Goldman Sachs just revised up profit forecasts for Zhongji Xuchuang (ZJXC) by 65%, 108%, and 119% for 2026-2028. This isn't a typo. It's a signal. A signal that the market is betting the house on AI infrastructure being the next railroad boom. But for those of us who chased shadows in the liquidity fog of 2017, such numbers trigger a different reflex. Not FOMO, but a forensic audit of the assumptions beneath the hype.
Context: The Infrastructure Play
ZJXC is the world's leading supplier of high-speed optical modules, specifically the 800G and upcoming 1.6T/3.2T transceivers that glue together AI training clusters. Think of them as the 'picks and shovels' for the AI gold rush, but with a critical twist: their products are the literal connective tissue of every GPU farm from Nvidia to Google. The bull case is simple: as AI models scale, the bandwidth between GPUs becomes the bottleneck. Solution? Faster, pricier optical modules. Goldman sees ZJXC as the 'TSMC of optical communication'—the dominant, indispensable supplier.
Core Analysis: The Yield as a Disguise for Risk
Goldman's core thesis rests on three pillars: surging silicon photonics module shipments, higher ASPs for 1.6T gear, and an unshakeable belief that 2027-2028 AI capex will still be accelerating. This is where the financial engineer in me begins to probe. Yields are just risk wearing a disguise. The 119% profit jump for 2028 implies a linear extrapolation of current exponential growth. But tech cycles don't work that way. They resemble logistic curves: rapid ascent, then a plateau.
Let's examine the structural assumptions. First, volume. The demand is tied directly to Nvidia's GPU roadmap. If GB200 or its successor doesn't require 1.6T optical modules in the projected quantities (due to improved in-package interconnect like NVLink or copper-based solutions for short reaches), the volume narrative breaks. History doesn't repeat, but it rhymes in code. The 400G to 800G transition saw a massive overshoot in some forecasts.
Second, price. Goldman's model builds in 'higher ASPs' for new products. This is a classic trap. Competition in optical modules is fierce. As soon as 1.6T becomes a standard, Coherent, Lumentum, and others will price aggressively to gain share. The real question isn't if ASPs decline, but how fast. My own analysis of past hardware cycles shows that monopoly-equivalent margins are fleeting. It's a game of constant engineering optimization, not structural rent extraction.
Third, the client concentration risk. ZJXC's biggest customer is likely Nvidia. That's a strength now, but a vulnerability later. If the hyperscalers (Microsoft, Google, Amazon) continue their in-house optical interconnects (like Microsoft's Lyra), they will eventually self-supply, reducing dependency on third-party vendors. Systemic rot is hidden in the fine print of these supply agreements. The 2022 crash taught us that over-leveraged concentration in any single counterparty is a ticking bomb.
Contrarian Angle: The Decoupling Mirage
The prevailing narrative is that ZJXC is a pure AI play, decoupled from macro cycles. This is the siren song of fools. Correlation is the siren song of fools. Look at the macro-liquidity picture. AI capex is a function of cheap money, low interest rates, or massive sovereign wealth fund appetite. We are currently in a high-rate, QT environment in the West. The AI spending is real, but it's a specific, finite pool. If a recession hits in 2026, corporate IT budgets freeze. The first thing to get cut? Not cloud subscriptions (essential), but hardware refreshes at the cutting edge. The 1.6T module is a luxury, not a necessity, for many enterprises.
Furthermore, the very idea of decoupling from the broader tech cycle is an illusion from a macro perspective. AI infrastructure is the most capital-intensive tech build-out since the internet. Its funding relies on the same global liquidity flows that drive everything else. When the liquidity tide goes out (e.g., due to a credit event in commercial real estate or sovereign debt), even the 'best' picks will get dragged down. The structuralist in me sees ZJXC as a high-beta proxy for the NASDAQ, not a hedge against it.
Takeaway: Positioning for the Cycle
The question isn't whether ZJXC will grow. It will. The question is whether the growth is priced for perfection. The 163% upside target implies zero operational friction for 3 years. That's an incredibly high bar. Innovation often precedes regulation by a decade, but here, the innovation is already priced. As a macro watcher, I see this as a peak-narrative moment. The real alpha might be in shorting the ETF that tracks these optimistic forecasts, or waiting for the inevitable 'supply chain normalization' headline that will smash the price. The cycle always wins. Chasing 119% growth numbers is how people get caught in the liquidity fog of 2028.
Volatility is the tax on certainty. Goldman is selling certainty. I'm buying a ticket to watch the show.
