Goldman Sachs just told you there's 163% upside in a Chinese optical module maker. That should make you nervous.
Let me be clear: I don't trade on sell-side forecasts. I trade on liquidity mechanics. And when a blue-chip bank publishes a 65%, 108%, 119% profit growth projection for a single supplier over three consecutive years, the first question isn't "Is it real?" It's "Who's getting out before the last buyer?"
Zhongji Xuchuang is the world's leading supplier of 800G and 1.6T optical modules for AI data centers. Think of them as the plumbers of the AI gold rush—every NVIDIA H100 and B200 cluster needs their fiber to talk to each other. Goldman's argument is straightforward: AI infrastructure expansion is non-linear, optical modules are the bottleneck, and Zhongji's silicon photonics tech lets them capture the entire upgrade cycle from 800G to 1.6T to 3.2T. Each transition comes with 30-50% higher ASP. Gross margins expand. Earnings explode.
I've seen this narrative before. In 2020, every DeFi protocol was "non-linear," and the ones that survived were the ones with the most cynical treasuries—not the most ambitious roadmaps. During the Terra collapse, I watched $60 billion evaporate because the exit liquidity was never questioned. The same discipline applies here.
Let's dissect the forecast with the tools I use for options strategies—not hope, but order flow.
First, the core assumption: that AI capex from hyperscalers (Microsoft, Google, Amazon, Meta) will compound at 40-50% CAGR through 2028. That's not just a bet on model scaling—it's a bet on infinite datacenter buildout, infinite power subsidies, and infinite demand for inference compute. One slowdown in AI training efficiency, one regulatory clampdown on power consumption, one breakthrough in model compression that reduces compute needs by an order of magnitude, and that capex curve flattens. Optical modules are a derivative of that capex, not a prime mover. When the tide goes out, every supplier's revenue gets cut in half.
Second, the competitive dynamics. Goldman paints Zhongji as the "TSMC of optical." That metaphor is dangerous. TSMC holds a near-monopoly on leading-edge semiconductor manufacturing because the capital cost and process complexity create an unbreachable moat. Optical modules have lower barriers: die-to-die interconnects, laser integration, and packaging yield are challenges—but not monopoly-level ones. Coherent, Lumentum, and competitors like Eoptolink are all scaling silicon photonics. Worse, hyperscalers are increasingly self-developing optical interconnects—Microsoft's Lyra, Google's light-based switching. When your biggest customers become your competitors, the pricing power Goldman assumes disappears. Arbitrage doesn't ask for permission, but it always asks for price.
Third, the timeline risk. 1.6T modules are still in qualification with NVIDIA. Volume production is scheduled for late 2025. The 3.2T spec isn't even finalized. Goldman's 2027-2028 estimates assume flawless execution across two technology generations and zero supply chain disruptions from DSP chips (Broadcom, Marvell) or laser diodes. In my experience auditing smart contracts, the most catastrophic failures come from dependencies that seem trivial—a reentrancy bug in a function you assumed was safe. Here, the dependency is on 800G DSP availability. If Broadcom delays its next-gen Tomahawk switch, the 1.6T module market doesn't exist. Terra’s code was poetry; Luna’s exit was prose. This is prose with broken syntax.
Now, the contrarian angle: this article originated from a blockchain/Web3 media source, not a financial wire. Why? Because the crypto crowd loves narratives with clear catalysts and high multiples. They're prone to treat a sell-side report as confirmation of their thesis—exactly when liquidity is being built for the exit. Options don't care about your thesis. They only care about the gap between someone's belief and reality.
In 2022, I watched the same pattern with Luna: a rapidly growing ecosystem, backed by top-tier VC forecasts, with a bottleneck (the UST-Luna mint mechanism) that everyone knew existed but no one priced correctly. The optical module boom has a similar structural fragility: it's long-dated, high-conviction macro bet in a single industrial vertical. The retail buyer of Zhongji shares today is effectively providing exit liquidity for the early adopters who got in during 2023. Every Goldman upgrade attracts more buyers, but the fundamental question remains: who gets to sell before the music stops?
Let me add some real numbers from my own trading. I ran a delta-neutral ETF arbitrage strategy in 2024, capturing the basis spread between spot Bitcoin ETFs and the underlying. The biggest lesson was that liquidity is not continuous—it's episodic. When a narrative flips from bullish to neutral, the bid disappears faster than any forecast can adjust. The optical module thesis relies on continuous AI capex growth. One earnings miss from NVIDIA, one hyperscaler announcing a pause in datacenter expansion, and the bid on Zhongji evaporates. In a thin market, the move down is 3-5x the move up.
The article also misses the real bottleneck: power. AI datacenters consume 10-20 MW per cluster—and that's before the 1.6T modules, which draw significantly more power per port. In Virginia, the world's largest datacenter hub, utilities are already warning that new connections can't keep up. If power constraints slow deployment, the optical module demand curve shifts right by 12-18 months. Goldman's model assumes no such delay. Risk isn't a number. It's the gap between belief and reality.
So what's the takeaway? The market is pricing Zhongji as if the AI infrastructure buildout is deterministic. It's not. The next 12 months will bring either validation (massive 1.6T order from NVIDIA) or reality check (delays, competition, or capex slowdown). For the disciplined trader, the entry is not at current levels—it's at a level where the downside is capped by tangible asset backing and a clear exit signal. I'd wait for the first miss on guidance, see how the stock reacts, and only then consider a position with a defined risk limit. HODLing blind is just gambling with extra steps.
The real question isn't whether Zhongji can grow 119% in 2028. It's whether you can sell before the exit liquidity dries up. In a bull market, everyone thinks they're the DJ. They're not. They're the dance floor.
Set your exit now. When the narrative shifts, it shifts fast.
Terra’s code was poetry; Luna’s exit was prose. Options don't care about your thesis. Risk isn't a number, it's the gap between belief and reality.
