Intel will sell a cheap GPU with 32GB VRAM next week
Intel launches a budget-friendly 32GB VRAM GPU next week, priced at $949 for AI workloads.
Intel is making a direct play for the AI workstation market with the launch of its Arc Pro B70 GPU on March 31. Priced at $949, the card's headline feature is its substantial 32GB of VRAM, coupled with a memory bandwidth of 608 GB/s and a thermal design power (TDP) of 290W. This specification sheet is designed to compete with offerings like NVIDIA's RTX 5070, providing a more affordable entry point for developers and researchers who need high memory capacity for large language models and other AI inference tasks.
The primary use case Intel is targeting is local AI model deployment. The 32GB VRAM buffer is sufficient to run sizable quantized models, such as a 4-bit quantized version of Qwen 3.5 27B, entirely on a local machine. This enables faster iteration, enhanced data privacy, and eliminates cloud inference costs for professionals. By selling the GPU directly to consumers, Intel is aiming to disrupt the current market dynamics and provide a viable third option beyond NVIDIA and AMD for AI-focused hardware.
- Launches March 31 with a direct-to-consumer price of $949.
- Features 32GB of VRAM with 608 GB/s bandwidth and a 290W TDP.
- Specifically positioned to run quantized local AI models like Qwen 3.5 27B.
Why It Matters
It provides a more affordable, high-VRAM option for developers to run large AI models locally, reducing cloud dependency and cost.