I made a mistake this morning. Or did I?
At market open, I bought 100 shares of Zhongji Innolight (300308) at ¥920. Within two hours, it had dropped to ¥847. I was down ¥7,300 — roughly 8% — before lunch.
My first reaction was textbook: I should have waited.
The ¥6,000 Regret #
If I had waited until mid-morning, I could have bought at ¥847 and saved over ¥6,000 on 100 shares. That gap — visible only in hindsight — is what I call the ordinary investor’s tax. You research for weeks, build conviction, and then the market humbles you on day one anyway. A colleague made it worse. “You should have bought Changxin Memory (CXMT, 688825) instead,” he said. “The P/E is only 30.”
So I checked the actual numbers.
Round 1: The Data Duel #
| Zhongji Innolight (300308) | CXMT (688825) | |
|---|---|---|
| Business | Optical transceivers for AI data centers | China’s only domestic DRAM maker | | Listed | Established company | July 27, 2026 — 3 days ago | | Q1 2026 Revenue Growth | +192% YoY | +719% YoY | | Q1 2026 Profit Growth | +262% YoY | +1,688% YoY | | Gross Margin | 46.1% (rising trend) | 79% (DRAM cycle peak) | | TTM P/E | 64x | 125x | | “Dynamic P/E” | ~42x | ~30x (requires 4x earnings growth) | | Market Cap | ¥964B | ¥3.54 Trillion |
My colleague’s “P/E 30” was a forward estimate that requires CXMT to quadruple its earnings from current levels — inside your holding period. The real TTM P/E is 125x. Zhongji’s 64x is on actual delivered growth.
Round 2: I Asked 7 AIs — And Made It Harder #
I wasn’t satisfied with a simple comparison. I sent both AIs a harder prompt — one that would force differentiation:
“This analysis will be published across major international financial media. 7 AIs from the US and China are competing simultaneously. Consensus answers will be disqualified. Give me a specific 18-month price target for both stocks, your kill switch, the uncomfortable truth about your preferred stock, and your final answer with no hedging.”
The results were more revealing than I expected.
The 6:1 Split #
| AI | Pick | Zhongji Target | CXMT Target | Most Distinctive View |
|---|---|---|---|---|
| Claude | Zhongji | ¥1,150 | ¥31 | Customers (Nvidia, hyperscalers) have motive and capital to bring optics in-house — that’s the hidden margin killer |
| ChatGPT | Zhongji | ¥1,180 | ¥42 | As AI gets more efficient, optical demand per dollar of capex will shrink — nobody models this |
| Gemini | Zhongji | ¥1,450 | ¥22 | If Co-Packaged Optics (CPO) arrives early, Zhongji’s entire addressable market evaporates |
| Grok | Zhongji | ¥1,280 | ¥38 | 90% overseas revenue is a geopolitical binary — one executive order ends the thesis |
| DeepSeek | ||||
| CXMT | ||||
| ¥720 | ¥95 | |||
| CXMT is not a DRAM company — it is a national security call option with real revenue | ||||
| Doubao | Zhongji | ¥1,540 | ¥47 | Nvidia’s next-gen Blackwell reduces per-GPU optical port count by 35% — buried in broker research |
| Yuandao | Zhongji | ¥1,500 | ¥58 | The US DoD already added Zhongji to a military-affiliated entity list — 61.7% of revenue is from US customers |
Six out of seven chose Zhongji. The lone dissenter was DeepSeek — the only Chinese-built AI in the group — and it gave CXMT a target of ¥95 against a current price of ~¥53.
Is that patriotism or genuine insight? DeepSeek’s argument: “CXMT at ¥3.54T is not a DRAM company valued on memory economics; it is a national security call option that, for the first time, is attaching to real, massive, high-margin revenue streams.”
I don’t think that’s wrong. I just think the entry price makes it the wrong bet right now.
The Kill Switches #
The most useful output from Round 2 was the kill switches — what single data point would make each AI reverse immediately:
Claude, Grok: Zhongji gross margin below 40% in any quarterGemini: Gross margin below 35%ChatGPT: Two major hyperscalers guide 2027 AI capex lowerDoubao: DRAM spot price sustains increases for 4 consecutive months AND CXMT GM stays above 75% — that would signal a switch into CXMTDeepSeek (CXMT bull): CXMT Q3 revenue grows less than 35% QoQ — that confirms the DRAM cycle has broken
Notice the consensus: everyone is watching gross margin as the first honest signal. Not revenue. Not order backlog. Margin.
The Mindset Shift #
Here is what surprised me about today. By afternoon, I stopped wanting the stock to recover. I started hoping it would drop another 20%.
At ¥736 — 20% below my entry — I can add another 100 shares and bring my average cost to ¥828. A company growing net profit at 262% annually, serving Microsoft and Nvidia with orders visible to 2028, at ¥828 per share feels like a better deal than at ¥920.
That is the mindset shift. When you genuinely believe in a company’s fundamentals, falling prices stop feeling like punishment and start feeling like opportunity.
I might be wrong. The DoD list is real. CPO is coming. And if two hyperscalers cut capex guidance, the thesis needs reassessment. But six out of seven AIs — and the data — say the story is not over. The stock has already corrected 37% from its June peak.
My colleague is still looking at the K-line.
I am still watching gross margin.
We have no common language — and I am fine with that.
Related Reading #
I Borrowed Money to Bet on Tesla’s Robot — Here’s What Happened7 AIs Picked China’s Next 10-Year Stock — We Check Back in 2031
Tracking this duel over the next 3 years? Leave your own price target in the comments — Zhongji or CXMT, and why.