Morgan Stanley Analysis: AI Network Market Heads Toward $70 Billion, Why Does Copper Still Get the First Mover Advantage?

TL;DR

• Morgan Stanley estimates the AI networking market opportunity in 2030 at approximately $70 billion, representing a more than 4x increase from last year's estimate.
• Copper cables will remain dominant in scale-out networking through 2026-2027; CPO is not expected to achieve 20%-30% penetration until 2029-2030.
• Keysight Technologies, Astera, Broadcom, and Semtech are earlier beneficiaries; Corning, Lumentum, and Coherent face later upside.

In its latest report, Morgan Stanley has raised its estimate of the AI networking market opportunity in 2030 to approximately $70 billion and has re-evaluated the role of copper cables in AI clusters.

This is not a story of an imminent CPO breakout. AI clusters are expanding from single racks to multiple racks, requiring denser and higher-speed connections between GPUs, thereby enlarging the overall size of the back-end network. However, as long as power consumption, distance, and bandwidth density haven't truly reached their limits, short-reach connections will maintain strong inertia towards copper cables.

The timeline provided in this report is conservative: CPO penetration in scale-out networking will be near zero in 2026-2027; small-scale deployment begins in 2028; meaningful adoption levels of 20%-30% are not expected until 2029-2030. While the market opportunity has been significantly upgraded, the real impact of optics on scale-out networking requires both larger GPU domains and a more mature supply chain to align.

The $70 Billion Opportunity Comes from Multi-Rack Deployments; Optical Modules Aren't the First to Benefit

The core driver behind this upgrade is the significantly increased demand for connections within servers and between racks as AI clusters expand.

In traditional single-rack scenarios, the short distances between GPUs mean copper cables maintain advantages in cost, latency, and power consumption. For short-reach connections, especially within 7-9 meters, copper remains the most straightforward solution. Over the past few years, stronger SerDes, retimers, and PAM4/PAM6 technologies have continuously extended the lifespan of copper cables, repeatedly pushing back the timeline for optical replacement.

The shift occurs as clusters continue to grow. As training and inference clusters expand from one rack to multiple racks, GPUs need to communicate across racks, and signal speeds advance from 100G to 200G and 400G. With longer distances and higher speeds, challenges related to electrical loss, insertion loss, and noise management increase, pushing copper cables closer to their performance limits.

Back-end network revenue forecast for 2024-2030; scale-out networking revenue rises rapidly, reaching a market opportunity of ~$70 billion in 2030.

For investors, this determines the order of beneficiaries. The primary beneficiaries are not necessarily CPO suppliers, but the chip and module companies that enable copper cables to run faster and farther. The upside for optical engine, passive photonics, laser, and testing equipment companies will become clearer only when multi-rack clusters become more prevalent.

2026-2027 Remains Copper's Window; CPO Won't Boom Until After 2029

CPO's appeal lies in bringing optical components closer to the switch ASIC or compute chip, reducing the propagation distance of high-speed electrical signals on the PCB, thereby improving power consumption and bandwidth density. The difficulty lies in the fact that this is not just a cable swap; it involves changes in packaging, manufacturing, testing, maintenance, and supply chain responsibilities.

This is precisely why CPO will not see a full-scale breakout in 2026. CPO penetration in scale-out networking will be near zero in 2026-2027, with small-scale introduction in 2028. Truly meaningful adoption is not expected until 2029-2030. At that point, if the expansion of multi-rack GPU domains proceeds as planned, CPO penetration in scale-out networking might reach 20%-30%.

CPO penetration forecast: scale-up vs. scale-out; scale-out CPO adoption only rises to 20%-30% in 2029-2030.

This leaves a window of at least two years for the copper supply chain. Astera Labs' Scorpio X-Series has entered initial volume shipments. Broadcom has connectivity opportunities in the AMD MI400/Helios and custom ASIC ecosystems. Semtech participates in the transition phase with its CopperEdge low-power copper cables and linear optical solutions.

More importantly, copper and optics are not simple substitutes. Hyperscalers will mix and match DAC, ACC, AEC, AOC, NPO, and CPO based on distance, power, cost, maintainability, and reliability. Short-reach, intra-rack, and near-rack connections may retain significant copper usage, while CPO will likely tackle the high-density, longer-distance, and more power-constrained segments.

Nvidia's Roadmap Boosts Optical Demand, but Timing Depends on Platform Rollout

The point at which CPO truly becomes important is directly linked to Nvidia's next-generation AI platform roadmap.

According to Nvidia's official technical blog, the Vera Rubin Ultra NVL576 will connect 8 racks of 72 GPUs into a 576-GPU NVLink domain, using copper and direct optical connections. The Feynman-era Kyber NVL1152 is designed for even larger-scale interconnection and will also use similar direct optical schemes.

As the GPU domain expands, the demand for optical engines will not just increase linearly. The report estimates that the number of optical engines per GPU could rise from approximately 2 today to a range of 35-70. In other words, once the architecture shifts, the optical content per GPU will increase significantly.

Comparison of XPU cluster size and OE demand; as the GPU domain expands from 72 to 576/1152 GPUs, the number of OEs per GPU increases from 2 to 17-70.

This is also why Corning (GLW), Lumentum (LITE), and Coherent (COHR) are included in this main narrative. Corning benefits from passive photonics and glass-related content; Lumentum and Coherent are more tied to lasers, optical engines, and optical components. After incorporating scale-out CPO adoption rates into its model, Morgan Stanley indicates that the earnings upside for these companies depends more heavily on the adoption cadence.

However, this remains an upside "if adoption occurs," rather than realized revenue. There is disagreement in the market regarding Nvidia's roadmap itself. Some industry analysts suggest certain configurations of Kyber or Rubin Ultra might be delayed, while Nvidia maintains the roadmap is unchanged. For the optical supply 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 similar connectivity pathways.

Keysight Resembles a "Pick-and-Shovel" Play; Test Equipment Doesn't Bet on One Single Route

Within this narrative, Keysight Technologies' (KEYS) logic differs from that of optical module companies. It doesn't have to bet on whether copper or CPO will ultimately win because the more diverse AI network architectures become, the greater the demand for test and validation.

Currently, there is no unified standard for AI back-end networking. Nvidia has NVLink and its future expansion roadmap, while the non-Nvidia camp includes UALink, SUE, PCIe, and various proprietary interconnect schemes from different cloud providers. Each architecture requires testing for signal integrity, bit error rate, interoperability, power efficiency, and reliability.

According to Investing.com, Morgan Stanley has upgraded Keysight from Equalweight to Overweight, raising its price target from $350 to $400, citing AI investments, diversifying network architectures, and increased testing requirements for 800G, 1.6T, and 3.2T. Keysight's AI-related revenue accounts for a mid-teens percentage of its total revenue.

In contrast, the upside for optical component companies is more concentrated on CPO adoption rates and the cadence of specific platforms. If Nvidia's roadmap progresses smoothly, Corning, Lumentum, and Coherent will benefit more directly. If copper cables continue to extend their lifecycle through 2026-2027, the short-term certainty for Astera, Broadcom, and Semtech is higher.

CPO Will Eventually Move to Center Stage, but Hyperscalers Aren't Ready for a Leap Yet

The counter-intuitive aspect of this report is that it simultaneously acknowledges CPO's long-term core position and emphasizes that the short-term role of copper cannot be underestimated.

The obstacles facing CPO are not insignificant. Hyperscalers fear supplier lock-in; deep integration of optical components into switch or compute packages complicates subsequent replacements, repairs, and multi-vendor sourcing. Manufacturing yields, thermal management, maintainability, and quality risks also impact the adoption pace. If the cost premium isn't offset by power savings and bandwidth density improvements, adoption will be delayed.

Architectural divergence also exists. While Nvidia's roadmap may drive a higher proportion of optical connections, proprietary architectures like Google's TPU use different topologies that might reduce reliance on traditional CPO schemes. While the non-Nvidia XPU ecosystem creates opportunities for companies like Broadcom and Astera, the lack of standardization means the supply chain cannot rapidly scale based on a single solution.

Therefore, the upward revision of the $70 billion market opportunity is more about the expansion of the entire AI back-end networking pie than any single technology route having secured victory. In 2026-2027, copper will remain dominant for intra-rack and short-reach scenarios. Optics will begin moving to a more central position after 2028. By 2029-2030, CPO may achieve truly meaningful penetration in scale-out networking. The most common market misconception is equating "CPO will eventually arrive" with "CPO is about to explode."