Northeast Securities has released a research report indicating that the OCS industry trend is currently dominated by overseas major manufacturers, with domestic companies gradually increasing their participation. It is recommended to focus on opportunities in component supply or outsourcing. The report highlights three major investment areas: companies already involved in overseas OCS supply/outsourcing, companies with potential to enter overseas OCS supply/outsourcing, and optical module manufacturers actively preparing for OCS.
Northeast Securities’ main viewpoints are as follows: OCS optical switching is a technology that switches optical signals between fiber ports without needing optoelectronic conversion (O/E/O). The principle of OCS involves physically reconstructing the path of optical signals, establishing dedicated optical paths between input and output ports. Implementing OCS in data centers can significantly improve overall network performance, operational efficiency, and sustainability, providing substantial advantages.
OCS can be applied in three types of networks within data centers: TPU cluster interconnections (scale-up networks), Spine layer (scale-out networks), and DCI cross data center interconnections (scale-across networks). OCS is expected to see resonance both domestically and internationally, with a broad market outlook. Overseas, OCP has announced the establishment of an optical switching OCS sub-project, which is expected to promote industry standardization. Domestically, telecommunications operators and the Ministry of Industry and Information Technology are actively advancing OCS implementation.
According to Cignal AI, by 2025, the OCS market, led by Google’s MEMS OCS, is estimated to reach approximately $400 million; by 2029, the market size for OCS is projected to exceed $1.6 billion, with a four-year CAGR of around 41%. LightCounting predicts that OCS shipments will surpass 50,000 units by 2029, with a CAGR of 15% from 2025 to 2030, as more cloud manufacturers beyond Google are expected to drive market growth.
The four major technological routes will evolve together, with a comprehensive interplay of costs, performance, and technological difficulty. OCS primarily includes four mainstream technological routes: MEMS, liquid crystal, piezoelectric, and silicon photonic waveguides. Currently, the application pace of MEMS and liquid crystal solutions is faster. Each technological route has its advantages and disadvantages, with no single solution currently holding a clear advantage. The firm believes that the four technological routes will evolve together in the future, considering costs, performance, and technical difficulty as significant factors for terminal application manufacturers.
Given the different characteristics between electrical switches and optical switches, CPO and OCS are expected to coexist in data center networks. Traditional electrical switching is more suitable for scenarios with significant dynamic traffic flow and frequent adjustments to data transmission paths, while optical switching technology is better suited for relatively stable traffic patterns with clear port mapping relationships and no frequent switching needed, thus avoiding the drawback of long switching times.
In scale-out networks, CPO and OCS can be used in tandem: CPO is expected to accelerate its penetration at the Tor/Leaf layer, where the traffic characteristics are strongly bursty with numerous connections, but the data volume for each link is relatively small. The CPO electrical switch has a faster switching speed, allowing for flexible and rapid responses to traffic demands at the Leaf layer; OCS is anticipated to provide scalable replacements at the Spine layer, where traffic patterns are generally more predictable and stable, enabling OCS to offer stable, high-bandwidth direct connection data flows.
In scale-up networks, there may be competition between CPO and OCS. Currently, the overall implementation pace of OCS is outpacing that of CPO. Risk factors include potential underdevelopment of AI, increasing market competition, and the risk of related technology developments not meeting expectations.
Comments