On 23 February 2026, a young Chinese humanoid-robotics company, AI² Robotics (智平方), announced a Series B worth more than one billion yuan — roughly 144 to 145 million dollars — lifting its valuation beyond ten billion yuan, or about 1.4 billion dollars. The figure is large, but it is not what merits a pause. What merits a pause is the list of those who signed the cheque.
For the investors in this round are not, for the most part, venture-capital funds. Among them are Baidu's investment arm, the broker Guotai Haitong Securities, the fintech services provider Yusys Technologies, the tyre manufacturer Sentury Tire — and, above all, CRRC Corp., China's state-owned railway manufacturer. A state industrial group is taking a stake in a robot maker. The question is not how much it put in, but what it puts in exchange, and what it expects in return.
The shareholder is also the customer
The answer fits in a single sentence, and it is that sentence which changes everything. CRRC brings not merely capital: it brings large-scale industrial environments in which to test and deploy the robots. Baidu, for its part, brings its AI models. In other words, those who finance the machine are also those who will put it to work. The fund that normally invests bets on future demand it does not control; here, the investor is the demand.
The same statement confirms it without ambiguity: AI² Robotics has secured an order for 1,000 robots over three years from HKC, the world's third-largest LCD-panel maker. Morgan Stanley sees it as the largest single order in the world in this sector. The purchase order and the funding round are part of the same operation: the factory is capitalised and it is guaranteed in advance who will buy its output. The funds raised will go primarily to developing the onboard AI model GOVLA and to expanding production capacity for the AlphaBot range — that is, to manufacturing precisely what the shareholders have just pre-ordered.
The numerical target follows the same logic: AI² Robotics is aiming for 10,000 AlphaBot robots produced in 2026, against a capacity of just 1,000 units a year only reached in September 2025. A tenfold increase in a matter of months, backed not by a market to be won but by an order book already partly filled by the financiers themselves.
And the loop does not stop at production. The CEO of AI² Robotics says he wants to take the company public within one to two years, and the presence of Guotai Haitong Securities among the investors confirms that an IPO is already in the works. The broker who will value the offering on the market has come in upstream: it is buying low to sell high an asset whose order book it knows to be pre-filled by other shareholders. The supply chain finances itself, orders from itself and will list itself.
In the West, seven robots are leased on trial
Four days earlier, on 19 February 2026, the West was carrying out the exact opposite operation — and it is this counter-shot that sharpens the contrast. Agility Robotics and Toyota Motor Manufacturing Canada announced a commercial agreement for the deployment of Digit humanoid robots at the Woodstock plant in Ontario, dedicated to the RAV4. The first commercial deployment of humanoids in automotive production in Canada.
The scope, however, is nothing like the Chinese one. The agreement covers 7 robots, of which only 3 will be put into service from April 2026; the remaining 4 are conditional on the success of this first wave. And the financial structure is the antithesis of the AI² model: it is a Robots-as-a-Service contract. Toyota takes out a subscription including maintenance and software updates, without ever owning the robots outright. The agreement itself follows a one-year pilot run in three phases — development, proof of concept, on-site trial.
On one side, a shareholder-customer ordering 1,000 units over three years from a subsidiary it has just capitalised. On the other, an industrial firm leasing 3 robots on trial and reserving the right to hand back the keys. This is not the same market at two speeds: these are two market structures. In the East, industrial and state capital pre-orders the volume it will finance, and will cash it in as a capital gain on the day of the stock-market listing. In the West, the customer tests on a lease and keeps control of the off switch.
Twofold: the measure of what is sold to oneself
One figure remains, which the week leaves lying around without comment, and which says it all. Projections for humanoid deliveries in China in 2026 diverge twofold. The Chinese consultancy GGII, which puts domestic shipments for 2025 at about 18,000 units — a 650% growth year on year —, forecasts 62,500 units in 2026. Morgan Stanley, for its part, projects only 28,000.
This gap is not a quarrel among analysts: it is the accounting translation of the thesis. If demand is pre-ordered by shareholder-deployers such as CRRC or HKC, then the high figure describes an order book captured in advance by the supply chain, independently of any end market. If one expects real demand, exposed to competition and risk, the low figure holds up better. The factor of two between 62,500 and 28,000 is, near enough, the measure of what is sold to oneself in advance.
Since the autumn, we have seen the demo parade past, the word "autonomous", the televised set piece and the calculations of the number one. This week steps down a notch, to the level of capital. The right question is not whether the Chinese market is moving twice as fast as the West. It is to look at who signs the purchase order — and to note that he also signed the cheque for the capital increase. In the West, seven robots are leased on trial. In the East, the shareholder orders a thousand from the subsidiary he has just capitalised.