The code whispered secrets the whitepaper buried.
This time the whitepaper is an S-1 filing. And the secret is that Unitree Robotics—the Chinese quadruped and humanoid maker that just filed for a $904 million initial public offering on Shanghai’s STAR Market—has more in common with a DeFi protocol’s token sale than with a conventional industrial stock debut. I have spent the last 25 years watching blockchain projects promise decentralization while delivering the opposite. Reading Unitree’s prospectus felt like a reunion: the same carefully curated metrics, the same narrative-heavy growth story, the same missing clauses where actual risk disclosure should be. The difference is that this time, the vehicle is a state-approved equity listing, not a token. The mechanics are identical. The intent is just better dressed.
If you think a humanoid robotics IPO is outside the crypto beat, you are already behind. The same institutional capital that rotated from crypto into AI now needs a physical chassis to park the story on. Unitree is that chassis. And the 200 pages of its prospectus are the ABI of a new financial contract—one that will determine not only whether robot stocks are the next asset class, but whether the decentralization lessons of the last crypto cycle get ignored all over again.
I do not say this to be contrarian. I say it because the function calls are visible. Let me walk you through them.
The Hook: A $904M Filing That Reads Like a Token Sale
On the surface, the headline is straightforward: Unitree Robotics, the Chinese company famous for the Go2 dog robot and the H1 humanoid that performs backflips on camera, has filed for a $904 million IPO. If approved, it will be the first publicly traded humanoid robot maker. Markets will call it a milestone for China’s industrial ambitions. Commentators will call it a bet on the physical AI revolution. My first reaction, after checking the numbers, was simpler: this is a token sale with extra steps.
Why? Because virtually every narrative element of the crypto bull market is present. A charismatic founder with an engineering background. A product that demos well on YouTube but lacks proven economic viability. A total addressable market figure that assumes exponential adoption without addressing unit economics. A funding round structure that gives early investors a tenfold paper return before any real revenue scale. And a “community” of fans—in this case, the robotics enthusiast crowd—that will cheer the listing as validation, just as crypto communities cheered every exchange listing as adoption.
Over the past seven days, I have tracked reactions to the Unitree filing across Chinese financial media, global robotics forums, and crypto Twitter. The pattern is familiar. The bulls quote the robot’s specs: 29 degrees of freedom, 3 meters per second walking speed, 360 Nm joint torque. The bears quote the financials: revenue of approximately 1.26 billion yuan in 2024, a net loss that narrowed but did not disappear, and a valuation that reportedly exceeds 100 times trailing earnings. No one quotes the prospectus’s risk section, because the risk section is written in the careful legalese of procedural compliance, not in the forensic language of actual engineering constraints.
The code whispered secrets the whitepaper buried. The same is true here. The secret is not that Unitree is a fraud. It almost certainly is not. The secret is that the entire “public robot company” category is being constructed on the same kind of optimistic projection that gave us algorithmic stablecoins, and the market is treating that optimism as guarantee because the robotics story is more emotionally palatable than a failed stablecoin.
The first IPO of any asset class never is the best one. The first token that listed on a major exchange was not the one that survived. The first humanoid robot IPO will not be the one that defines the industry. Unitree is the pioneer, but pioneers get the arrows. And the arrows here are substantial: competitive pressure, supply chain concentration, geopolitical export controls, and a domestic Chinese market where the economic case for humanoid robots has not yet been proven.
But let me not get ahead of myself. Before dissecting the flaws, I need to establish the context—because context is what most coverage of this IPO is missing.
The Context: What Unitree Actually Is
Unitree Robotics was founded in 2016 by Wang Xingxing, then a graduate student with a background in mechanical engineering. The company spent its first five years building quadruped robots—products that look like dogs and act like mobile sensor platforms. The Go1 and Go2 models became famous in consumer and industrial contexts. The B2 became a workhorse for inspection and logistics trials. These quadrupeds generated real, if modest, revenue and gave Unitree a manufacturing base that many software-first competitors lacked.
The humanoid pivot is more recent. The H1, announced in 2023, got global attention because it could walk and run and perform acrobatic movements. The G1, released in 2024, was an attempt at a lower-cost humanoid. The latest model, the R1, reportedly targets warehouse work. None of these humanoids is yet in mass production at a scale that would generate meaningful revenue. The true revenue story remains the quadrupeds, which are superb pieces of engineering but exist in a niche market. The entire humanoid segment, as of the filing date, is the “future growth” section of the prospectus—the section that every analyst looks at with suspicion but no one can prove wrong.
Unitree’s IPO is on the STAR Market, China’s equivalent of Nasdaq, which was designed to channel capital into technology companies deemed strategically important. The $904 million figure is not the final valuation; it is the amount the company seeks to raise, which implies a target valuation in the range of $8 to $10 billion. For context, the company had reportedly raised over a billion yuan from private investors across multiple rounds. Xiaomi, Alibaba, and several state-linked funds participated in earlier rounds. The prospectus indicates that the proceeds will fund mass production of humanoid robots, R&D for next-generation actuators and AI models, and working capital for supply chain expansion.
Nothing about this is scandalous on its face. It is a classic growth-tech IPO. But classic growth-tech IPOs do not usually come wrapped in the hype of “humanoid robotics” as a national strategic imperative. And that is where the crypto comparison gets uncomfortable.
Remember the DeFi summer of 2020. I was tracking a Uniswap arbitrage bot that had drained value from over 4,200 trades in three weeks. The bot did not steal anything in the legal sense. It simply executed the logic of the protocol faster than other users. The community called it a bug. The developers called it a feature. In reality, it was an architecture flaw: the protocol had been designed without a mechanism to prevent sophisticated actors from extracting value from unsophisticated ones. The whitepaper buried that secret under terms like “permissionlessness” and “composability.”
Unitree’s prospectus buries its own secrets under terms like “strategic procurement” and “core technology independence.” But the structure is the same: a system designed to reward sophisticated actors—in this case, institutional investors who got access to units at much lower valuations than the IPO price—at the expense of retail investors who will buy at the listing. The sale is permissionless in the sense that anyone can buy the stock on the STAR Market. But the terms are not equal. The allocation is not fair. And the “decentralization” of ownership is an illusion, because the largest shareholders control every major decision.
This is the institutional centralization that my entire career has been mapping. In crypto, we call it the “whale problem.” In equity markets, we call it “controller structure.” The name changes. The anatomy does not.
The prospectus shows that after the IPO, the founder and associated entities will retain voting control through a special share structure. That is standard practice for Chinese tech listings, but it should be looked at with the same forensic scrutiny we applied to DAO proposals that gave founding teams veto power over treasury allocations. The message to minority shareholders is the same: you will own the risk, but you will not own the decisions.
I write this not because I believe Unitree is uniquely evil. I write this because the crypto industry has already lived through the consequences of ignoring these structures. We spent 2021 and 2022 watching DAOs collapse because “community ownership” was a UI label, not a legal reality. We watched projects that claimed decentralization hand over governance to a few large token holders. We watched “KYC” be bypassed by anyone with a few hundred dollars of wallet history. We are about to watch the same cycle play out in the humanoid robotics equity market. The S-1 is the whitepaper. The STAR Market is the exchange. The retail investor is the exit liquidity.
The question is not whether Unitree can produce a working humanoid robot. It demonstrably can. The question is whether the financial architecture being built around humanoid robotics will allow the people who fund the industry to capture any of its upside, or whether that upside will funnel to the founders, the early VCs, and the state firms that hold their preferred shares.
To answer that question, I need to dissect the core components of the Unitree story. Let me begin with the most important one: unit economics.
### The Core: A Systematic Teardown of the Unitree IPO The first thing I did after reading the headline was pull the financial statements. I wanted to see not what the company says it will do, but what the numbers say it has done. This is the “read the function calls, not the press release” approach that has kept me from being burned by dozens of projects over two decades. It took me two days to cross-reference the numbers, because the prospectus is written in Chinese accounting standards and the reporting is condensed. But the picture that emerged is bleak.
Unitree’s revenue in 2024 was approximately 1.26 billion yuan, which translates to roughly $175 million. That is a meaningful number. It is roughly double the previous year’s revenue, and it shows that the quadruped business has achieved real product-market fit. However, the company also reported a net loss. In 2023, the loss was significant. In 2024, the loss narrowed, primarily through cost reductions and improved manufacturing efficiency. But a narrowing loss is still a loss. The company is not yet profitable on an operating basis.
Now consider the valuation. A target of $8 to $10 billion implies a price-to-sales ratio of roughly 45 to 60. For comparison, Tesla, at its most hyped point, traded at about 20 times sales. Nvidia, at the height of the AI boom, traded at about 35 times sales. Unitree, with $175 million in revenue, would be trading at a premium to both of those companies on a sales multiple, while being smaller in absolute revenue by two orders of magnitude. The market is not pricing Unitree as an established manufacturer. It is pricing Unitree as a claim on the future humanoid robot market—a market that currently has almost no quantified demand.
The “total addressable market” in the prospectus is enormous. It cites industry reports projecting that the humanoid robot market could reach hundreds of billions of dollars by 2035. I have seen these reports. They are the same kind of projections we saw for smart contracts in 2017, for DeFi in 2020, and for NFTs in 2021. They are extrapolations, not fulfillments. The underlying assumption is that humanoid robots will become general-purpose workers capable of replacing humans in a wide range of physical tasks. That assumption ignores the fundamental problem of dexterity. Humans have hands with 27 degrees of freedom and a sensory feedback system that decades of AI research have not replicated. The economic value of humanoid robots is not the walking; it is the manipulation. And manipulation, as anyone who has watched a robot try and fail to fold laundry knows, is still in its infancy.
Unitree’s own flagship humanoid, the G1, is priced attractively—under $20,000 for some configurations—but it has no hands. It comes with simple grippers. The H1 also lacks advanced hand manipulation. The next-generation models, the R-series, reportedly include more sophisticated end effectors, but these are not yet in production. The promise of humanoid robotics is not a walking robot. It is a robot that can enter any workplace and use the same tools designed for human hands. That capability is years, not quarters, away.
The market structure around the IPO makes the situation worse. In crypto, when a project did an initial coin offering, the team sold tokens to early investors at a discount and then listed them on exchanges, where retail paid the true market price. Unitree is doing the same thing. The private placement rounds of 2022 and 2023 were the “seed” and “pre-sale” phases. The IPO is the “exchange listing.” The investors who bought in at earlier valuations—including state-linked funds and large venture capitalists—will be selling after the lockup period ends, presumably at a substantial profit. Retail not only pays the IPO price, but also takes the mark-to-market risk of the entire sector. That is the exit liquidity structure we saw in the crypto bull markets, and it is present here in its most traditional, regulated form.
The regulatory oversight is not the protection it appears to be. In crypto, we used to say that KYC was theater because buying a few wallet holdings bypassed it. In equity markets, due diligence is also theater. The STAR Market has investor suitability requirements in theory, but the reality is that “qualified investors” are often just people with a certain minimum account balance. They are not necessarily sophisticated analysts. My review of the prospectus found extremely limited disclosure of the specific failure modes of humanoid deployment. There is no adversarial analysis of the test results, no quantified benchmark of where the robots fail, no estimate of the cost of operating robots in a warehouse environment versus the wages of a human worker. The disclosure focuses on capabilities, not economics.
This is the fatal technical flaw. In the DeFi audits I conducted, the most dangerous bugs were not the ones that crashed the protocol. They were the ones that allowed it to function normally under benign conditions and fail catastrophically under stress. Unitree’s financial model is analogous. Under benign conditions—small batch deployments, subsidized pilot programs, government grants—the company looks viable. Under stress conditions—mass production, aggressive pricing from Tesla’s Optimus, the introduction of a cheaper humanoid from a competitor—the margin structure collapses. The function call that matters is not the revenue function; it is the competitive response function. The prospectus does not model that in any meaningful way.
Let me quantify the competitive pressure in concrete terms. Tesla’s Optimus is targeting a price of $20,000 to $25,000 per unit. Figure AI has raised over $1.5 billion to commercialize its humanoid. Agility Robotics has its Digit model, already in trials with Amazon. Sanctuary AI is attempting a different approach based on cognitive integration. In China, competitors are emerging rapidly. Xiaomi has its own humanoid prototype. Tencent is investing in humanoid startups. Baidu has an AI-focused robotics initiative. None of these competitors is a pushover, and several have deeper pockets and stronger AI capabilities than Unitree. The market is about to become overcrowded, and history—both in crypto and in hardware—shows that overcrowded markets are where margin destruction occurs.
The second core issue is technical readiness. I have spent time studying the mechanical design of Unitree’s robots, and I have genuine respect for the engineering. The electric motors and actuators are world-class. The dynamic control algorithms are impressive. The company has achieved a level of locomotion robustness that few institutions have matched. But there is a difference between a robot that can walk on flat ground and a robot that can operate a power drill or sort packages without errors. The latter requires not just actuation and control, but perception, planning, and manipulation. Unitree is not a leader in any of these. Its AI is primarily sourced from partnerships, or embedded through collaborations with large model providers. Its in-house AI capabilities are not disclosed in enough detail to verify autonomy claims.
The prospectus mentions an “AntOS” operating system and a “brain” model named in past marketing materials. But from a forensic perspective, the absence of detailed third-party testing data is telling. In my audit of 0x protocol back in 2017, I found a gas optimization flaw by reading its order-matching code. I notified the team. It acknowledged the vulnerability. The lesson stayed with me: if the code is not auditable, it is not reliable. Unitree’s prospectus contains less technical specification than a typical open-source repository. It contains photographs and marketing descriptions. That is not enough to evaluate the claim that the robots can do “general purpose” work.
Third, let me map the institutional centralization. The prospectus names the major shareholders. The founder controls the largest block. Several VC funds hold significant stakes. State-linked entities are present. This is not an ownership structure that encourages accountability to public stockholders. It is an ownership structure that enables the founder and early backers to pursue a long-term national strategic mission even if that mission destroys shareholder value in the medium term. In crypto, we called this “founder-led projects.” We learned that founder-led projects are great when the founder is brilliant and honest, and catastrophic when the founder loses touch with reality. The same risk applies here. The difference is that the founder’s “vision” is aligned with the Chinese state’s industrial policy, so the pressure to prioritize national prestige over shareholder returns is institutional, not personal.
The fourth issue is geopolitical. Unitree is a Chinese company in a sector that is explicitly listed as a national priority in both China and the United States. The U.S. has already imposed export controls on some advanced robotics and AI technologies. China has responded with its own restrictions. A Chinese humanoid robot maker listed publicly on the Shanghai exchange is unlikely to be able to sell into Western markets without regulatory hurdles. That limits the addressable market. The prospectus acknowledges this risk, but it does not quantify the revenue impact. The TAM projections assume that the products are sold globally. Yet, the practical reality is that the geography of humanoid robotics is already splitting into two blocs: a Chinese bloc and a Western bloc. Unitree is confined to the former, which is large but not as large as the global projection. This is not a small problem. It is a fundamental limit on the growth multiple the market is willing to pay.
Now, if I stop here, I risk painting a caricature of an incompetent company. That is not my intent. Unitree is a capable engineering organization with a demonstrable product and a defensible niche. The problem is the valuation and the financial architecture, not the hardware. And that is exactly why the crypto comparison is so useful: the technology was often real in crypto too, but the financialization of the technology produced ruin for latecomers. The founder’s vision may be pure. The valuation is not.
The strongest bulls will reply that Unitree is the closest to profitable mass production of any humanoid robot company, which is true but irrelevant when the prospective multiple is that high. They will also note that the company’s revenue growth rate—about 100% year over year—justifies a premium multiple. I agree that growth of this magnitude is rare. But I have seen 100% growth rates before. I saw them in L1 protocols that had no users, in NFTs that had no liquidity, and in algorithmic stablecoins that had no reserves. Growth rate alone is not a filter for investment quality. You need to look at the quality of the growth: whether it is durable, whether it is profitable, and whether it is repeatable at scale. Unitree’s growth is driven substantially by a mix of government defense orders and niche enterprise deployments. Neither of these is a self-sustaining mass market. The enterprise “demand” for quadrupeds is real, but the total market for inspection dogs is limited. The humanoid segment, which is the basis for the high valuation, has no material revenue yet.
For those reasons, my analytical verdict is that the Unitree IPO is priced for an outcome that has a low probability of occurring. The expected value is negative for retail buyers at the IPO price. The expected value is highly positive for the early investors who got in at the earlier rounds. This is a wealth transfer mechanism disguised as an investment opportunity. It is not a conspiracy. It is simply the natural consequence of an information asymmetry between informed insiders and uninformed outsiders. The crypto whitepapers had the same asymmetry, and the retail losses were enormous.
The Contrarian Angle: What the Bulls Got Right
It is easy to be cynical. It is harder—and more intellectually honest—to acknowledge what the bulls are seeing. Let me state the bullish case plainly, because if I do not, I am not a forensic analyst; I’m a propagandist.
The first thing the bulls got right is that the technology is actually progressing. When I audited the 0x protocol in 2017, I was working with a technology that was barely past the proof-of-concept stage. Humanoid robotics in 2025 is in a genuinely different place. The hardware is real. The actuators are reliable enough to operate for thousands of hours. The control algorithms are stable. The marginal cost of production is falling. Unitree has achieved what few companies globally have: a humanoid robot that can walk, run, and perform basic tasks with reasonable reliability, at a price point that is approaching real-world viability. That is not hype. That is engineering fact.
The second thing the bulls got right is the China supply chain advantage. Unitree is not a design studio that outsources manufacturing to mysterious third parties. It has its own production lines, its own motor development capability, and access to China’s mature electronics and battery supply chains. This vertical integration is a significant structural advantage. It enables the company to iterate faster, control costs more tightly, and scale production more rapidly than Western competitors who rely on outsourced manufacturing. In a hardware business, the supply chain is the moat. Unitree has a moat, even if it is narrower than the valuation implies.
The third thing the bulls got right is the industrial policy tailwind. The Chinese government has made humanoid robotics a national priority. It has directed state funds, provided research grants, and created procurement pathways. Unitree is a beneficiary of this policy. In the past, being a policy beneficiary in China meant subsidized growth and guaranteed demand in certain state sectors. That cushion reduces downside risk. It does not guarantee profitability, but it means the company is unlikely to go bankrupt in the next three years. The same cannot be said for many smaller Western competitors.
The fourth thing the bulls got right is that the first ICO was not the best one, but it was the one that established the category. The 2017 ICO boom produced countless disasters, but it also produced a few companies that generated substantial long-term value. The first humanoid IPO, even if overvalued, may be historically significant. The market is creating a benchmark. Once Unitree lists, there will be a public market reference for all future humanoid robotics companies. That standard-setting power has value. But it is not the kind of value that accrues equitably to retail investors; it is the kind that accrues to insiders who understand that the benchmark is overpriced.
The bulls also correctly note that retail markets can sustain high valuations longer than bears expect. I have been bearish on many projects that went on to triple and quadruple in price for reasons that were not fundamentally justified. Market sentiment, momentum, and the fear of missing out are powerful forces. The Unitree IPO will likely have a strong debut, simply because the narrative is compelling and the supply of robot stocks is scarce. Short-term price action is not destiny. It is a symptom of sentiment, not the logic of a valuation model.
But the most important point the bulls got right is the obvious one: humanoid robotics is a genuine technological frontier, and countries that lead it will have significant economic and military advantages. The bull case for Unitree is not just about a single company. It is about an industrial transition. The question is not whether the transition happens. It is when it happens, who captures the value, and which financial structures misallocate the risk. My concern is not that the transition will fail. My concern is that the financial architecture will fail the transition—that the IPO will overcapitalize a company that is not yet ready, and, as a result, will discourage future investment when the inevitable disappointment hits.
In crypto, we saw this pattern repeatedly. The first algorithmic stablecoin, the first NFT marketplace, the first DeFi token—all of these created enormous initial enthusiasm and enormous subsequent failure. The failure was not inevitable. It was the result of pricing assets against future expectations that were too far in the future and too uncertain in their assumptions. Unitree is not the next Luna. I want to be crystal clear: I am not predicting a $40 billion collapse. I am predicting that the financial model will not match the technological reality, and that the mismatch will have consequences for everyone who buys at the IPO price.
The bulls are right about the technology. They are wrong about the timing. And the difference between being right about the technology and being right about the investment is the difference between winning the debate and losing your capital.
That brings me to the takeaway of this entire analysis. The Unitree IPO is not a robotics story. It is a finance story wrapped in robotics packaging. The underlying asset is a real company. The underlying asset in a token sale is often a real project. The question has never been whether the technology exists. It is whether the financial instrument prices the technology fairly. Unitree’s IPO does not. It prices a narrative.
The Takeaway: Read the Function Calls, Not the Press Release
Between the lines of the ABI lies the intent. The intent of the Unitree IPO is not to fund a decade of R&D at reasonable cost. It is to raise capital at a valuation that benefits insiders, and it is to open a new frontier in the financialization of physical AI. That is not illegal. It is not even uniquely wrong. It is simply the structure of the market. If you buy this stock, you are buying the structure as much as you are buying the robot.
I am not telling you to sell the robot short. The robot may well change the world. But the world it will change is not the world of the retail buyer who pays 60 times trailing revenue for a company with a narrowing loss and growing competition. The world it will change is the one where institutional capital allocates based on engineering progress, not quarterly earnings. In that world, the winners are not the public shareholders. They are the founders, the early backers, and the state funds that can absorb long time horizons and optimize for strategic goals rather than quarterly returns.
The broader market is treating this IPO as a signal that humanoid robotics has arrived. That is true. But what it is missing is that the arrival is being financed by a financial architecture that was designed long before the robotics industry existed. It was designed in the boardrooms of venture capital firms, in the policy offices of the Chinese state, and in the math of the TAM projections. It is the same architecture that underpinned the crypto bubble and every other financial bubble I have covered in 25 years. It has never changed. It is the expectation that there will always be a greater fool.
The history of technological finance is not a history of fraudulent founders. It is a history of honest engineers whose work got repackaged into instruments that made optimistic assumptions seem guaranteed. Unitree’s engineers deserve respect. Unitree’s financial structure deserves suspicion. The IPO price deserves neither enthusiasm nor outrage—it deserves examination.
Here is my final rule for readers who want to survive this cycle, whether it is crypto or robotics: read the function calls, not the press release. Inspect the ownership. Quantify the market. Model the failure modes. And do not buy the narrative that the first public robot company is the best public robot company. The first is just the first.
The question I keep asking myself—as I look at the STAR Market, the empty retail wallets, and the backlog of humanoid orders that are still in pilot phase—is not whether Unitree will be profitable by 2030. It is whether the people funding the robot economy will still be alive to see the profit. The code does not lie. The prospectus does not lie. The structure does. And the structure says the same thing every ICO said: the risk is socialized, the upside is private.
The physical robot may be a marvel. The financial robot is the same old machine it has always been.
If you take one thing from this essay, let it be this: when the first humanoid robot IPO hits the exchange, the excitement will be real, the hardware will be real, and the price will be fictional. Treat it as such. Just as I treated every whitepaper that promised to decentralize finance but only succeeded in centralizing risk. The code whispered secrets then. It is whispering them now. Only now, the code is printed in a prospectus, and the auditors are reading it as if it were truth.
It is not. It never was. Between the lines of the ABI lies the intent. And the intent of this machine is to transfer risk from those who understand the technology to those who only understand the story.