Connect | APAC // Debate: AI compute across training and inference - Can data centers keep pace?
- —
- Compute, Storage & Networking
- Open Session
Speakers
Luke Mackinnon
SVP and Managing Director, Asia, NEXTDC
Rowan Peck
Director, Mission Critical Systems
Theo Krewinkel
Senior Solution Architect
Compute, Storage, and Networks Editor, DCD
Across APAC, data centers have scaled to support AI training through large, centralized campuses built for high-density workload. As AI adoption grows, the compute that supports it is becoming more complex and evolving rapidly. Rising rack density, power demand, and cooling requirements are putting pressure on infrastructure not designed for this pace of change. At the same time, inference is taking a larger share of demand, introducing different requirements around latency, utilisation, and deployment. The question is whether data centers can keep pace with how AI compute is evolving across both training and inference.
This session will explore:
- How AI workloads are changing compute requirements and deployment models
- Whether data centers can support rising density and hardware shifts
- What needs to change across power, cooling, and overall infrastructure
- How inference creates different requirements for latency, utilization, and distribution
- What this means for cost, capacity planning, and investment across APAC