By Santhosh Viswanathan
A health worker in a tier-3 district in Uttar Pradesh struggles to process an AI-generated tuberculosis alert during a server outage. A farmer in Andhra Pradesh abandons an agricultural advisory app because its recommendations are in English rather than Telugu. On a factory floor in Ho Chi Minh City, a delay in cloud-based defect detection allows faulty products to pass unchecked. These scenarios point to a common challenge for AI deployment across Asia Pacific: even the most capable models are of limited use when connectivity, language or latency prevents them from working where they are needed.
This is where edge computing and AI PCs can make a difference. By running AI workloads directly on devices or local systems, they can enable faster responses, support regional languages and reduce dependence on continuous cloud connectivity. A diagnostic tool can process an alert locally, an agricultural app can deliver advice offline, and a factory vision system can identify defects in real time. The challenge is no longer just about building more powerful AI models, but about creating the infrastructure that makes them useful at the point of need.
A region that demands a different architecture
Asia Pacific is the world’s most linguistically, economically, and infrastructurally diverse technology market. India has 22 scheduled languages and 650 million internet users, many accessing digital services on affordable devices in regional languages. Indonesia’s archipelago geography makes cloud-dependent AI structurally unreliable for most of its population. Malaysia’s data sovereignty requirements mean sensitive citizen data must stay within domestic environments. Vietnam’s factory floors need real-time inference where cloud latency is a disqualifier, not an inconvenience.
The answer is not to choose between cloud and device. It is to build AI as a continuum—from data centres to edge servers to AI-capable devices—where each workload runs on the most appropriate tier based on connectivity, latency, data sensitivity, and cost. This is what hybrid AI computing makes possible. The AI PC—integrating CPU, GPU, and NPU to deliver up to 180 TOPS of on-device compute—is the device that makes the last mile of that continuum viable, bringing genuine AI capability to the point of use without cloud dependency, in any language, at any connectivity level.
India is already building some of the most thoughtful sovereign AI infrastructure in the world to support it. Bhashini enables real-time translation across 36 Indian languages; BharatGen Param2 17B supports 22 Indian languages with applications already deployed in agriculture, governance, and defence. The IndiaAI Mission and Semicon 2.0 are building the compute and policy foundation that gives these models reach. Similar initiatives are gaining momentum across Southeast Asia—a shared recognition that AI built for Asia must start with Asia’s constraints.
Sustainability cannot be an afterthought
As AI infrastructure scales across Asia Pacific—a region already facing acute energy stress, with data centre power consumption projected to double globally by 2030—sustainability is not a secondary concern. On-device and edge inference addresses this directly. Purpose-built AI hardware, designed specifically for inference workloads rather than repurposed from training architectures, consumes a fraction of the power that cloud-routing every workload demands. Building distributed AI infrastructure is not just critical to expanding access, but also to building AI responsibly and sustainably.
The opportunity ahead
The question is not whether AI can transform healthcare, public services, retail, and citizen engagement across this region. It demonstrably can. The question is whether the infrastructure beneath those applications is being built with the same ambition.
Asia Pacific has the engineering talent, the policy momentum, and the sovereign AI clarity to answer decisively. The last mile is not the hardest part of the AI journey. It is the most important—because it is where intelligence finally becomes useful to people.
The writer us VP & MD, Asia Pacific & Japan, Intel.
Disclaimer: The views expressed are the author’s own and do not reflect the official policy or position of Financial Express.
