I recently discovered a quite interesting phenomenon — the entire investment market is re-exploring the next wave of high-growth targets. In the past, AI server hardware experienced a surge, but the base was already quite high, and now capital is shifting toward fields like silicon photonics, which are just beginning to emerge. Honestly, if you haven't yet understood what CPO and silicon photonics are, it's definitely time to get familiar with them now.

Simply put, as AI computing demands explode, traditional copper wire data transmission has hit a bottleneck — too hot, too slow, and too energy-consuming. The emergence of silicon photonics and CPO aims to solve this problem, with the core idea of replacing "electric" signals with "light" to transmit data. It sounds sci-fi, but this isn't the future — it's happening right now.

Silicon photonics technology basically involves shrinking the bulky optical components—like lasers, detectors, and modulators—down to the size of microprocessors and integrating them onto silicon substrates. Meanwhile, CPO (Co-Packaged Optics) moves the optical transceiver modules directly next to the CPU or GPU, packaging them on the same board. What's the benefit of doing this? It can save over 30% of energy and significantly boost transmission speeds. That’s why these two concepts are often linked — silicon photonics is the core technology of CPO, and CPO is currently the most promising application of silicon photonics.

The industry landscape is quite clear now. U.S. tech giants hold patents, chip design, and communication protocols, while Taiwan leverages its world-leading semiconductor manufacturing and packaging capabilities to form an integrated foundry ecosystem. This is also why Taiwan’s silicon photonics concept stocks are so closely watched.

Looking from the upstream down, TSMC is not just manufacturing chips but also defining CPO packaging standards. Its COUPE platform is considered central to silicon photonics development, and the CPO packaging technology expected to be mass-produced this year is led by TSMC. Following TSMC’s "big alliance" members — such as InnoLight-KY and ASE Group — these advanced packaging and testing companies will be prioritized for certification and orders.

In the optical components sector, LianYa provides InP epitaxy, which is a key material for external light sources in CPO. Optical Ring and China Star Optical have entered the AI supply chain through CW laser foundry services. Midstream, Advanced Optoelectronic and TSMC are collaborating deeply on FAU fiber array technology, which is crucial for bringing light into chips. The market generally believes that Advanced Optoelectronic benefits the most from this "interface."

In the U.S. stock market, Broadcom leads in CPO, with its Tomahawk series launched in 2026 becoming the standard for AI data centers. Marvell is also strong in high-speed optical interconnect chips and has deep cooperation with NVIDIA. Microchip recently acquired DustPhotonics, gaining direct control of photonic integrated circuit technology, with complete solutions from 800G to 1.6T.

However, investing in silicon photonics concept stocks also involves some risks. First is yield issues — CPO involves packaging optical components with chips, and if any part has problems, the expensive GPU could be scrapped. When reviewing financial reports, pay close attention to gross margin trends; if revenue increases but gross margin declines, it may indicate yield struggles. Second, there’s competition from LPO (Laser Plug and Play Optical modules), a more traditional and cheaper plug-in module that’s easier to maintain. Before the widespread adoption of 1.6T solutions, LPO might take a significant market share.

Another key point is to look at the actual optical communication revenue proportion of a company. If a company claims to be a silicon photonics concept stock but has a very low optical communication revenue share, be cautious — it might just be riding the trend. Lastly, geopolitical factors matter — the U.S. broadband infrastructure plans will directly impact optical communication demand, and U.S.-China tech tensions could impose restrictions on advanced technologies like silicon photonics, adding uncertainty.

Ultimately, silicon photonics isn’t a short-term theme but a structural growth trend over the next 5 to 10 years. One simple principle suffices — in the U.S., focus on "standard-setting," while in Taiwan, focus on "supply chain performance." While chasing new themes, it’s essential to return to fundamentals: prioritize companies that have been certified by major players and show a clear increase in optical communication revenue share. Only then can you avoid the noise in this fast-paced race and seize truly valuable silicon photonics concept stocks.