AI Infra Capex Is Still Climbing, and SG Retail Should Watch the Second Order Names

The headline AI names get all the attention, but the more interesting story for me as a Singapore retail investor sits one layer underneath. Power, cooling, networking, and the memory that actually feeds these compute clusters.

When the hyperscalers keep nudging their capex guidance higher, that spend does not stop at the GPU. It flows into data center build outs, electrical and thermal gear, and the storage stack. Those second order names often move later and with a lot less crowding than the obvious leaders.

My approach is simple. I am not trying to call the exact top of the AI trade, because I will get that wrong. I am asking which parts of the supply chain still have room before expectations get fully stretched.

From an SG angle I also watch how this feeds back into our own market, mainly through the data center exposed REITs and the regional tech supply chain that quietly rides the same wave.

Not financial advice, just how I am framing it right now. Curious what the community thinks the real bottleneck becomes over the next year, chips, power, or memory.