Why CPO matters
AI scaling is no longer compute-bound alone. It is interconnect-bound. As clusters move from tens of thousands to millions of GPUs, copper becomes a bottleneck in:
Power consumption
Latency
Signal integrity
CPO reduces power per bit and enables denser rack-scale designs. If NVIDIA controls optical capacity into 2027–2030, it protects the next scaling phase of Blackwell successors.
Is there a “second curve”?
The first curve was training acceleration.
The second curve is likely:
1. Inference at planetary scale
2. Network dominance via NVLink + optical fabric
3. Full-stack integration from silicon to system to interconnect
If NVIDIA owns the fabric layer, it widens the moat beyond GPUs.
Valuation question
A trillion-dollar valuation requires:
Sustained data centre revenue growth
High-margin inference monetisation
No severe hyperscaler insourcing shock
The risk is not technology. It is capex discipline from customers. If AI spend shifts from “grab compute” to “prove ROI”, multiples compress before fundamentals fail.
Long term, I remain structurally constructive.
But the stock’s path will be cyclical, even if the platform thesis compounds.
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.
