Staff writer, with CNA
The global artificial intelligence (AI) industry is entering a new phase in which growth is shifting beyond AI model development to infrastructure expansion and real-world deployment on devices and across industries, Market Intelligence and Consulting Institute (MIC) director-general Chris Hung (洪春暉) said yesterday.The current wave of AI development is “no longer just about models,” but increasingly about the broader ecosystem needed to support deployment, Hung said at the Computex Best Choice Award ceremony in Taipei.“AI applications are no longer confined to the cloud and are beginning to enter edge computing, end devices and manufacturing sites,” he said.
A graphic showing the letters for artificial intelligence and the logo for the AI ChatBot smartphone application are pictured on Tuesday last week.
Referring to a layered AI infrastructure framework popularized by Nvidia chief executive officer Jensen Huang (黃仁勳), Hung said the AI ecosystem spans energy, chips, infrastructure, models and applications.AI has evolved into “a new generation of national infrastructure,” Hung said.
The expansion of AI workloads is driving demand for servers, networking, cooling systems and power infrastructure, as data centers scale up to support AI training and inference, he added.Citing projections by the government-funded institute, Hung said global AI server shipments would continue rising through 2030, while AI-related semiconductors are expected to account for nearly half of the overall semiconductor market by 2028.Meanwhile, the focus of AI computing has been gradually shifting from cloud-based training to edge and end-user applications, with AI PCs expected to reach a “rapid adoption stage” this year as more devices become capable of running AI workloads locally, he said.“AI has become a full ecosystem,” with the technology increasingly spreading into transportation, healthcare, manufacturing and office applications, Hung said.AI-enhanced Wi-Fi development is also evolving from a focus on peak speeds to improving reliability and low-latency performance in real-world environments, he added.
