The week humanoids went public: Agility SPAC, Unitree in Shanghai and Tesla converts its factory for Optimus

The week humanoids went public: Agility SPAC, Unitree in Shanghai and Tesla converts its factory for Optimus

🔎 Three signals in one week: the humanoid leaves the lab for the trading floor

Between July 7 and 12, 2026, humanoid robotics experienced a turning point. Not just another prototype in a video commented on by influencers. Three structural events that all say the same thing: the industry is moving from the pitch deck to financial reporting.

Agility Robotics closed its SPAC at $2.5 billion in late June. Unitree received the regulatory green light for its IPO in Shanghai, targeting a $619 million raise. And Tesla began physically dismantling its Model S/X lines in Fremont to install production for Optimus Gen 3.

The temporal clustering is no coincidence. Each player wants to be priced before another sets the reference multiple. Once a listed pure-play humanoid publishes its first quarters, the market will have a framework for analysis. The others want to negotiate their valuation in the fog, not after the first disappointment.

The viral context helps. In mid-July, a video of a robot equipped with superhuman fingers made the rounds on social media, reminding the general public that physical AI is no longer science fiction. But behind the buzz, the real story is playing out in the investment banks' Excel spreadsheets.

The essentials

• Agility Robotics goes public via SPAC at $2.5B with $300M in pre-orders for Digit v5, becoming the first listed humanoid pure-play in the United States.
• Unitree Robotics receives approval for its IPO on Shanghai's STAR Market, targeting $619M in fundraising with a potential valuation of $7B.
• Tesla physically converts its Model S/X lines in Fremont for Optimus Gen 3 production, an irreversible commitment without a dedicated listing.
• None of the three has demonstrated viable unit economics at scale. The IPOs will force transparency on real costs, uptime rates, and operating margin.

Recommended tools

Agility: the pure-play warehouse robot company taking the risk of going public

Agility Robotics has become the first publicly traded humanoid robot manufacturer in the United States. The merger with Churchill Capital Corp XI, announced in late June 2026, values the company at $2.5 billion and raises $620 million in fresh capital.

It is a bold gamble. SPACs have had a bad reputation since 2021, and no investor has forgotten the disaster of space vehicles or fintechs that followed this path. But Agility has an argument that others do not: concrete pre-orders.

$300 million in pre-orders for Digit v5 is something tangible in a sector drowning in announcements. According to TechCrunch, the startup spun out of Oregon State University has already deployed Digit in manufacturing, distribution, and logistics environments. The names mentioned by the Wall Street Journal include GXO Logistics, Toyota, and Mercado Libre.

Digit v5: what changes compared to previous generations

Digit v5 integrates significant hardware improvements: better energy autonomy, improved grasping, and above all, reliability targeted for continuous industrial shifts. The Robot Report specifies that the $620M raised will be used directly to industrialize this version and fulfill the orders.

Agility's positioning is clear: warehousing and logistics. No promises of a domestic robot, no talk of a life companion. Just boxes to move, trucks to load, repetitive tasks in semi-structured environments. This is the segment closest to potential profitability, and it is the one where B2B clients are ready to sign multi-year contracts.

Agility's specific risk

Being the first to go public means being the first to have to publish earnings. Every quarter, investors will see the gap between pre-orders ($300M) and recognized revenues. Between a signed contract and an invoiced robot, there is manufacturing, delivery, integration at the customer's site, and validation in real-world conditions. The first public quarter will be a moment of truth.

Unitree: the Chinese giant rewarded by Shanghai for volume

Unitree's story is fundamentally different from Agility's. Where Agility is a US humanoid pure-play with a few hundred deployments, Unitree is a Chinese manufacturer that has already shipped more than 5,000 robots in 2025, claiming around 39% of the global market share according to AI Weekly.

On July 2, 2026, the Chinese regulatory authority gave the green light to Unitree's IPO on the Shanghai STAR Market. The company raises 4.2 billion yuan ($619M), with a potential valuation reaching $7 billion according to KraneShares.

Figures that tell an ambivalent story

The financial data published by Caixin Global paint a nuanced picture. In the first quarter of 2026, Unitree generated 420M yuan in revenue, up 68.5% year-on-year. But net profit fell by 47.7% to 50M yuan.

This growth/margin divergence is classic in hardware at scale. Unitree is lowering its prices to gain market share. Rest of World confirms this: the IPO reveals continuously falling prices and a maximum volume strategy. It's the classic Chinese model — crush costs, dominate by scale, monetize later.

The Unitree G1 deployed at Haneda Airport illustrates this strategy: a highly visible presence in an international environment, serving as both a commercial showcase and a technical pilot.

Why the STAR Market, and not Hong Kong or the US

The choice of the STAR Market is not trivial. This market is designed for high-growth Chinese tech companies. It tolerates higher P/E ratios and shorter profitability track records than Western exchanges. Shanghai rewards hardware with real revenue — exactly Unitree's profile.

UBTECH, another Chinese humanoid player, is already listed in Hong Kong according to EVSint. But Unitree chose Shanghai, probably because the STAR Market offers a better valuation for deep tech companies with growing revenues but compressed margins.

Tesla: the most expensive physical conversion in the history of robotics

The third signal is not an IPO, but it is perhaps the most significant. Tesla is not listing Optimus separately. But what the company is doing in Fremont in July 2026 is an irreversible commitment that even an IPO does not allow.

According to Yahoo Finance and Assembly Magazine, Tesla has stopped the production of the Model S and Model X and has started converting the lines for Optimus Gen 3. TeslaRati confirms that Musk indicated a production start at the end of July or August 2026.

46 days to dismantle 14 years of automotive history

The speed of the conversion is staggering. TechTimes reports that Tesla dismantled the line in 46 days. Fourteen years of high-end vehicle production erased in a month and a half.

The cost of goods sold (COGS) target for Optimus at scale is $20,000 per robot. This is aggressive, but it is consistent with Tesla's logic: using automotive scale to compress the costs of a product that, at Agility or Unitree, still costs much more to produce.

Why the Fremont conversion changes everything

Unlike an IPO, which is a reversible financial event (a SPAC can be canceled, an IPO can go wrong), dismantling an automotive production line is an irreversible physical act. Tesla is sacrificing existing automotive revenue to bet on a market that does not yet exist.

It is also the most ambiguous of the three signals. Because Optimus does not have its own listing, its figures will be buried in Tesla's overall reporting. Investors will never know precisely how much the program costs, how many robots are produced, or at what utilization rate they are running. The transparency of Agility and Unitree, imposed by pure-play listings, will not exist here.

The three strategic bets dissected

Each player embodies a distinct business model. Comparing their valuations without understanding these differences makes no sense.

The Agility bet: logistics specialization

Agility is going all-in on a vertical segment. No service robot, no consumer humanoid. Warehouse, distribution, logistics. This is the segment where the ROI is most calculable for a B2B client: how much does a human cost per shift, how much does a robot cost per shift, what is the tipping point.

The risk is that this segment is also where non-humanoid competition is the strongest. Traditional robotic arms, AGVs (Automated Guided Vehicles), conveyors — this entire ecosystem already exists and works. The humanoid must prove it is more flexible without being significantly more expensive. Deployments like the full 8-hour factory shifts achieved by Figure Helix-02 show that feasibility exists, but not yet the economic advantage.

The Unitree bet: volume at all costs

Unitree is the only one of the three that has significant volume proof. 5,000 robots in 2025, 39% market share. But the number that counts is not the number of units shipped, it's the margin per unit. And there, the Q1 2026 data showing a 47.7% drop in net profit say something clearly: Unitree is buying its market share.

The strategy can work. If production costs continue to drop with scale, and if recurring revenues (maintenance, software, updates) materialize, the model becomes profitable. But it's a long-term bet that requires burning cash now. The $619M IPO is designed to finance exactly that.

The "physical AI closed-loop" stack presented by Foxconn at VivaTech shows that the Chinese manufacturing ecosystem is structuring itself around these large-scale deployments. Unitree is not isolated: it fits into a complete industrial value chain.

The Tesla bet: scale as the only argument

Tesla's bet is the most radical and the most fragile. The argument is simple: no one else can produce millions of robots at $20,000 each because no one else has Tesla's manufacturing scale. It's the iPhone argument applied to robots.

But the iPhone had an existing market of hundreds of millions of smartphones. The Optimus must create its own market. And unlike a smartphone, a humanoid robot must physically interact with an unpredictable world. Software is the hardest problem, and converting a production line does not solve the problem of embodiment — the fact that intelligence must be embodied in a body that walks, falls, grabs, drops.

The elephant in the room: unit economics don't work yet

This is the point that every enthusiastic press article omits. None of the three players has published data demonstrating that unit economics work at scale.

What "viable unit economics" actually means

For a humanoid robot to be a business, its total cost of ownership (TCO) over its lifetime must be lower than the cost of the human labor it replaces. This TCO includes:

• The purchase or lease price of the robot
• Maintenance costs (parts, technicians, downtime)
• Energy
• Software and updates
• Initial integration at the customer site
• Insurance and liability

None of these three players has published a complete table of these costs for a real deployment of more than 6 months. The announcements talk about "pre-orders," "pilot deployments," "full shifts." But a full shift is a technical demonstration. A business is 300 days a year for 5 years with a 95% availability rate.

BOM cost vs COGS vs selling price

TechTimes cites a COGS target of $20,000 for Optimus. This is the manufacturing cost, not the raw BOM (Bill of Materials). But even if Tesla reaches a $20,000 COGS, the selling price will have to be significantly higher to cover R&D, SGA, and margin. At $30-40,000 per robot, the economic calculation changes completely.

For Agility and Unitree, costs are likely higher at this stage. Unitree's volume helps, but 5,000 units a year is still industrial craftsmanship compared to what Tesla envisions.

The actual labor replacement rate

A robot that does 80% of a human's work does not replace 80% of the cost. If a human costs $50,000 a year and a robot does 80% of their tasks but requires 20% human supervision + 15% maintenance, the net gain is much lower than what investor deck presentations suggest.

IPOs will force the publication of these metrics. This is precisely what makes this week in July 2026 historic: for the first time, the market will have quarterly data to judge.

What IPOs change for the entire sector

Agility and Unitree going public doesn't just affect these two companies. It changes the rules of the game for the entire humanoid robotics ecosystem.

The end of the secrecy era

As a private startup, you publish what you want. The figures are selective, the metrics are favorable, the narrative is controlled. As a public company, you publish what the SEC or the CSRC demands. Every quarter, the 10-K and 10-Q (or their Chinese equivalents) will reveal:

• Real revenues by segment
• Cost of revenues and gross margin
• R&D expenses
• Number of employees
• Declared risks under penalty of prosecution

This forced transparency is an existential threat to companies whose narrative outpaces reality. And it is a major opportunity for those that actually have improving unit economics.

The benchmark effect on private startups

Once Agility publishes its first quarter, every private humanoid robotics startup will be judged against this grid. Venture capital investors will ask: "Why is your gross margin lower than Agility's?" or "Your cost per robot is 3x that of Unitree, how do you plan to reduce it?"

Subsequent rounds will be harder. The money that used to come in based on video demos will now have to be justified against the public data of listed companies. It's a brutal but necessary filter.

The deflation risk is real

If Agility's first quarters show negative margins, delivery delays on the $300M in pre-orders, or high return rates, the domino effect will be immediate. Not just on Agility's stock price, but on all valuations in the sector. VC fund LPs will call their GPs. Series B and C rounds will close more slowly. Hiring sprees will turn into layoffs.

This is the scenario the entire sector dreads. And it is exactly why the clustering of IPOs occurred: better to be priced before the first public quarter sets the benchmark.

❌ Common mistakes

Mistake 1: Confusing pre-orders and revenue

$300M in pre-orders at Agility is a signal of demand, not revenue. A pre-order can be canceled, renegotiated, or converted on a schedule spread over several years. Comparing them directly to Unitree's revenue (420M yuan in Q1 2026) is a major analytical error.

Mistake 2: Adding valuations to deduce market size

Agility at $2.5B + Unitree at 7B yuan does not make a $9.5B "humanoid market". These valuations reflect future growth expectations, not the current market size. The real market for commercially deployed humanoid robots in 2026 is likely less than 1B yuan in consolidated annual revenue.

Mistake 3: Treating Tesla Optimus as a direct comparable

Optimus has no standalone P&L, no public listing, and no declared external customers. Including it in a comparison with Agility and Unitree as if the three were equivalent is misleading. It is an internal project of an $800B market cap company, not a startup that has to prove its business model.

Mistake 4: Ignoring Chinese regulatory risk

Unitree's IPO on the STAR Market depends on Chinese regulatory approval, which can be revoked, delayed, or conditional. Western investors who treat this IPO as a purely economic event forget the political risk inherent in any listing on a Chinese market.

❓ Frequently Asked Questions

Why are these three events happening at the same time?

The clustering is not accidental. Each player wants to be valued before a peer publishes its first public figures. Once Agility goes public, the market will have a reference multiple. The others prefer to negotiate in uncertainty rather than after the first concrete data point.

Which of the three bets is the riskiest?

Tesla is the riskiest in terms of execution (factory conversion, no commercial robotic track record), but the least risky in terms of survival (funded by automotive cash flow). Unitree is the riskiest in terms of business model (compressing margins). Agility is the riskiest in terms of market expectations (first listed pure-play, every quarter will be scrutinized).

Will unit economics improve?

Probably, but the question is on what timeframe. The history of automation shows that costs decrease with volume, but that real-world reliability takes years to reach the levels promised in the lab. The next 3 to 5 years will be a painful ramp-up phase.

Should you invest in these companies' stocks?

It depends on your risk tolerance. A humanoid robotics IPO is a bet on a category that has not yet proven its business model. The public data that will arrive with Agility and Unitree's quarterly reports will allow for a more informed judgment. Before that, it is speculation.

Will the transparency required by being listed help or hurt the sector?

Both. It will help solid companies by validating the category with institutional investors, thereby unlocking more capital. It will hurt companies whose narrative is overvalued by revealing the gap between promises and figures. On balance, it is positive for the sector in the long term, but painful in the short term.

✅ Conclusion

The week of July 7-12, 2026, marks the transition of humanoid robotics from the age of demos to the age of numbers. Agility, Unitree, and Tesla have each made an irreversible decision — SPAC, IPO, factory conversion — that publicly commits them. None have yet proven that the business model works. But for the first time, the market will have the data to judge it. The bubble will inflate or deflate next quarter.