CSCO
NVDA
SMCI
Not just more GPUs
Cisco is expanding its Secure AI Factory with Nvidia, and this time Super Micro Computer is in the mix too. The add-on includes high-density liquid- and air-cooled GPU systems, which is a fancy way of saying: the AI arms race is now partly a plumbing race.
Why investors should care
As AI clusters get denser and hotter, the old “just throw fans at it” approach starts looking about as useful as a desk fan in a sauna. Cisco says power and cooling now help decide where infrastructure can even be built, which tells you how quickly this market is maturing.
What’s in the new setup?
- rack-to-fabric liquid cooling
- Cisco liquid-cooled networking gear
- Supermicro liquid-cooled servers
- support for Nvidia’s Vera Rubin NVL72 and HGX Rubin NVL8 platforms
The bigger AI supply-chain story
This is where the story widens beyond Nvidia. TrendForce expects liquid cooling penetration for AI chips to jump from 33% in 2025 to 53% in 2026, then 60% in 2027. That’s not a niche upgrade anymore — that’s a regime change.
And the ripple effect is obvious: the more powerful the chips get, the more money flows into networking, power management, and cooling. In other words, the AI gold rush is getting a lot more like digging ditches — just with nicer servers.
Big picture: Cisco’s move suggests the next phase of AI infrastructure won’t be defined by compute alone. It’ll be defined by whether the data center can keep its cool long enough to run the thing.