INFINITIX Unveils Next-Gen AI-Stack at COMPUTEX 2025

PR Newswire

Mon, May 19, 2025, 8:56 PM 3 min read

Leading the Next Frontier in GPU Resource Management

, /PRNewswire/ -- INFINITIX Inc., Taiwan's foremost AI infrastructure innovator, today announced the launch of its upgraded AI-Stack platform at COMPUTEX Taipei 2025, unveiling four transformative modules that address the full lifecycle of AI application — from model training to deployment and monetization.

The announcement coincides with INFINITIX receiving two prestigious recognitions: the COMPUTEX Best Choice Award – Computer & System Category and the Taiwan Artificial Intelligence Association's AI Award – Best Solution (Excellence), underscoring the company's role as a global frontrunner in GPU compute resource management.

Driving AI Infrastructure Transformation

With COMPUTEX 2025 centered on the theme "AI Next", INFINITIX is fully aligned with the industry's strategic focus on intelligent computing, robotics, and next-generation technologies. At Booth I0006, Hall 1 of the Taipei Nangang Exhibition Center, INFINITIX showcases how its flagship platform, AI-Stack, is revolutionizing the future of AI infrastructure — ushering in a new era of intelligent and scalable digital transformation.

The company is highlighting four key pillars of innovation:

• AI Infrastructure Innovation

• High-Efficiency AI Computing Management

• AI Industry Enablement

• Sustainable AI Operations

"In today's AI-driven era, GPU compute power has become a critical strategic asset," said Wen-Yu Chen, CEO of INFINITIX. "Our proprietary technology in GPU partitioning, aggregation, and orchestration empowers enterprises to elevate GPU utilization from just 20–30% to over 90%. AI-Stack ensures unparalleled compute efficiency across all major GPU environments."

Introducing Four New Modules: ET, MaaS, GaaS, TaaS

At COMPUTEX, INFINITIX unveiled an upgraded version of AI-Stack with four modules designed to streamline the AI development and deployment journey:

• ET (Elastic Training): Enables intelligent scaling of container resources across training stages and nodes, solving common scheduling bottlenecks.

• MaaS (Model-as-a-Service): Offers one-click model deployment with version control and rolling updates, supporting monetization through marketplace integration.

• GaaS (GPU-as-a-Service): Provides flexible GPU rental options, from single-card sharing to full-machine leasing, with hourly billing and tenant-level accounting.

• TaaS (Token-as-a-Service): Introduces a token-based billing and API management system that enables cost control, usage-based pricing, and revenue sharing for SaaS models.