In its latest report, Morgan Stanley has pegged the 2030 AI Scale-Out Network market opportunity at around $700 billion and has brought copper cabling back into the spotlight within AI clusters.

This is not a story of "CPO exploding immediately." AI clusters are transitioning from single-rack to multi-rack setups, requiring more dense and high-speed connections between GPUs, thereby amplifying the demand on the backend network. However, until power consumption, distance, and bandwidth density approach their limits, copper cabling will still have a strong presence in short-distance connections.

The timeline provided in this report is rather restrained: in 2026-2027, the CPO penetration rate in Scale-Out Networks is close to zero; a small introduction begins in 2028; and it may only reach a meaningful level of 20%-30% by 2029-2030. While the market opportunity has been significantly adjusted upwards, optical technologies will only truly capture the lion's share of Scale-Out Networks when the larger GPU domain and a more mature supply chain are simultaneously in place.

A $700 Billion Opportunity Comes from Multi-Rack Setups; Optical Modules Aren't the First to Enlarge

The core of this upward revision lies in the increased demand for connections within servers and between racks as AI clusters expand.

In the traditional single-rack scenario, where the distance between GPUs is short, copper cabling still holds advantages in cost, latency, and power consumption. For short-distance connections, especially within 7-9 meters, copper cabling remains the most direct solution. Over the past few years, stronger SerDes, retimers, PAM4/PAM6, and other technologies have continuously extended the lifespan of copper cabling, repeatedly delaying the point of optical substitution.

The shift occurs as clusters continue to grow. Training and inference clusters are expanding from one rack to multiple, requiring inter-rack communication between GPUs, and signal speeds are advancing from 100G to 200G and 400G. As distances increase and rates rise, electrical losses, insertion losses, and noise management difficulties all escalate, bringing copper cabling close to its performance limits.

For investors, this determines the benefit order. The first to benefit may not necessarily be the CPO supplier, but rather the chip and module companies that can make the copper cable run faster and farther; it will not be until multi-rack clusters become more widespread that the elasticity of optical engines, passive photonics, lasers, and testing equipment will become more pronounced.

2026-2027 Is Still the Copper Cable Window, CPO Explosion to Occur Post-2029

The attractiveness of CPO lies in bringing optical devices closer to switching chips or computing chips, reducing the transmission distance of high-speed electrical signals on the board, thereby improving power consumption and bandwidth density. The challenge is that this is not just about changing a wire, but rather altering the packaging, manufacturing, testing, maintenance, and supply chain responsibilities.

This is also why CPO will not fully explode in 2026. The penetration rate of CPO in scaled networks will approach zero in 2026-2027, see a slight introduction in 2028, and meaningful adoption is expected only in 2029-2030. By then, if the multi-rack GPU domain expansion progresses as planned, CPO's penetration rate in scaled networks may reach 20%-30%.

This leaves at least a two-year window for the copper cable chain. Astera Labs' Scorpio X-Series has already entered initial mass production, Broadcom has connection opportunities in the AMD MI400/Helios and custom ASIC ecosystem, while Semtech is participating in the transition phase through the CopperEdge low-power copper cable and linear optics solutions.

More importantly, copper cables and optics are not simply a substitution relationship. Large cloud players will use a mix of DAC, ACC, AEC, AOC, NPO, and CPO based on distance, power, cost, maintainability, and reliability. Short-distance, intra-rack, and near-rack connections may still heavily rely on copper cables, with CPO taking on more high-density, long-distance, and higher power-consumption scenarios.

NVIDIA Roadmap Boosts Optical Demand, But Platform Deployment Pace Key

The significance of CPO is directly related to NVIDIA's next-generation AI platform roadmap.

The official NVIDIA technical blog shows that the Vera Rubin Ultra NVL576 will consist of 8 72-GPU racks forming a 576-GPU NVLink domain, utilizing copper cables and direct optical connections; the Feynman-era Kyber NVL1152 is aimed at larger-scale interconnects and will utilize a similar direct optical approach.

Following the expansion of the GPU domain, the demand for optical engines will not just increase linearly. In this calculation, the number of optical engines per GPU could increase from the current approximately 2 to a range of 35-70. In other words, once the architecture switch occurs, the optical content will significantly rise.

This is also why Corning (GLW), Lumentum (LITE), and Coherent (COHR) are included in this narrative. Corning benefits from passive photonics and glass-related content, while Lumentum and Coherent are more associated with lasers, light engines, and optical devices. With scaled CPO adoption rates integrated into the model by Morgan Stanley, the earnings elasticity of the relevant companies depends more on adoption pace.

However, this is still elasticity "if adoption occurs" rather than realized revenue. NVIDIA's roadmap itself has market divergences, with some industry analysts suggesting that parts of Kyber or Rubin Ultra configurations may be delayed, while NVIDIA has responded that the roadmap remains unchanged. For the optical chain, the key is not the name of a single product generation but whether the large GPU domain enters mass production as planned, and whether the non-NVIDIA XPU ecosystem adopts a similar connectivity path.

Keysight Technologies is more like a "shovel seller," testing equipment without betting on a single path

In this narrative, Keysight Technologies (KEYS) differs from logic and optical module companies. It does not have to bet on whether copper cables or CPO will ultimately win because the more AI network architectures there are, the higher the testing and validation requirements.

The current AI backend networks have not yet standardized into one type. NVIDIA has NVLink and future expansion roadmaps, while the non-NVIDIA camp has UALink, SUE, PCIe, and various cloud providers' proprietary interconnect solutions. Each architecture requires signal integrity, bit error rate, interoperability, power consumption, and reliability testing.

According to Investing.com, Morgan Stanley has upgraded Keysight Technologies from Equalweight to Overweight, raising the target price from $350 to $400. The reasons include AI investments, network architecture diversification, and the increasing demand for 800G, 1.6T, and 3.2T testing. Keysight Technologies' AI-related revenue accounts for a mid-teens percentage of total revenue.

In contrast, the elasticity of the Optical Devices companies is more focused on CPO adoption rate and specific platform cadence. If NVIDIA's roadmap progresses smoothly, Corning, Lumentum, and Coherent will directly benefit more; if copper cables continue to be extended in 2026-2027, Astera, Broadcom, and Semtech's short-term certainty will be higher instead.

CPO Will Eventually Take a Core Position, But Cloud Providers Are Not Ready for a One-Step Transition Yet

The counterintuitive aspect of this report is that it acknowledges that CPO will eventually take a core position while also emphasizing that copper cables should not be underestimated in the short term.

CPO faces significant obstacles. Hyperscale cloud providers are concerned about vendor lock-in; once optical devices are deeply integrated into switching or compute packages, subsequent replacements, maintenance, and multi-vendor procurement will become more complex. Manufacturing yield, thermal management, maintainability, and quality risks will also influence the adoption pace. If the cost premium is not offset by power savings and bandwidth density improvements, adoption will be delayed.

There are also architectural divergences. NVIDIA's roadmap may drive a higher proportion of optical connections, but internally developed architectures like Google TPU use different topologies, potentially reducing reliance on traditional CPO solutions. While the non-NVIDIA XPU ecosystem creates opportunities for companies like Broadcom and Astera, the lack of standardization also implies that the supply chain will struggle to scale rapidly under a single approach.

Therefore, the $70 billion market adjustment seems more like the AI backend network pie being enlarged, rather than a single technological path already securing victory. In 2026-2027, copper cables will still dominate within the rack and short-distance scenarios; after 2028, optics will begin to take on a more core position; by 2029-2030, CPO may finally achieve meaningful penetration in scaled networks. The most misconstrued aspect of the market is equating "CPO will eventually arrive" directly with "CPO will immediately break out."

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia