Intel is back at COMPUTEX 2025 with a full stack of new announcements, marking its 40-year partnership with Taiwan’s tech ecosystem. This year, the company is going big on AI, graphics, and developer tools — rolling out new Intel Arc Pro GPUs, Gaudi 3 PCIe accelerators, and making its AI Assistant Builder publicly available on GitHub.
Intel Arc Pro B-Series: More VRAM, More Flexibility for Workstations and AI
Ahead of COMPUTEX, Intel launched the new Arc Pro B60 and Arc Pro B50 GPUs on May 19, aimed at creators, engineers, and AI developers who need workstation-class performance at a solid price point.
- Arc Pro B60 comes with 24GB of VRAM and is designed for AI inference and 3D workflows in professional environments.
- Arc Pro B50 targets the high-volume design and engineering segment, delivering great value with strong performance for content creation and compute-heavy tasks.
Both GPUs expand the Arc Pro family with larger memory configurations and broader software support, including more ISV certifications. They’re built for prosumers and enterprise users looking to accelerate their workflows without breaking the bank.
The GPUs are already available from Intel’s hardware partners, including major OEMs and system integrators.
Project Battlematrix: Workstation Platform for Medium-Sized AI Models
Intel also previewed Project Battlematrix, a new Xeon-based workstation reference platform that supports up to eight Arc Pro B60 GPUs with a total of 192GB VRAM.
This setup is built for training and running medium-scale AI models, with support for up to 150 billion parameters — making it ideal for researchers and developers who want a cost-effective, in-house AI workstation without going full data center.
Gaudi 3 PCIe: Scalable AI Inference for Data Centers
Intel’s Gaudi 3 AI accelerators are now available in PCIe form, giving businesses an easier way to scale AI inference inside existing server infrastructure.
The key pitch here is flexibility — whether you’re running small LLMs or enterprise-scale models, Gaudi 3 PCIe cards are designed to scale up or down depending on your compute needs.
These cards are a great fit for enterprises that want to tap into AI workloads without having to overhaul their data center layout.
Intel AI Assistant Builder Now on GitHub
To make AI development even easier, Intel announced the public release of its AI Assistant Builder, available now on GitHub.
This toolkit is aimed at helping developers build, fine-tune, and deploy AI agents faster — especially those working with Intel’s Gaudi or Xeon platforms. It includes reference workflows, performance-tuned libraries, and ready-to-use code snippets to get AI projects up and running without starting from scratch.
Final Thoughts
From new GPUs and AI accelerators to open-source tools, Intel’s COMPUTEX 2025 announcements are all about giving developers, creators, and businesses the power to build and scale AI and graphics-intensive workloads.
Whether you’re a workstation user upgrading your rendering rig or an enterprise IT lead looking at scalable AI infrastructure, Intel has something in the mix this year.
