AI Spending Grows in Asia Pacific Despite Infrastructure Challenges
Investment in artificial intelligence across the Asia Pacific region continues to surge, yet many organizations find it difficult to realize tangible returns on their AI projects. A critical factor behind this gap is the AI infrastructure, which often fails to support the speed and scale required for real-world inference workloads.
Industry analyses reveal that numerous AI initiatives fall short of their ROI targets, even after significant expenditure on generative AI tools. This reveals the pivotal role that infrastructure plays in determining AI performance, cost-efficiency, and scalability within the region.
Edge Computing Emerges as a Solution to Inference Bottlenecks
Akamai, in collaboration with NVIDIA, is addressing these challenges through its Inference Cloud platform powered by NVIDIA’s latest Blackwell GPUs. The core concept focuses on relocating AI inference closer to end-users instead of relying on centralized data centers, thereby reducing latency and operational costs.
Jay Jenkins, Chief Technology Officer of Cloud Computing at Akamai, highlights to AI News that enterprises are now forced to rethink AI deployment strategies. He emphasizes that inference—the process of running models for decision-making—has become the primary bottleneck, overshadowing the training phase.
Infrastructure Limitations Hamper AI Adoption
Jenkins explains that the transition from AI experimentation to full-scale production is wider than many organizations anticipate. High infrastructure costs, latency issues, and difficulties scaling models often impede progress, despite strong interest in generative AI.
Many companies still depend on centralized cloud environments with large GPU clusters. However, as AI usage expands, these setups become prohibitively expensive, especially in locations distant from major cloud hubs. Latency adversely affects user experience and business value, while multi-cloud environments and regulatory compliance add layers of complexity to deployment.
Inference Demands Surpass Training in AI Workloads
As AI moves beyond pilot projects to widespread application, daily inference workloads consume more computing resources than periodic training. Across Asia Pacific, enterprises deploy language, vision, and multimodal models in multiple markets, increasing the need for rapid, reliable inference.
This demand intensifies challenges for centralized systems that are ill-equipped to provide the low latency and responsiveness required for AI models operating under diverse linguistic, regulatory, and data conditions.
Benefits of Localized Edge AI Infrastructure
By shifting inference processes closer to users, devices, or autonomous agents, companies can significantly reduce data travel distances, speeding up response times and lowering costs associated with data transmission between cloud centers.
Edge-based AI is crucial for physical systems such as robots, autonomous vehicles, and smart city applications, where millisecond decision-making is vital. Akamai’s analysis shows that enterprises in countries like India and Vietnam particularly benefit from cost reductions when deploying AI inference at the edge, due in part to improved GPU utilization and decreased egress fees.
Industries Leading Edge AI Adoption
Retail and e-commerce sectors are among the early adopters of edge inference due to their need for real-time, personalized customer experiences. Slow AI responses often result in decreased user engagement and lost revenue.
Financial services also depend heavily on low-latency AI for critical tasks such as fraud detection, payment authorization, and transaction scoring. Processing these workloads near the data source enhances speed and ensures compliance with stringent data regulations.
Collaborations Between Cloud Providers and GPU Makers Drive Innovation
The increasing scale of AI workloads has fostered stronger partnerships between cloud providers and GPU manufacturers. Akamai’s collaboration with NVIDIA exemplifies this trend, deploying GPUs, DPUs, and AI software across thousands of edge locations.
This distributed “AI delivery network” approach not only improves performance but also facilitates regulatory compliance by enabling local data processing. Jenkins notes that nearly half of large organizations in the region face challenges due to varying data laws, making edge AI essential.
Security and Compliance in Distributed AI Systems
Security is integral to edge AI infrastructure, with zero-trust models, data-aware routing, and safeguards against fraud and bot attacks embedded into the technology stack. These measures are particularly critical for industries like finance, where data protection is paramount.
Preparing for an Edge-Centric AI Future
Looking ahead, enterprises must adapt to a more decentralized AI lifecycle where models are updated and managed across multiple edge sites. This necessitates enhanced orchestration tools and comprehensive visibility into system performance, costs, and errors.
Data governance will become more complex but manageable as inference processes remain localized, helping organizations navigate diverse regulatory landscapes.
Finally, security protocols must expand to cover every edge location, ensuring robust protection of APIs and data pipelines against emerging threats.
Photo by Igor Omilaev
Related Articles
Obvio Launches AI-Powered Stop Sign Cameras to Enhance Pedestrian Safety
Phi-4 Demonstrates Data-First Fine-Tuning as a Game Changer for Efficient LLM Reasoning
India’s Arya.ag Defies Global Crop Price Decline by Attracting Investors and Maintaining Profitability
Hasbro Confirms Cyberattack, Recovery Expected to Take Several Weeks
