Organizations expanding AI across their operations risk inflating infrastructure costs and energy consumption by applying a one-size-fits-all compute approach, while a widening gap between board-level governance and technological advancement will further constrain adoption.

Executives of Japan-based  information technology firm NTT Data and US-based edge computing company Blaize Holdings, have recently shared their views in separate interviews with TNGlobal in conjunction with the World AI day (July 16).

No single AI compute model fits all workloads: Blaize

Joseph Sulistyo, Senior Vice President of Marketing at the United States-based edge AI computing company Blaize Holdings, emphasized that organizations expanding AI across their operations risk inflating infrastructure costs and energy consumption by applying a one-size-fits-all compute approach.

The right mix of cloud, edge, and hybrid AI deployments should be matched to specific workloads rather than defaulted to a single architecture, he said.

The argument carries particular weight in manufacturing, telecommunications, healthcare, and smart infrastructure, where organizations need AI to process data in real time and close to where it is generated.

For many of these use cases, Sulistyo said routing every workload through the cloud is not the most efficient approach, and edge AI helps reduce latency, optimize bandwidth, and improve responsiveness where timely insights are critical.

Sulistyo noted that enterprise conversations about AI across Asia-Pacific have become more pragmatic over the past year, with fewer organizations questioning AI’s potential and more asking where it can deliver measurable impact and how to scale without adding unnecessary complexity or cost.

“AI sovereignty has also emerged as a strategic dimension influencing partner ecosystem and technology choices across the region,” Sulistyo said.

Looking ahead, he said the trends he expects to have the greatest impact in APAC are the rise of physical AI, such as intelligence applied directly to physical systems and operations, from factory floors to city infrastructure, alongside sovereign AI programs now taking shape across the region. Organizations that establish the necessary infrastructure foundations now will be better positioned to scale sustainably as the ecosystem matures, he said.

Gap between board-level AI governance, technology advancement can slow APAC adoption: NTT DATA

The widening gap between AI governance at board and management level and the pace of technological advancement will continue to constrain the speed and scale of enterprise AI adoption in Asia-Pacific, said Jan Wuppermann, Senior Executive and Head of Service Assurance and Data & Analytics at Japan’s NTT DATA.

Wuppermann said organizations across the region have largely demonstrated the ability to run successful pilots. But turning those pilots into capabilities embedded across the organization and linked to measurable outcomes is the harder task, he added.

The real opportunity lies in integrating AI across infrastructure, data, systems, and workflows in an end-to-end fashion rather than applying it in isolated pockets.

To make that progress securely, responsibly, and at scale, businesses have to secure some requirements: governance and security frameworks, scalable AI infrastructure, trusted data and information architecture, and a workforce that embraces AI at all levels.

The strongest returns, Wuppermann said, are currently being seen in areas where AI can improve the speed, quality, and consistency of core business processes, including customer service, software engineering, IT operations, supply chains, and manufacturing.

“In software engineering AI agents are being incorporated into development and modernization workflows alongside governance controls covering privacy, access, and policy enforcement,” he said, citing NTT DATA’s work with AI coding tool Cursor as one example of this model.

In operational environments, Wuppermann pointed to NTT DATA’s work with Hyster-Yale Materials Handling, applying vision sensors, edge AI, and physical AI models to quality assurance on the factory floor as an example of intelligence being embedded directly into production workflows.

As adoption scales, he said organizations will need to manage trade-offs around infrastructure, cost, cybersecurity, energy use, and skills. Model and infrastructure choices should be guided by the needs of each workload, with many enterprises in the region likely to combine cloud, private environments, and edge computing. AI sovereignty has also emerged as a strategic consideration affecting partner ecosystems and technology choices, Wuppermann said.

Workforce training and change management are as important as the technology itself because humans must remain accountable for outcomes, with clear boundaries around the decisions AI can support or execute.

Looking ahead, Wuppermann said the convergence of agentic AI and physical AI will have significant impact across the region. Organizations establishing the necessary foundations now will be better positioned to scale sustainably.

 

APAC enterprises shifting AI focus from time savings to competitive advantage: LinqAlpha executive