If you can use copper, you use copper: Semiconductor analyst

Key Concepts

• Copper vs. Optics: The debate regarding the physical medium for data center connectivity; copper remains preferred for its cost-efficiency and lower power consumption, while optics are used for high-speed, long-distance requirements.
• Active Electrical Cables (AEC): Cables integrated with semiconductors at each end to extend the signal reach and performance of copper.
• AI Infrastructure: The foundational hardware (networking, processors, accelerators) required to support AI workloads.
• Hyperscalers: Large-scale cloud providers (e.g., AWS, Google, Microsoft) that drive demand for custom AI accelerators and networking components.
• ASIC (Application-Specific Integrated Circuit): Custom-designed chips for specific tasks, such as AI acceleration.

1. The Resilience of Copper in AI Networking

Contrary to the prevailing market sentiment that "copper is dead" in favor of optics, the expert argues that copper remains a fundamental building block for data center infrastructure.

• The "Rule of Thumb": If copper can be used, it should be used because it is fundamentally cheaper and consumes less power than optical alternatives.
• Extending Copper’s Life: Companies like Credo are vital because they manufacture Active Electrical Cables (AEC). By embedding semiconductors at the ends of these cables, they extend the signal integrity and lifespan of copper, delaying the need for more expensive optical solutions.

2. Strategic Market Players

The discussion highlights several key companies positioned to benefit from the ongoing AI infrastructure build-out:

• Credo: Identified as a "pure play" in the networking space. Beyond their copper-extending technology, they are expanding into the optical market, further aligning themselves with the AI trade.
• Networking/ASIC Providers: The expert maintains a "Buy" rating on companies involved in AI accelerators. As the industry shifts from general training to more compute-heavy, dedicated AI tasks, hyperscalers are increasingly developing custom processors. Investors are beginning to recognize the value of these networking components in the AI ecosystem.
• Switch Chip Manufacturers: These companies are described as "cousins" to the networking players. Despite previous market skepticism, their stocks are recovering as investors realize copper connectivity will remain relevant for years. A key catalyst for growth is the upcoming release of the "Scorpio" switch chip, which is expected to open significant new market opportunities.
• Legacy Tech Reinvention: The expert upgraded a long-standing technology firm to a "Buy" rating, citing a significant inflection point in their ability to generate cash flow and return capital to shareholders, proving their ability to reinvent themselves for the modern AI era.

3. Market Dynamics and Investment Perspective

• The "AI Trade" Misconception: The expert argues that the market overreacted by bailing out of copper-related stocks. The current trend is a correction as investors realize that networking is a fundamental requirement for AI, not just the processors (like NVIDIA) themselves.
• Shift in Spending: There is a noticeable shift in how hyperscalers allocate capital. As AI models become more complex, the demand for specialized networking and custom silicon (ASICs) is increasing, providing a tailwind for companies that provide these specific components.

4. Synthesis and Conclusion

The core takeaway is that the "copper vs. optics" narrative is nuanced. While optics are necessary for extreme speeds, copper remains the backbone of data center connectivity due to its economic and power-efficiency advantages. Companies that innovate within the copper domain—specifically those extending its reach through semiconductors or developing advanced switch chips—are essential to the AI infrastructure stack. The expert suggests that the recent market volatility in these stocks was a miscalculation by investors, and the long-term outlook for these networking-focused companies remains strong as they continue to support the evolving needs of hyperscalers.