ASUS debuts high-priced Intel Arc Pro B70 GPU for AI developers

A premium entry into AI hardware

ASUS has officially launched the UGen Intel Arc Pro B70 graphics card in the Australian market. This specific model is aimed at professional workstation users and developers who require specialized hardware for artificial intelligence tasks. The card comes with a significant price tag of $3,299, placing it in a high-end bracket for hardware enthusiasts.

The pricing of this card is notably higher than the original Intel Arc Pro B70, which launched in March 2026 for approximately A$1,350. By positioning the product at this premium level, ASUS is directly competing with the price point of the NVIDIA GeForce RTX 5090. However, the two cards serve different primary functions despite their similar costs.

While the NVIDIA flagship is built for high-end gaming and creative work, the ASUS version is marketed as a workstation tool. It is designed specifically for local AI inference and professional applications. This distinction is important for buyers who need to choose between a versatile gaming machine and a dedicated AI workstation.

Technical specs for AI workloads

The UGen Arc Pro B70 is built on Intel’s Xe2 architecture. It includes 32 Xe2 cores and 256 XMX engines, which are components specifically designed to speed up AI processing. These features allow the card to handle complex mathematical operations required by modern machine learning models.

One of the most important specifications for this card is its 32GB of ECC GDDR6 memory. ECC, or Error Correction Code, is a type of memory that can detect and fix internal data corruption. This high memory capacity is a primary selling point because it allows developers to run large AI models directly on their own computers.

By running models locally, users can avoid the limitations of smaller consumer graphics cards. This is particularly useful for businesses that handle sensitive information and cannot risk sending data to external servers. It also provides a way for developers to reduce the ongoing costs associated with cloud-based processing services.

Power and cooling for workstations

ASUS has designed the card with a dual-slot blower-style cooling system. This type of cooling is often preferred in workstation environments because it can move air efficiently through a chassis. It ensures that the components stay cool even during the sustained heavy loads typical of AI training or inference.

The card features a 310W power rating, which is a substantial increase over the 230W rating found on Intel's reference Arc Pro B70. This extra power allowance provides the hardware with more performance headroom for demanding tasks. However, it also means the card requires a more robust power delivery system and better cooling in the host computer.

The increased power and cooling requirements are standard for high-performance workstation hardware. Because the card is intended for professional use, these specs ensure it can maintain high performance over long periods. It is built to handle the intensity of professional workloads rather than the bursty nature of standard consumer applications.

Scaling with multi-GPU setups

For users with even more demanding needs, ASUS supports multi-GPU configurations with this card. The company is promoting systems that can house up to four Arc Pro B70 cards in a single workstation. This setup would provide a combined 128GB of graphics memory across the four individual units.

A four-card configuration would cost roughly $13,196, not including the cost of the rest of the workstation hardware. This makes it a very expensive solution for heavy-duty AI development. It is intended for organizations that need to process massive datasets or run extremely large models that exceed the limits of a single card.

It is important to note that the combined 128GB of memory does not function as one single shared pool. The usable capacity will depend on the specific software and the type of workload being performed. This limitation means developers must still optimize how they distribute tasks across the multiple GPUs in the system.

Competition and market position

ASUS faces a significant challenge in the AI market because NVIDIA has a very established position. NVIDIA's dominance is largely due to its CUDA software ecosystem, which many developers already use. To succeed, Intel and ASUS must prove that their hardware and software support can be just as reliable for professional users.

The success of the Arc Pro B70 will likely depend on how well it performs in real-world AI applications. While the memory capacity is impressive, the software compatibility and performance efficiency are key factors. Buyers will need to see if the card can handle their specific workflows as effectively as the current industry standards.

For now, the UGen Arc Pro B70 represents a high-end option for those seeking substantial memory for local AI processing. It offers a specific path for developers who need to move away from the cloud. Whether the premium price is justified will depend on the actual results delivered by the Xe2 architecture in professional settings.