TAIPEI, Taiwan – May 19, 2025 — Intel marked its 40th anniversary in Taiwan with major product announcements at Computex 2025, unveiling a powerful new lineup of GPUs and AI accelerators aimed at professionals, AI developers, and workstation users.
Headlining the launch is the introduction of the Intel® Arc™ Pro B60 and Arc™ Pro B50 GPUs—new entries in Intel’s Arc Pro family designed for advanced AI inference workloads and high-performance workstations. Alongside the GPU updates, Intel also announced broader availability of its Gaudi 3 AI accelerators, now offered in both PCIe and rack-scale systems, and introduced the Intel AI Assistant Builder, a public tool for building local AI agents optimized for Intel hardware.
Arc Pro GPUs Target AI and Professional Workloads
Built on Intel’s Xe2 architecture, the new Arc Pro B60 and B50 GPUs feature Intel® Xe Matrix Extensions (XMX) AI cores and enhanced ray tracing capabilities. With 24GB and 16GB of video memory, respectively, and support for multi-GPU scalability, the B-Series is purpose-built for architecture, engineering, construction (AEC), and inference-based applications.
Intel says the new cards are certified by leading independent software vendors (ISVs) and support both Windows pro/consumer drivers and containerized Linux stacks to simplify AI development and deployment.
The GPUs will power Intel’s new “Project Battlematrix” workstation platform—an Intel® Xeon®-based system supporting up to eight B60 GPUs for medium-scale models with up to 150 billion parameters and 192GB of video memory.
The Arc Pro B60 will begin sampling through partners like ASRock, Gunnir, and Sparkle in June 2025, while the Arc Pro B50 is expected to ship from Intel-authorized resellers in July 2025.
Gaudi 3 AI Accelerators Go Modular
Expanding Intel’s AI offerings, Gaudi 3 AI accelerators are now available in two form factors:
PCIe Cards: These are aimed at enhancing existing data center infrastructure with scalable inferencing capabilities for models like Llama 3.1 8B and Llama 4 Maverick, and will be available in H2 2025.
Rack Scale Systems: Supporting up to 64 accelerators per rack, these high-density systems offer 8.2TB of high-bandwidth memory, liquid cooling, and an open, modular design to reduce vendor lock-in and simplify deployment.
ADVERTISEMENT
Intel’s Gaudi systems are optimized for low-latency inferencing at scale, reinforcing its commitment to open and flexible AI infrastructure for enterprise and cloud environments.
AI Assistant Builder Now in Public Beta
Rounding out Intel’s Computex announcements is the AI Assistant Builder, a lightweight development framework that allows companies and developers to create custom, locally deployed AI agents on Intel-powered PCs. After its CES 2025 preview, the tool is now available on GitHub, with early implementations already showcased by Acer and ASUS.
A Milestone Year in Taiwan
This product launch coincides with Intel’s 40th anniversary in Taiwan, underscoring decades of partnership with one of the world’s leading tech ecosystems. Intel CEO Lip-Bu Tan emphasized this legacy, stating, “For the past 40 years, the power of our partnership with the Taiwan ecosystem has fueled innovation that has changed our world for the better…Together, we will create great products that delight our customers and capitalize on the exciting opportunities ahead.”
Looking Ahead
With growing demands in AI compute and professional workloads, Intel’s 2025 lineup represents a significant expansion in its GPU and AI accelerator strategy—offering scalable, open, and developer-friendly platforms for the next generation of intelligent computing.
For more information, visit intel.com.
