AI data center construction demand remains strong; Morgan Stanley is bullish on the outlook for Ethernet and optical communications, favoring Lumentum (LITE.US) and Coherent (COHR.US) amid the stock pullback.

On Friday, Morgan Stanley noted in its latest research report that although the expansion of artificial intelligence (AI) data centers still faces multiple challenges such as power supply, network connectivity, and infrastructure efficiency, demand for computing power from hyperscale cloud providers remains strong. As AI clusters continue to scale up, the importance of Ethernet, optical communications, and related testing equipment has further increased. The bank is bullish on the development opportunities for companies such as Arista Networks (ANET.US), Cisco Systems, Inc. (CSCO.US), Lumentum (LITE.US), Coherent (COHR.US), and Keysight Technologies (KEYS.US) during the upgrade of AI infrastructure. The Morgan Stanley research team attended the 2026 Open Compute Project Global Summit held in San Jose, California, from October 12 to 15. This year's conference attracted about 15,000 attendees, while another approximately 3,000 potential participants were unable to enter due to venue capacity limits, reflecting the AI infrastructure industry chain's strong interest in next-generation computing and Network-1 Technologies, Inc. However, the bank noted that compared with previous years, this year's OCP conference brought relatively limited entirely new industry information. The main challenges facing data center expansion, including power supply capacity, network architecture, and optical communications technology roadmaps, had already been widely discussed at multiple previous industry conferences. Nevertheless, Morgan Stanley believes that this conference once again made clear an important trend, namely that hyperscale cloud providers are still actively seeking to expand computing scale, and demand remains robust for power, GPUs, CPUs, network connectivity, and optical communications equipment. Ethernet accelerates into AI clusters; Arista and Cisco Systems, Inc. poised to benefit Morgan Stanley believes that AI network infrastructure is shifting from simply pursuing higher bandwidth to simultaneously emphasizing computing resource utilization, network stability, and overall operational efficiency. As AI training and inference clusters continue to expand, networks not only need to connect more computing accelerators, but also need to reduce system interruptions and improve the actual utilization efficiency of existing GPUs and CPUs. Against this backdrop, Ethernet is gradually expanding from traditional data center networks into Scale-up networks within AI clusters. Scale-up networks are mainly used to connect multiple accelerators within the same computing system or cluster so that they can efficiently collaborate on AI workloads. Morgan Stanley pointed out that as Ethernet switches supporting Scale-up begin shipping and the industry continues to promote unified network standards, the application prospects for Ethernet in AI infrastructure are becoming clearer. This will expand the potential market space for Arista Networks and Cisco Systems, Inc. However, the bank also cautioned that the promotion of Ethernet architecture does not necessarily mean that NVIDIA Corporation's (NVDA.US) competitiveness in the AI networking market will decline. NVIDIA Corporation also supports multiple network operating systems and, by collaborating with industry chain partners, is expanding its Network-1 Technologies, Inc. ecosystem. Therefore, the key to future industry competition lies not only in which technology standard gains market recognition, but also in the mass production deployment of specific products and which equipment suppliers customers ultimately choose. In addition, growing AI inference demand is also expanding the scope of application for networking equipment. At the conference, NVIDIA Corporation introduced storage networks optimized for key-value cache (KV Cache), as well as Scale-in network architecture connecting users, models, and data. Morgan Stanley believes that these emerging applications are expected to create more opportunities for Arista and Cisco Systems, Inc., so that their potential market is no longer limited to high-speed interconnection within AI training clusters. Optical communications upgrades are imperative; Lumentum and Coherent usher in new opportunities In addition to Ethernet, optical communications technology was also an important focus of discussion at this year's OCP conference. As AI clusters expand, traditional electrical connections face increasingly greater limitations in transmission distance, power consumption, and bandwidth density, and the application of optical interconnect technology within AI systems is becoming more important. Morgan Stanley believes that the penetration of optical communications into Scale-up networks has become a long-term trend, and near-packaged optics (NPO) and co-packaged optics (CPO) will be important technology routes driving this change. Among them, NPO places optical components close to compute or switch chips, while CPO further integrates the optical engine and processing chip into the same packaging system to shorten electrical signal transmission distance and improve power performance. However, the industry has not yet reached a unified conclusion on the best application scenarios for NPO and CPO. According to LightCounting data cited in the research report, customers more closely tied to the NVIDIA Corporation ecosystem may adopt CPO earlier, while other large cloud computing companies are still actively exploring different technology solutions such as NPO. The institution expects that by 2031, in the 3.2Tbps high-speed port market with single-channel rates reaching 400Gbps, CPO and NPO will become the main technology forms. Morgan Stanley believes this trend is favorable for optical communications companies Lumentum and Coherent. However, the expansion of the overall optical communications market does not mean that all suppliers will achieve the same proportional revenue growth. The laser configuration, number of optical components, and supplier selection adopted by different architectures will still affect the ultimate degree of benefit. Against the backdrop of recent sell-offs in related stocks, the bank is particularly focused on investment opportunities in Lumentum and Coherent, believing that subsequent commercialization progress in NPO technology could become a new catalyst for both companies. But the bank emphasized that this year's OCP conference did not materially change its previous investment judgment on the two companies. Meta discloses CPO reliability progress; laser supply remains key Regarding the commercialization of CPO technology, the latest test results disclosed by Meta Platforms (META.US) drew attention. According to Morgan Stanley, Meta said that its relatively mature laboratory CPO system improved interruptions by about 6 times compared with traditional pluggable optical modules. At the same time, in tests totaling 1.35 million system operating hours, the related systems did not experience failures that could not be repaired through maintenance. The bank believes that this result enhances the feasibility of large-scale future deployment of CPO technology, but different suppliers and subsequent product platforms still need to complete reliability verification separately, and the existing test results cannot be directly generalized to the entire industry. At the same time, traditional pluggable optical modules will not quickly exit the market. During the conference Q&A session, NVIDIA Corporation mentioned that USFP products are expected to enter the production stage in 2027, and listed Cisco Systems, Inc. as one of the related development partners. Morgan Stanley pointed out that optical communications solutions based on electro-absorption modulated lasers (EML) still face supply chain challenges. At present, 200G EML supply is already tight, and as the industry further moves toward 400G EML, procurement difficulty for related products may increase. On external lasers, Lumentum said its ELSFP modules have already entered production. However, whether to deploy lasers on the equipment panel or integrate them inside the system remains an important industry debate. The former facilitates on-site maintenance and component replacement, while the latter may have advantages in power consumption and assembly complexity. The bank believes that investors need to pay attention to the laser solutions ultimately adopted by different platforms, as well as the value of optical components contained in each architecture, rather than simply assuming that growth in the CPO market will benefit all laser suppliers simultaneously. Demand grows for cross-data-center interconnection of AI clusters; Ciena and Keysight Technologies also in focus As AI computing tasks become increasingly complex, some cloud computing providers have begun exploring collaborative computing across data centers, namely Scale-across network architecture. This model requires connecting computing resources in different geographic locations, placing higher demands on network latency, congestion management, and failure recovery capabilities. Morgan Stanley believes that cross-data-center interconnection will increase demand for long-distance optical transmission equipment, especially benefiting Ciena (CIEN.US). At the same time, different vendors are exploring more flattened network structures to reduce switching layers, lower power consumption, and improve network efficiency. For example, the multi-plane network architecture demonstrated by Meta reduced switching layers from three to two, while the hardware fault detection and traffic rebalancing technology introduced by NVIDIA Corporation can shorten the related processing time to within 3 milliseconds. Morgan Stanley pointed out that these technological advances mean that future growth in the networking equipment market cannot be judged solely based on bandwidth demand, but also requires consideration of the amount of networking equipment required per accelerator and changes in overall system architecture. In addition, because AI data centers are simultaneously testing multiple networking and optical communications solutions, demand for testing and verification of related equipment will also increase. The bank believes that Keysight Technologies is expected to benefit from the increasing diversification of optical communications architectures and the expansion of high-speed optical component testing scale. AI networking market still has room for growth, but supplier competition will become more complex Morgan Stanley believes that the 2026 OCP conference did not bring technological breakthroughs sufficient to completely change market expectations, but it further confirmed that demand for AI infrastructure remains strong. As hyperscale cloud providers continue to expand computing deployment, network connectivity, power efficiency, and optical communications technology will become important factors determining the pace of AI data center expansion. In terms of specific investment directions, the bank believes that in the short term, the expansion of the Ethernet Scale-up market is most positive for Arista Networks and Cisco Systems, Inc., while the continued development of NPO and CPO technologies provides new growth opportunities for Lumentum, Coherent, and Keysight Technologies. However, the bank cautioned that the industry is promoting more open network standards and interoperable hardware platforms, and customers also want to reduce the number of devices and system power consumption. Therefore, expansion of the AI networking market size does not necessarily translate directly into proportional revenue growth for all equipment suppliers. In the future, whether suppliers can secure orders in specific customer deployments, maintain a high share of product value in new architectures, and continuously improve equipment energy efficiency and operational stability will become key to determining their long-term competitiveness.