The surge in AI computing power has introduced terms like CPO, optical modules, and optical communication, often leaving general investors confused. What do these terms mean? Why have they suddenly gained prominence? How are they related to AI investing?
Li Fan, a fund manager at浦银安盛基金, provides clear answers: optical communication is expanding from connections "between servers" to "inside servers," potentially multiplying the market size, with 2027 likely marking a critical starting point for CPO volume growth. This article presents his latest analysis on AI computing power, the optical communication supply chain, and CPO technology trends in a Q&A format, helping investors cut through the complexity and grasp the core investment themes in AI computing.
Q: Many investors struggle to differentiate between "CPO," "optical modules," and "optical communication." Could you use a simple analogy to clarify their relationship? A: "CPO" (Co-packaged Optics) is a next-generation packaging technology that integrates the optical engine with the switch chip to address power consumption and signal integrity issues at high speeds. Think of it as "soldering the translator directly onto the chip." Previously, the translator was separate from the chip, leading to signal loss and higher power usage. CPO brings them together, resulting in faster speeds and lower power consumption.
"Optical modules" are the core product form currently used for optical communication in data centers, responsible for converting electrical signals to optical signals and vice versa. They act as "translators," converting electrical signals into light for transmission over fiber optics, then converting them back at the destination. In today's AI data centers, thousands of optical modules perform this translation work.
"Optical communication" refers to the broader industry direction, encompassing the complete chain from optical fibers to components to systems.
Using an analogy: optical communication is the "highway system," optical modules are the "toll booths/vehicles," and CPO is like "building the factory right at the highway entrance"—reducing transport distance and improving efficiency.
Q: Within the AI investment theme, why should investors care about these concepts? Which key sectors deserve attention? A: As AI computing demand grows, traditional optical modules are evolving towards CPO. This upgrade process is expected to create a multi-fold expansion in market potential. Companies that gain an advantage in this transition could become leaders in the next phase.
An analogy: a large AI model is like a super factory, GPU chips are the machines inside, and optical communication acts as the "conveyor belts" connecting these machines. The speed and efficiency of the conveyor belts determine the factory's overall production capacity. With the explosion in AI computing needs, the number of GPU chips is growing exponentially, but these chips require high-speed interconnects to work together effectively. If the "conveyor belts" are too slow, even numerous machines cannot operate efficiently. Therefore, optical communication has become one of the most critical bottlenecks in AI computing power.
Q: The market often discusses "CPO" and "optical modules" together. Are these two distinct technology paths? From an investment perspective, are they competing or sequential? How should general investors view their investment timelines? A: They represent an "evolutionary and co-existing relationship." Currently, traditional pluggable optical modules remain the primary solution for connections between servers, while CPO is gradually penetrating connections inside servers. We anticipate 2027 will be a key inflection point for CPO adoption. Regarding investment timing: from 2026 to 2027, traditional optical modules will still benefit from high growth in AI computing demand (with a compound growth rate exceeding three times). After 2027, the CPO-related supply chain is expected to see significant volume growth, opening new incremental market opportunities. The two technologies will be relevant at different stages, rather than being a simple substitution.
Q: We often hear the phrase "optical replacing copper," but some argue copper still holds a cost advantage for short-distance interconnects. What is your view on the competition between optical communication and copper interconnects inside AI servers? A: Our assessment is clear: copper interconnects have hit a physical ceiling. As AI computing density increases and transmission rates advance towards 800G/1.6T, copper cables face growing challenges with signal attenuation and power consumption. We expect optical communication to expand from "between servers" to "inside servers," potentially unlocking a market more than five times larger and offering roughly ten times the bandwidth improvement. In practical engineering, copper solutions might still be used for extremely short distances within a rack, but the penetration rate of optical communication will continue to rise.
Q: Beyond CPO, terms like LPO (Linear-drive Pluggable Optics) and OIO (Optical I/O) can be overwhelming for investors. What specific problems do these technologies address? How can investors distinguish between "thematic speculation" and "genuine industry trends"? A: We need to filter for "real industry trends." First, "return to first principles"—assess whether a technology solves a genuine industry bottleneck. Second, "analyze the industry's pain points"—understand the physical limits, power consumption, and cost issues of existing technologies.
Third, "comprehensively evaluate cost-effectiveness"—determine if the new technology offers sufficient advantages over current solutions. Finally, "conduct deep industry verification"—monitor industry conferences like OCP and GTC to see if leading companies are adopting the technology and if there is a clear commercialization timeline. Using these criteria, if a technology remains only conceptual, lacks backing from major players, and has no defined mass-production timeline, investors should be cautious of "thematic speculation" risks.
Q: Concerns about an AI bubble are prevalent. You stated that "the AI industry is seeing the dawn of monetization." What is the basis for this view? What signal does the price increase in cloud computing convey? A: The recent adjustments in US software stocks and the Hang Seng Tech Index do not reflect a destruction of industry value but rather the disruptive impact of booming AI applications on traditional business models. The cloud computing industry, which saw consistent price declines over the past 20 years, has begun to increase prices. Computing power is becoming a core, directly monetizable asset, highlighting the industry characteristic where "computing power equals revenue." This signals that the AI industry is transitioning from an investment phase to a monetization phase.
Risk Disclosure: Funds carry risks; investors should exercise caution. Data source: Wind, as of March 25, 2025. The opinions and commentary provided herein are for reference only and are time-sensitive; they do not constitute any operational advice or recommendation regarding the mentioned securities. Any consequences resulting from investment decisions or actions based on this information are the sole responsibility of the investor. This material is the property of our company; no institution or individual may alter its content in any way that distorts the original meaning without written permission. The fund manager is committed to managing fund assets with integrity, diligence, and responsibility but does not guarantee fund profitability or a minimum return. Investors should understand that fund investments may result in loss of principal. Past performance of a fund is not indicative of its future results. The performance of other funds managed by the fund manager does not guarantee the performance of this fund. The fund operation history in China is relatively short and may not reflect all stages of stock market development. Before investing, investors must carefully read the "Fund Contract" and "Prospectus" legal documents. To purchase this fund, please comply with investor suitability regulations, complete a risk assessment in advance, and purchase fund products matching your risk tolerance based on the results. Mentions of specific stocks, industries, sectors, or indices are for illustrative purposes only and are not investment recommendations; past index performance does not guarantee future returns.