At Computex 2025 in Taipei, Taiwan, Intel made a bold statement in the AI and workstation computing space by unveiling its new Intel Arc Pro B-Series GPUs (B50 and B60) and expanding deployment options for its Gaudi 3 AI accelerators. These announcements, made on May 19, 2025, underscore Intel’s ambition to challenge industry giants like NVIDIA and AMD in the rapidly evolving AI and professional computing markets. As Intel celebrates 40 years of collaboration with Taiwan’s tech ecosystem, these new products signal a strategic push to capture market share in AI inference, workstation performance, and scalable enterprise solutions. Let’s dive into the details of these offerings, their competitive positioning, and the pros and cons for developers, enterprises, and prosumers.

Intel Arc Pro B-Series GPUs: Powering AI Workstations

Intel introduced the Arc Pro B50 and B60 GPUs, built on the Xe2 architecture and tailored for professional workstations and AI inference tasks. These GPUs are designed to deliver high memory capacity and competitive performance at accessible price points, targeting prosumer and enterprise workstation users.

Key Features of Arc Pro B50 and B60

1. Arc Pro B50: Priced at $299, this GPU offers 16GB of VRAM and delivers up to 170 TOPS (Tera Operations Per Second) of performance. It’s a compact, 70W small form factor (SFF) GPU optimized for AI inference and virtualization, with a focus on affordability.
2. Arc Pro B60: Priced around $500, the B60 steps up with 24GB of VRAM and 197 TOPS, catering to more demanding workstation applications. It supports advanced features like hardware-accelerated ray tracing and Intel Xe Matrix Extensions (XMX) AI cores for enhanced AI compute.
3. Project Battlematrix: Intel also previewed a configurable AI development platform, codenamed Project Battlematrix, which supports up to eight Arc Pro B60 GPUs, offering a combined 192GB of VRAM. This platform, powered by Intel Xeon processors, is designed to handle AI models with up to 150 billion parameters, targeting mid-sized AI applications.

Availability

The Arc Pro B50 and B60 GPUs are currently being sampled to Intel’s partners, with availability expected in Q3 2025 (July for B50, slightly later for B60). Initially, these GPUs will be integrated into complete workstation systems, with standalone sales potentially starting in Q4 2025. Intel is collaborating with board partners like ASRock, Gunnir, Lanner, Maxsun, Onix, Senao, and Sparkle to drive adoption.

Gaudi 3 AI Accelerators: Scalable AI Solutions

Intel’s Gaudi 3 AI accelerators, first unveiled in April 2024, received significant updates at Computex 2025. Now available in PCIe and rack-scale configurations, Gaudi 3 is positioned as a scalable, open-source alternative to NVIDIA’s dominant AI accelerators, targeting enterprise and cloud AI workloads.

Key Features of Gaudi 3

1. PCIe Configuration: The PCIe add-in cards integrate seamlessly into existing x86 servers, supporting a range of AI models, from smaller Llama 3.1-8B to larger Llama 4 Scout and Maverick configurations. These cards are set to ship in the second half of 2025.
2. Rack-Scale Systems: Gaudi 3 rack-scale systems support up to 64 accelerators per rack, delivering 8.2TB of high-bandwidth memory (HBM). Featuring liquid cooling and an open, modular design aligned with Open Compute Project (OCP) standards, these systems are optimized for large-scale AI inferencing with low-latency performance.
3. Performance Claims: Intel claims Gaudi 3 offers up to four times the computational performance of its predecessor, Gaudi 2, for key large language models (LLMs), with significant improvements in memory bandwidth. Posts on X from 2024 suggest Gaudi 3 delivers 50% faster training and 40% better power efficiency compared to NVIDIA’s H100, though these claims require real-world validation.

Additional Tools

Intel also announced the public availability of its AI Assistant Builder on GitHub, an open-source framework that enables developers to create custom AI agents for Intel-based AI PCs. This tool, initially debuted at CES 2025, complements Intel’s hardware advancements by simplifying AI application development.

Competitive Landscape: Intel vs. NVIDIA and AMD

Intel’s announcements at Computex 2025 position it as a direct competitor to NVIDIA and AMD in the AI and workstation markets. Here’s how Intel stacks up:

Intel vs. NVIDIA

1. GPUs: NVIDIA’s professional GPUs, like the RTX A6000, dominate high-end workstation and AI inference tasks with robust CUDA ecosystems and higher VRAM options (up to 48GB). Intel’s Arc Pro B50 and B60, with 16GB and 24GB respectively, target a more budget-conscious segment but lack the software maturity of NVIDIA’s CUDA-X libraries, which are optimized for accelerated computing. Intel’s XMX AI cores aim to close this gap, but adoption depends on software ecosystem growth.
2. AI Accelerators: NVIDIA’s H100 and upcoming Blackwell platforms remain the gold standard for AI training and inference. Intel’s Gaudi 3, priced competitively (a Gaudi 3 kit with 8 GPUs costs ~$125,000 vs. $65,000 for Gaudi 2), emphasizes open-source flexibility and lower total cost of ownership (TCO) with liquid-cooled rack systems. However, NVIDIA’s NVLink Fusion and established market dominance pose challenges for Intel’s market penetration.
3. Market Sentiment: Posts on X highlight Intel’s aggressive pricing strategy to undercut NVIDIA, but skepticism remains about Gaudi 3’s real-world performance compared to NVIDIA’s H100 and Blackwell platforms.

Intel vs. AMD

1. GPUs: AMD’s Radeon Pro W7900 offers 48GB of VRAM, outpacing Intel’s Arc Pro B60 in memory capacity but at a higher price point. Intel’s $299 B50 and $500 B60 target cost-sensitive users, potentially appealing to small businesses and prosumers. However, AMD’s established presence in professional graphics gives it an edge in software support and driver stability.
2. AI Accelerators: AMD’s Instinct MI300 series competes with Gaudi 3 in the AI accelerator space, focusing on high-performance computing (HPC) and AI workloads. Intel’s open, modular Gaudi 3 design and PCIe compatibility provide deployment flexibility, but AMD’s broader ecosystem integration with EPYC CPUs may attract enterprises with existing AMD infrastructure.
3. Market Positioning: Intel’s focus on affordability and open standards contrasts with AMD’s emphasis on raw performance, giving Intel an edge in cost-sensitive markets but potentially lagging in high-end performance.

Pros and Cons of Intel’s Offerings

Arc Pro B-Series GPUs

Pros:

1. Affordability: At $299 (B50) and ~$500 (B60), these GPUs offer competitive pricing for workstation-grade hardware, undercutting NVIDIA and AMD equivalents.
2. High Memory Capacity: 16GB (B50) and 24GB (B60) VRAM are substantial for AI inference and professional applications like DaVinci Resolve, as demonstrated at Computex.
3. Compact Design: The B50’s 70W SFF design suits space-constrained workstations, ideal for small businesses or edge computing.
4. AI Optimization: XMX AI cores enhance AI inference performance, making these GPUs viable for mid-sized AI workloads.
5. Partner Ecosystem: Collaborations with ASRock, Gunnir, and others ensure broad system integration and availability.

Cons:

1. Software Maturity: Intel’s GPU software ecosystem lags behind NVIDIA’s CUDA and AMD’s ROCm, potentially limiting adoption for developers reliant on established frameworks.
2. Delayed Features: Features like SRIOV, VDI, and manageability software won’t be available until Q4 2025, which may deter early adopters.
3. Initial Availability: Limited to complete workstation systems until Q4 2025, restricting standalone purchases.
4. Performance Gap: While competitive, the B50 and B60 may not match the raw performance of NVIDIA’s high-end GPUs or AMD’s Radeon Pro W7900 for demanding workloads.

Gaudi 3 AI Accelerators

Pros:

1. Scalability: PCIe and rack-scale options cater to diverse use cases, from small-scale inferencing to large data center deployments.
2. Open-Source Flexibility: Alignment with OCP standards and an open, modular design reduce vendor lock-in, appealing to enterprises seeking cost-effective solutions.
3. Performance Gains: Up to 4x the performance of Gaudi 2 and competitive power efficiency make Gaudi 3 a strong contender for LLM inference.
4. Liquid Cooling: Rack-scale systems with liquid cooling enhance performance and TCO for large-scale AI deployments.
5. AI Assistant Builder: The GitHub availability of Intel’s AI Assistant Builder complements Gaudi 3, enabling developers to create custom AI agents efficiently.

Cons:

1. Market Dominance: NVIDIA’s entrenched position in AI accelerators, backed by CUDA and a mature ecosystem, poses a significant barrier to Gaudi 3 adoption.
2. Unproven Claims: While Intel claims superior performance and efficiency over NVIDIA’s H100, real-world benchmarks are needed to validate these assertions.
3. Delayed Shipping: PCIe cards won’t ship until the second half of 2025, potentially missing the current AI deployment wave.
4. Ecosystem Development: Intel’s AI software stack, while improving, is less mature than NVIDIA’s, which may limit developer adoption for complex AI workloads.

Strategic Implications and Market Outlook

Intel’s Computex 2025 announcements reflect a dual strategy: offering affordable, high-memory GPUs for prosumers and small businesses, and scalable AI accelerators for enterprises. The Arc Pro B-Series GPUs position Intel as a value-driven alternative in the workstation market, appealing to users who prioritize cost over cutting-edge performance. The Gaudi 3 accelerators, with their open-source approach and competitive pricing, aim to disrupt NVIDIA’s dominance in AI infrastructure, particularly for cost-conscious enterprises.

However, Intel faces challenges in software ecosystem development and market perception. NVIDIA’s CUDA and AMD’s ROCm have years of refinement, while Intel’s XMX and AI Assistant Builder are relatively new. The success of these products will hinge on Intel’s ability to deliver robust software support, validate performance claims through independent benchmarks, and leverage partnerships with system integrators like Dell (via Dell AI Factory) and board partners.

The broader AI market is heating up, with NVIDIA’s NVLink Fusion and AMD’s Instinct MI300 series raising the stakes. Intel’s focus on open standards and affordability could carve out a niche, particularly in edge computing and mid-sized AI deployments, but it must overcome NVIDIA’s ecosystem lock-in and AMD’s performance edge to gain significant traction.

Conclusion

Intel’s Arc Pro B50 and B60 GPUs and Gaudi 3 AI accelerators mark a significant step in its quest to compete in the AI and workstation markets. The Arc Pro GPUs offer compelling value for budget-conscious professionals, with high VRAM and AI-optimized cores, while Gaudi 3 provides scalable, open-source solutions for enterprise AI. However, software maturity, delayed feature rollouts, and fierce competition from NVIDIA and AMD present hurdles. As Intel continues to invest in its AI ecosystem and Taiwan partnerships, the success of these products will depend on delivering on performance promises and building developer trust. Computex 2025 has set the stage for an exciting AI race—stay tuned for real-world benchmarks and adoption trends in the coming months.

Sources: Computex 2025 announcements and industry reports from ServeTheHome, HPCwire, TechTimes, and posts on X.