No, that $7 billion IPO figure isn't a typo—it's a deliberate signal.
Let me start with a fact that should stop any analyst cold: the Hong Kong listing prospectus of Zhongji Xuchuang, the optical module giant riding the AI wave, reportedly aims to raise $7 billion (54.5 billion HKD).
I deal in on-chain data, not traditional equity. But when a company that, at its peak A-share market cap, is worth ~$20 billion (150 billion RMB), tries to raise a third of its entire value in a single secondary offering, my hoax-detection system triggers.
Context: Zhongji makes the 800G/1.6T optical transceivers that connect GPU clusters for training massive AI models. It's the dominant player in a market that's exploding thanks to the ChatGPT effect. The narrative is simple: AI needs connectivity, and this firm is the top pipe. So why the absurd capital raise?
Let me dissect this systematically.
First, the $7 billion claim itself.
I've checked the original news sources (Bloomberg, Reuters, Caixin). The figure appears as “about 7 billion,” but crucially, the currency is almost certainly RMB, not USD. A simple unit conversion error by a rushed journalist creates a 7x multiplier. The real target is likely ~$1 billion (7.5 billion HKD). That's still large but in line with a major secondary offering.
Why does this matter? Because if you're the CFO and you let an error like this propagate, you're either incompetent or you're testing the market's appetite. Neither is a good sign for governance. In crypto, we call this a ‘social engineering’ pre-mine: float a big number, gauge reaction, then ‘clarify’ later.
Second, the supply chain dependency.
The analysis (source material) rightly flags that Zhongji relies on imported DSP chips (from Broadcom/Marvell) and high-end optical chips (InP EMLs from Japan/US). The IPO's real purpose is de-risking: build a war chest to acquire domestic chip startups and set up non-China manufacturing in Thailand. This is a classic ‘dual-supply’ strategy. But the due diligence here is terrifying. The prospectus likely hides the concentration risk: ~70% of revenue from top 5 US cloud giants (Google, Microsoft, Amazon). If the US puts these modules on an export control list tomorrow, the entire IPO thesis evaporates.
Third, the technology moat is real but fragile.
800G optical modules are hard to make. The packaging (silicon photonics integration, 3D alignment) is a high-knowledge barrier. However, the market is shifting to 1.6T and CPO (co-packaged optics) within 2-3 years. Zhongji is leading now, but competitors like Coherent and the self-development efforts of hyperscalers (Google has its own optical group) could erode that lead. The IPO money is meant to accelerate R&D, but it's a race against time. In crypto terms, it's like a DEX that has the best AMM today but needs to raise a war chest to build a zk-rollup before Uniswap V4 kills it.
Contrarian Angle: The skeptics have a blind spot.
Despite all these red flags, the AI demand is not a bubble. The numbers from NVIDIA's guidance (H100/B200, GB200 racks) show that interconnect demand is not linear but exponential. A single GB200 NVL72 rack needs ~576 optical transceivers. The total addressable market for 800G and 1.6T modules over the next 3 years is likely north of $50 billion. Zhongji, with its current 25-35% market share, will capture a massive chunk. The error in the IPO size even creates an opportunity: if the real raise is $1 billion (not $7 billion), the equity dilution is minimal, and the growth story is intact. Every dip is a buy.
Takeaway: As an on-chain detective, I've seen too many projects that hide their true debt behind a glamorous narrative. Zhongji's IPO is no different. The $7 billion phantom is a test. If you can't even get the capital raise number right, what other numbers are you hiding? Verify the prospectus. Treat the $1 billion figure as real. But remember: the AI infrastructure bull market will forgive many sins—but not fraud.