A first-day close at roughly 5.7 times the offer price is information about the auction, not about output [6]. The proceeds were fixed when the book closed. Everything after that changed hands between investors, in a market that S&P Global Market Intelligence's Ruiying Zhao describes as heavily driven by retail participants [9], and that she credits to investor appetite for AI and robotics [20]. The fab's output does not change because the ticker moves.
Back the scale out before treating this as the centre of global issuance. Divide the combined Hong Kong and Shanghai proceeds by their 21% global share and the world pool lands near $257 billion, of which Nasdaq's 55% is about $141 billion [2]. SpaceX's single June listing at $75 billion is therefore about 1.4 times everything the two Chinese venues raised between them [13][3]. On the stated floors, those venues are up about 17% year on year [1]. Growth is real. The base is still smaller than one American deal.
CXMT's revenue is the line I would underwrite against. More than 700% growth means at least eight times, so dividing the quarter by eight puts the year-earlier period at about 6.4 billion yuan, under $1 billion [4]. That is a steep curve off a small base, at a company founded in 2016 [7], selling into a demand spike for chips needed for AI [6]. The raise is about 1.15 times a single quarter of revenue [5], which is a modest cheque for a business that adds capacity in fab-sized increments. For that curve to transfer into anything a buyer can plan around, memory pricing has to hold and the new output has to qualify at customers who currently buy elsewhere. The reporting says nothing about either, and nothing about what the proceeds fund.
Unitree supplies the only durability reading in the set, and it points down from where the debut ended [4]. Zhao's stated test for a durable cycle is sustainable revenue, visible profit margins and realistic valuations [14]. A debut pop scores zero on all three.
The venue choice has a mechanism behind it that is not mainly about price. Freshfields' Howie Farn says listing overseas typically takes more time than an IPO in China [15], and tighter US and Chinese scrutiny of strategically important sectors has kept some issuers closer to home [16]. Because mainland exchanges limit foreign purchases, a company that wants international capital runs a parallel Hong Kong book as well [17]. Shein worked through the US and London before settling on Hong Kong [18]. Hong Kong also took Luxshare Precision and Zhongji Innolight, which makes the optical transceivers that go into data centres [22]. WPIC's Jacob Cooke argues the AI investment cycle is absorbing the risk appetite that would otherwise have reached a company like Shein [23], and Shein priced at about $27 billion, a fraction of its earlier peak, partly because of US and EU moves against de minimis exemptions [19].
Two comps now exist for pricing a Chinese AI-hardware listing. They disagree with each other.
