makers โ Unitree, UBTech, Fourier, and a cascade of startups โ are competing for government contracts and market share simultaneously. This is not a single company's monolithic ecosystem; it is a managed market with competing architectures. The standards that emerge from this process will be the result of many actors jostling, not one committee dictating. It is closer to a permissioned blockchain: centralized settlement, but contested state transitions. Anyone who has worked in governance knows that contested state transitions, even within a permissioned framework, produce better outcomes than uncontested consensus. The IPO accelerates that contest by forcing Unitree to raise capital in public, disclosing its technology roadmap and financial health to competitors and regulators alike. Transparency through capital markets is still transparency.
And yet, I cannot fully convince myself that this detour ends in the right place. The centralized-clarity argument assumes that accountability will be exercised, that regulators will be competent, that shareholder pressure will produce safety rather than short-termism. The stablecoin experience undermines that assumption. Tether has been publicly listed on nothing, yet it has faced years of regulatory scrutiny without producing a genuinely independent audit. The mechanisms of accountability are available, but they are only effective when the culture demands their use. A public listing does not create that culture. It merely creates the possibility.
The Takeaway: Governance Before Deployment
The question I am left with is not whether Unitree's IPO will succeed. It will, in the narrow sense that matters to underwriters. The question is whether the human beings who will share this world with humanoid robots will have a governance stake in their behavior. Crypto has spent a decade building the tools for exactly this problem โ quadratic voting, multi-stakeholder governance, transparent data rights, human-in-the-loop architectures. The robots are coming whether or not we use those tools. The machines will not wait for our governance stack to reach feature parity with their deployment schedules.
I keep returning to a line I wrote after the pandemic, during the darkest weeks of the bear market: a ledger that only counts tokens will never balance what matters. The humanoid robotics industry is about to fill ledgers with billions of dollars of market cap. What it will not count is the trust of the communities that host its machines. That trust has to be earned through governance, not purchased through a share offering. Unitree is building something extraordinary. The robots are real. The ambition is real. The question is whether the architecture of accountability will be built with equal care โ before the first robot walks into your neighborhood, not after.
Code without compassion is cold. A robot without a voice in its own governance is a tool of whoever holds the largest share. Governance is not an interface; it is a relationship. We can build a different architecture. The question is whether we will build it in time, and for whom.
