01 / COMPUTE ARCHITECTURE
Workloads are diverging
Right-fit hardware →
AI inference, autonomous agents, quantum simulation and HPC each demand different hardware. Price-performance is workload-specific — and changes with every generation. Lock in the wrong architecture and you lock in inefficiency. Routing inference through external API gateways adds a margin tax, latency variance, and residency risk that co-located serving on owned silicon eliminates.
10×
Performance delta:
optimised vs. generic
02 / GEOGRAPHY & ENERGY
Location is strategy
Sovereign-by-design →
Where compute lives determines who governs it. Sovereignty and compliance are inseparable from physical location. As workloads intensify, energy cost becomes compute cost — and political control requires physical proximity.
40–60%
Energy cost reduction
vs. diesel-backed
03 / VALUE CREATION
The stack is being rebuilt
Vertical stack + ecosystem →
Consultancy is being automated. Vertical SaaS is commoditising. The new value chain runs from infrastructure ownership upward — through platforms, ecosystems and education. The stronghold is at the bottom.
$20M / MW
Cloud revenue
vs. $4M / MW colo