Intel Fortifies AI and Workstation Landscape with New Arc Pro GPUs and Gaudi 3 Accelerators at Computex 2025
Intel is making a significant push into the AI and professional workstation markets with a suite of new hardware and software announcements at Computex 2025. The company is expanding its Intel Arc Pro GPU lineup with the introduction of the B60 and B50 models, designed to meet the growing demands of AI inference and complex workstation applications. These new offerings, built upon Intel's Xe2 architecture, integrate advanced features such as Intel® Xe Matrix Extensions (XMX) AI cores and sophisticated ray tracing units, promising enhanced performance for creators, developers, and engineers.
Expanding the Intel Arc Pro GPU Lineup
The newly launched Intel Arc Pro B60 and B50 GPUs represent a strategic expansion of Intel's professional graphics portfolio. The B60 model boasts 24 gigabytes (GB) of memory, while the B50 offers 16GB, providing substantial capacity for memory-intensive AI tasks and large datasets. A key feature highlighted is the multi-GPU scalability, which allows for the creation of powerful, interconnected systems capable of handling extremely demanding computational workloads. This makes the Intel Arc Pro B-Series a versatile solution tailored for professionals working in fields such as architecture, engineering, and construction (AEC), as well as for AI developers who require robust performance for model training and inference.
Intel has focused on ensuring stability and performance through extensive Independent Software Vendor (ISV) certifications and optimized software. The Intel Arc Pro B-Series GPUs are compatible with both consumer and professional drivers on Windows, offering a familiar environment for many users. For Linux users, Intel provides a containerized software stack designed to simplify AI deployments and facilitate easier management of complex AI workloads. This approach, combined with high memory capacity and broad software compatibility, positions the Arc Pro B-Series as a scalable and cost-effective choice for businesses looking to enhance their AI and professional computing capabilities.
Further demonstrating its commitment to the professional market, Intel also unveiled Project Battlematrix, a configurable workstation-class platform based on Intel® Xeon® processors. This platform is engineered to reduce friction points for AI developers and is designed to support up to eight Intel Arc Pro B60 24GB GPUs. This configuration enables the development and deployment of medium-sized AI models, up to 150 billion parameters, with an impressive total of 192GB of video random-access memory, catering to the most advanced AI research and development needs.
Vivian Lien, vice president and general manager of Client Graphics at Intel, commented on the significance of the new lineup: "The Intel Arc Pro B-Series showcases Intel’s commitment in GPU technology and ecosystem partnerships. With Xe2 architecture’s advanced capabilities and a growing software ecosystem, the new Arc Pro GPUs deliver accessibility and scalability to small and medium-sized businesses that have been looking for targeted solutions." The Intel Arc Pro B60 GPU is slated for sampling from add-in board partners starting in June 2025, with the Intel Arc Pro B50 GPU expected to be available from authorized resellers in July 2025.
Intel Gaudi 3 Accelerators Expand Deployment Options
Complementing its GPU advancements, Intel is broadening the deployment flexibility for its Intel Gaudi 3 AI accelerators. This expansion is a key part of Intel's comprehensive AI strategy, offering more options for enterprises to integrate powerful AI hardware into their infrastructure. The new deployment options include:
- Intel Gaudi 3 PCIe cards: These cards are designed to support scalable AI inferencing directly within existing data center server environments. Leveraging popular AI models such as Llama, customers ranging from small businesses to large enterprises can benefit from the flexibility to run various model sizes, from Llama 3.1 8B to full-scale Llama 4 Scout or Maverick, thanks to the scalable configurations available. The Intel Gaudi 3 PCIe cards are anticipated to be available in the second half of 2025.
- Intel Gaudi 3 rack scale system reference designs: Built with flexibility and scale in mind, these systems can support up to 64 accelerators per rack and offer a substantial 8.2 terabytes (TB) of high-bandwidth memory. The design emphasizes an open and modular approach to help avoid vendor lock-in. Equipped with liquid cooling, these systems are engineered to deliver strong performance while helping to manage the total cost of ownership (TCO).
The Intel Gaudi rack scale architecture is specifically optimized for running large AI models and excels in delivering real-time inferencing with low-latency performance. These expanded configurations underscore Intel's dedication to providing open, flexible, and secure AI infrastructure solutions. They are designed to support both custom deployments and adhere to Open Compute Project (OCP) designs, catering to the needs of cloud service providers (CSPs) and large enterprises.
Intel AI Assistant Builder Now Publicly Available
Further enhancing the AI ecosystem, Intel announced that Intel AI Assistant Builder is now available in public beta on GitHub. This lightweight, open software framework allows developers to build and run custom AI agents locally on Intel-based AI PCs. Following its initial debut at CES 2025, the availability of AI Assistant Builder on GitHub aims to empower developers and partners to rapidly create and deploy AI agents for their organizations and end customers, potentially accelerating the adoption of AI solutions across various business sectors.
Celebrating 40 Years in Taiwan
These significant product announcements coincide with a special milestone for Intel: the company's 40th anniversary in Taiwan. This occasion highlights four decades of collaboration, innovation, and shared success with Taiwan's dynamic technology ecosystem, particularly around the x86 architecture. Intel CEO Lip-Bu Tan remarked, "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. This week, we are renewing our commitment to our partners as we work to build a new Intel for the future. Together, we will create great products that delight our customers and capitalize on the exciting opportunities ahead."
Intel's presence at Computex 2025, located at 1F, Nangang Hall 1 – J0306, runs from May 20th through May 23rd, showcasing these and other innovations to attendees.
AI Summary
At Computex 2025, Intel has made substantial announcements aimed at bolstering its presence in the AI and professional workstation markets. The company introduced the new Intel Arc Pro B60 and B50 GPUs, built on the advanced Xe2 architecture. These GPUs are specifically engineered with Intel® Xe Matrix Extensions (XMX) AI cores and enhanced ray tracing capabilities, targeting demanding AI inference workloads and professional applications. The Arc Pro B60 comes with 24GB of memory, while the B50 offers 16GB, providing ample capacity for complex tasks. The lineup also emphasizes multi-GPU scalability, making it a versatile choice for creators, AI developers, and engineers. Intel highlighted the optimization of these GPUs for AEC (architecture, engineering, construction) and inference workstations, ensuring stability and performance through a broad spectrum of Independent Software Vendor (ISV) certifications and tailored software. Compatibility extends to both Windows (with consumer and pro drivers) and Linux (via a containerized software stack), simplifying AI deployments. Intel also showcased Project Battlematrix, a configurable Intel® Xeon®-based workstation platform designed to support up to eight Intel Arc Pro B60 GPUs, capable of handling AI models up to 150 billion parameters with a total of 192GB of video RAM. Vivian Lien, vice president and general manager of Client Graphics at Intel, stated that the Arc Pro B-Series demonstrates Intel’s dedication to GPU technology and ecosystem partnerships, offering accessibility and scalability to small and medium-sized businesses. The Intel Arc Pro B60 GPUs are expected to begin sampling in June 2025, with the B50 available in July 2025. In parallel, Intel is expanding the availability of its Intel Gaudi 3 AI accelerators. New deployment options include Intel Gaudi 3 PCIe cards, designed for scalable AI inferencing within existing data center environments, supporting a range of AI models like Llama. These PCIe cards will be available in the second half of 2025. Furthermore, Intel introduced Intel Gaudi 3 rack scale system reference designs, engineered for flexibility and scale, accommodating up to 64 accelerators per rack with 8.2 terabytes of high-bandwidth memory. These systems feature an open, modular design with liquid cooling to optimize performance and TCO. The Gaudi rack scale architecture is optimized for large AI models and real-time inferencing with low latency. These advancements reinforce Intel's strategy for open, flexible, and secure AI infrastructure. The company also announced the public beta release of Intel AI Assistant Builder on GitHub, a framework for creating and running custom AI agents locally on Intel-based AI PCs. This initiative aims to reduce friction for AI developers. The announcements were made during Computex 2025 in Taipei, coinciding with Intel's 40th anniversary in Taiwan, a period marked by significant collaboration and innovation within the region's technology ecosystem. Intel CEO Lip-Bu Tan emphasized the enduring partnership with Taiwan's ecosystem in driving innovation. Intel's presence at Computex 2025, located at 1F, Nangang Hall 1 – J0306, runs from May 20th to May 23rd.