The demand for AI computing power has propelled the optical communication industry into a rare period of high prosperity. However, according to Guosheng Securities, whether this growth can be realized increasingly depends on global delivery capabilities, accurate technology roadmap judgments, and access to upstream core resources.
In an industry weekly report dated April 26, the Guosheng Securities team led by Song Jiaji stated that the optical communication sector is currently experiencing a super cycle driven by AI computing demand. Yet, behind this high prosperity and rapid growth, companies face a series of capability shortcomings.
The report breaks these challenges into four key areas: the growing pains of establishing overseas manufacturing facilities; the anxiety caused by technological generational shifts moving from "linear evolution" to multiple parallel development paths; structural shortages in upstream component/chip/foundry capacity ahead of the volume ramp for 1.6T products; and the reality that small and mid-sized companies are facing gaps in both customer acquisition and funding.
The conclusion is clear: competitive advantage will skew towards industry leaders. Those who can build robust global delivery capabilities, deploy multiple technology platforms, secure key supply chain resources in advance, and withstand high R&D and capital expenditure are more likely to sustain the long-term benefits brought by AI.
Overseas expansion is a hard constraint, with three major hurdles for setting up factories abroad.
The report frames globalization in pragmatic terms: core customers for global AI computing investment are concentrated in North America, making overseas markets vast. For domestic Chinese optical communication companies, growth ceilings will quickly become apparent if they cannot integrate into the global supply chain.
The question is "how to expand overseas." Guosheng Securities identifies three specific challenges for establishing overseas factories:
* Slow yield ramp-up: Southeast Asia, particularly Thailand and Vietnam, is currently the preferred location for new factories. However, the local industrial workforce base is weaker, and production line proficiency lags behind domestic levels, often requiring significantly more time to achieve comparable yields. * High training and assignment costs: Core technical positions often require staff assignments from China. Visa processing, living arrangements, and cultural adaptation can slow the process, while training local employees from scratch is a lengthy cycle. * Policy and compliance uncertainties: Companies must adapt to new environmental impact assessments, land planning, labor regulations, and tax policies. Furthermore, the global minimum tax policy, set to take effect starting in 2025 in jurisdictions like Thailand, Singapore, and Hong Kong, will increase compliance costs.
These factors collectively point to the same result: overseas expansion is far more complex than simply relocating capacity. Delivery stability will become a new critical differentiator.
From 800G to 1.6T: Faster iteration speeds amplify technology roadmap risks.
The report emphasizes that optical communication is in a period of simultaneous breakthroughs in materials, data rates, and integration architectures, meaning leadership windows could narrow rapidly. Data indicates that the iteration cycle for optical modules is accelerating, shrinking from the traditional 3-4 years for previous generations to just 1-2 years for the transition from 800G to 1.6T.
This shortened cycle has a dual effect: while it supports high gross margins, it also intensifies pressure on R&D and production line transitions. The report clearly outlines the "parallel path" scenario:
* Material platforms are no longer dominated solely by Indium Phosphide (InP); alternatives like Silicon Photonics, Thin-Film Lithium Niobate (TFLN), and heterogeneous integration are now competing. * In terms of technical architecture, pluggable optical modules remain mainstream, but new architectures like Co-Packaged Optics (CPO), Near-Packaged Optics (NPO), and Optical Circuit Switches (OCS) are developing in parallel. A large-scale shift by core chip manufacturers and Cloud Service Providers (CSPs) could reshape the entire industry landscape.
In this environment, corporate anxiety manifests as two critical choices: how to extend the lifecycle of existing products, and how to avoid having heavily invested directions being disrupted. The report suggests a strategy favoring larger players: employing a "multi-technology platform layout" to create room for error.
On the eve of the 1.6T ramp, upstream bottlenecks center on EML, DSP, and Silicon Photonics foundry capacity.
During demand upswings, upstream "critical components" are the first to face strain. The structural shortages are primarily focused on three areas:
* High-speed EML optical chips: Production capacity is highly concentrated among American and Japanese companies. Traditional Japanese giants tend to have conservative capacity expansion strategies, leading to a supply-demand gap for high-end EMLs as 1.6T demand surges. * DSP electrical chips: The supply landscape is a duopoly dominated by Broadcom and Marvell (via its acquisition of Inphi). For the 200G DSPs used in mainstream 1.6T solutions, the weekly report notes Broadcom holds a more significant technical advantage. * Silicon Photonics wafer foundry (FAB): Tower Semiconductor is described as a core foundry for Silicon Photonics tape-outs. Its specialized process platforms have scarce and tightly scheduled capacity. The ability to secure Tower's capacity in advance directly impacts the actual delivery capability of silicon photonics modules.
The underlying message here is straightforward: as data rates increase, competition shifts beyond "who can make it" to "who can secure key resources and ensure stable delivery."
Small and mid-sized companies face greater difficulties: slow customer adoption and rapid cash burn create a vicious cycle.
The report argues that in a market with high concentration, smaller companies are squeezed from both ends. On one end are customers. Gaining approval from leading customers involves long cycles and low tolerance for error; a lack of reference cases further raises the barrier to customer acquisition, creating a negative feedback loop. On the other end is funding. Optical communication is a sector characterized by high R&D investment and high capital expenditure. Establishing production lines, obtaining certifications, and maintaining inventory all require significant upfront capital, creating inherent cash flow pressure.
When upstream components are scarce, downstream customers are demanding, and technology roadmaps are shifting rapidly, the flexibility of smaller manufacturers can become a vulnerability. The weekly report crystallizes the "strong get stronger" dynamic into four core capabilities: global delivery capability, diversified technology platform strategy, proactive supply chain locking, and financial strength.
Comments