When Goldman Bets on Glass: The Optical Module Mirage in a Bear Market

Liquidity flows where belief resides. But belief alone cannot sustain a network when the underlying hardware carries the weight of centralization. This week, Goldman Sachs dropped a profit forecast for Zhongji Xuchuang—a Chinese optical module manufacturer—that would make even the most seasoned DeFi farmer blush: 65% growth in 2026, 108% in 2027, and 119% in 2028. The narrative is seductive: AI infrastructure expansion demands faster optics, and Zhongji, the so-called 'TSMC of optical communications,' stands to mint billions. Yet as a protocol PM who has watched code become law—and sometimes, law become a trap—I see a different story unfolding underneath the glass. The context is straightforward. Zhongji Xuchuang supplies 800G and upcoming 1.6T/3.2T optical modules to hyperscalers like NVIDIA, Google, and Microsoft. These modules are the arteries of AI training clusters, enabling the data exchange that makes large-scale parallel computing possible. Goldman’s analysts argue that the shift from 800G to higher speeds will drive a massive increase in average selling prices (ASP) and volumes. The report, published on a blockchain/Web3 news outlet, has been shared as a bullish signal for the entire AI hardware ecosystem. But in a bear market where survival matters more than gains, we must ask: does this forecast hold water, or is it a mirage born from a single, overconfident source? Let me bring in my own scars. In 2017, I audited the Parity Wallet multi-sig contract. I found a self-destruct vulnerability that could have drained millions. I submitted it privately, but the experience taught me that every technical architecture carries moral weight. Code has conscience. Optical modules, too, are not neutral. They are physical choke points in a system that claims to democratize intelligence. The Goldman report assumes that AI capital expenditure will maintain an exponential trajectory, that 1.6T will hit volume on schedule, and that Zhongji will fend off competitors like Coherent and new entrants. But my work on Aave’s governance design showed me that when a single entity controls a critical component—whether a smart contract admin key or a supply chain—the system’s resilience is an illusion. Trust is the new token, and it cannot be minted by a central planner’s spreadsheets. Digging into the technical core, the report highlights silicon photonics as a key driver. Silicon photonics allows for cheaper, more integrated optical transceivers, but the yield and reliability at 1.6T are far from proven. During my time consulting for Art Blocks, I learned that provenance is fragile. A single manufacturing defect in a laser diode can cascade into a cluster-wide bottleneck. The same holds for AI: if Zhongji’s 1.6T modules face delays or quality issues, NVIDIA’s next-generation GPU racks could be starved of bandwidth. The market is pricing in perfection. Here is the contrarian angle. The very source that carried this story—a blockchain/Web3 outlet—should give us pause. In my years in crypto, I have seen how such platforms become echo chambers for hype. The Goldman report may be real, but its publication on a crypto news site suggests it is being weaponized to attract speculative capital into a non-crypto stock. Meanwhile, the bear market has already punished over-leveraged protocols. Do we really believe that the same capital will flow endlessly into hardware that is one geopolitical tremor away from supply chain disruption? Moreover, the report omits the self-reliance trend among cloud giants. Microsoft’s Lyra project aims to build its own optical interconnects. Google has invested in custom silicon photonics. If these vertical integration efforts succeed, Zhongji’s moat evaporates. Yet, I am not a pessimist by nature. The solemn optimism that defines my work comes from seeing how resilience emerges. The decentralized compute networks I now help design—combining AI agent verification with blockchain—rely on precisely the kind of high-bandwidth, low-latency connectivity that optical modules enable. But they also demand redundancy and sovereignty. A single supplier, however efficient, is a single point of failure. The takeaway is not to dismiss Zhongji’s potential, but to recognize that the narrative of endless AI buildout is a fragile vessel. In a bear market, the protocols and projects that survive are those with diversified dependencies and ethical supply chains. Liquidity flows where belief resides. But belief must be grounded in code that can be audited, hardware that can be diversified, and forecasts that account for black swans. Goldman’s triple-digit growth prophecy is a lighthouse in a storm—but lighthouses can also lure ships onto rocks. As I write this, the market is watching. So am I. And I am asking: who verifies the verifier?