Why APAC Enterprises Are Moving AI Infrastructure to the Edge to Cut Inference Costs

AI spending across Asia Pacific continues to climb, yet a growing number of companies are struggling to extract measurable value from their AI investments. A core part of the problem lies in the infrastructure underpinning these projects — most systems are simply not engineered to run inference at the speed or scale that real-world applications demand.

📌 Industry studies show that many enterprise projects miss their ROI goals even after heavy investment in GenAI tools — largely due to infrastructure limitations that throttle performance and inflate operational costs.

This gap makes clear just how significantly AI infrastructure shapes performance outcomes, cost efficiency, and an organisation's ability to scale deployments across the region.

🧠 Akamai's Answer: Inference Cloud Powered by NVIDIA Blackwell

Akamai is moving to close this gap with its newly launched Inference Cloud, built in partnership with NVIDIA and powered by the latest Blackwell GPUs. The premise is straightforward: if the majority of AI applications require real-time decision-making, then those decisions should be executed close to the end user — not routed through distant, centralised data centres.

💡 Akamai's core argument: Distributed inference at the edge can help enterprises manage costs, reduce latency, and sustain AI services that depend on split-second responses — all at production scale.

💬 Why Inference — Not Training — Is the Real Bottleneck

Jay Jenkins, CTO of Cloud Computing at Akamai, spoke with AI News about why this moment is forcing enterprises to fundamentally reconsider how they deploy AI infrastructure — and why inference has emerged as the critical constraint, overtaking model training as the primary operational challenge.