Intel: Our upcoming AI chip will be cheaper, run cooler than Nvidia, AMD options

2026-06-01 · Source: AI - Ars Technica · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Cloud Computing & IT Infrastructure · Depth: Intermediate, quick

Summary

Intel plans to ship its new "Crescent Island" GPU, an AI chip designed for inference tasks, by the end of this year. This offering aims to challenge Nvidia and AMD by utilizing cheaper LPDDR5 memory and air-cooling technology, contrasting with rivals' expensive high-bandwidth memory (HBM) and liquid-cooling infrastructure. The chip, developed over 18 months, marks Intel's renewed push into the AI infrastructure market under CEO Lip-Bu Tan, following past struggles with AI training GPUs like "Gaudi." Intel intends to manufacture Crescent Island in its own foundries, further reducing costs compared to competitors relying on TSMC. The company is also evaluating compliance with US export controls for potential sales in China, where demand for such price points exists. Intel's shares have risen over 200 percent this year amidst this strategic shift.

Key takeaway

For AI Architects evaluating new inference hardware, Intel's upcoming "Crescent Island" GPU presents a compelling alternative to high-cost Nvidia and AMD options. Your decision-making should now factor in this air-cooled, LPDDR5-based chip, which promises significantly lower memory and cooling infrastructure expenses. This shift could enable more economical deployment of AI inference at scale, especially if your operations are sensitive to HBM and liquid-cooling costs. Investigate its performance benchmarks upon release to assess its fit for your specific inference workloads.

Key insights

Intel targets AI inference with a cost-effective, air-cooled GPU using LPDDR5 memory, aiming to undercut rivals.

Principles

• Cost-effective memory and cooling create market differentiation.
• In-house foundry manufacturing reduces chip production costs.
• Focus on specific market segments based on past experience.

In practice

• Develop AI hardware optimized for inference workloads.
• Explore LPDDR5 memory for cost-sensitive AI applications.
• Consider air-cooling solutions to reduce infrastructure complexity.

Topics

• Intel Crescent Island
• AI Inference
• LPDDR5 Memory
• Air Cooling
• Foundry Manufacturing
• Data Center GPUs
• US Export Controls

Best for: CTO, VP of Engineering/Data, MLOps Engineer, AI Hardware Engineer, AI Architect, Director of AI/ML

Related on AIssential

• Nvidia Plans New Chip to Speed AI Processing, Shake Up Computing Market — Nvidia is launching a new inference-optimized AI processor to accelerate model responses and reshape the AI hardware market.
• AMD's tiny AI PC points to a more local future for model inference — Local AI inference on compact hardware reduces cloud dependency, improves privacy, and lowers latency.
• The RTX 5080 Is 3x Faster Than NVIDIA’s $3,999 “AI Supercomputer” — And It Doesn’t Matter — Memory capacity, not raw inference speed, defines the DGX Spark's primary value proposition.
• What is LPU? Language Processing Units | The Future of AI Inference — LPUs offer deterministic, low-latency inference for LLMs by optimizing for sequential processing and on-chip memory.
• India’s $11B Semiconductor Bet, AI Power Shift: Meta Chips & The AI Cost Crisis — The global AI landscape is marked by a race for custom silicon, rising inference costs, and national strategic investments.
• AI inference costs dropped up to 10x on Nvidia's Blackwell — but hardware is only half the equation — AI inference cost reductions up to 10x are achieved by combining Blackwell hardware, optimized software, and open-source models.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI - Ars Technica.