TEL：+86 158 1857 3751 
In 2024, the Ethernet optical transceiver market experienced double-fold growth. Had InP chip suppliers possessed sufficient production capacity, the market size could have doubled again in 2025. However, it is predicted that shortages of EML and CW laser chips will constrain market growth until the end of 2026. By then, demand may gradually ease, helping to restore supply-demand balance. Although forecasts point towards a soft landing, such balance is often achieved through market corrections—and the earlier the correction occurs, the less severe the pain.
The chart below illustrates the impact of artificial intelligence on the Ethernet optical transceiver market, with forecasts covering Retimed optical modules, LPO, and CPO solutions adopted in both Scale-Out and Scale-Up networks. Between 2026 and 2030, demand from AI applications will continue to dominate the market, with growth potentially exceeding expectations. It is fair to say that AI has now fully seized control of market dominance.
Another interpretation of this situation raises the question: Has AI developed self-awareness and begun to influence human decision-making? This might explain why the investment frenzy has been so intense. Even if this is not yet the case, when it does happen, will we even realize it?
The key issue is whether the current investment boom is based on sound financial planning or simply on the fear of missing out. We must confront this situation with a clear and rational mindset.
This investment frenzy was ignited by the early success of ChatGPT, developed through the collaboration of OpenAI and Microsoft. In a bid to defend its dominance in search engines, Google has heavily invested in its Gemini project, which now rivals ChatGPT in performance. AI-enhanced advertising within search engines has become a trillion-dollar profit battleground for Google and Microsoft, justifying, for now, the scale of both companies' investments in AI.
Yet, are there other proven and profitable AI business scenarios? The answer is no. Most application scenarios remain in early development, with success far from guaranteed. It may only be a matter of time before investors lose patience and withdraw their capital. The current level of investor commitment to new technology rivals that seen during the internet and telecom bubbles 25 years ago—unless, of course, a self-aware AI is already manipulating them.
