Compal and Datasection scale AI factory infrastructure across Asia-Pacific

This announcement is part of a broader pattern: second‑tier hardware and ODM players are quietly becoming critical to the AI factory build‑out. Compal’s SGX30-2 platform sits underneath Datasection’s AI cloud, which is being positioned to serve generative models, coding assistants, video generation, and agentic workloads at scale across Asia-Pacific. While NVIDIA still owns the GPU layer, companies like Compal are increasingly controlling the reference designs, rack-scale integration, and power/cooling trade‑offs that determine whether AI capacity can actually be deployed in volume.

For the race to AGI, this matters because compute bottlenecks are as much about integration and supply chain diversity as about raw chip design. When Taiwanese manufacturers, Japanese AI infrastructure providers, and NVIDIA’s ecosystem are all aligned on “AI factories” as a design target, capacity ramps faster and becomes geographically more distributed. That undercuts the idea that only hyperscale U.S. cloud providers will control frontier‑class infrastructure. It also means more regional labs and enterprises can access high‑density systems tuned for large‑scale inference and agentic workloads, which in turn creates more experimental surface area for new model architectures and training regimes.

In short, this is infrastructure plumbing—but it is exactly the kind of plumbing that makes sustained AGI-scale experimentation economically viable outside a handful of U.S. tech giants.