Intel will release a new data-center chip this year to challenge Nvidia and AMD in the AI hardware market. The chip focuses specifically on AI inference workloads with cheaper memory and lower power consumption to attract cost-conscious data center operators.
Inference Workload Design
This chip targets the AI inference phase, where trained models generate outputs like chat responses or image analysis. Inference requires less raw computing power than training new models, letting Intel prioritize efficiency over peak performance. The design skips expensive components unnecessary for inference tasks, avoiding over-engineering for workloads the chip won't handle. Data centers run inference constantly as applications scale, making small per-unit savings critical at scale.
Memory and Power Focus
Memory costs and electricity consumption are major pain points for data centers running AI services. Intel's chip uses lower-cost memory configurations while maintaining acceptable inference speeds. Power draw is reduced through architectural tweaks that minimize energy waste during routine operations. Operators managing thousands of inference requests daily could see meaningful reductions in their operating expenses. The company's bet is that these practical savings will sway customers tired of high hardware costs.
Competitive Positioning
Nvidia dominates the AI chip market with its versatile GPUs, while AMD has made incremental gains. Intel enters with a narrow play targeting only inference workloads rather than competing broadly. The company hopes its cost and efficiency edge will resonate where Nvidia's premium pricing creates friction. This move addresses Intel's previous missteps by focusing on a specific need rather than trying to match rivals' full product stacks. Data center operators now have another option for inference-heavy deployments.
Shipping Timeline
Intel plans to start customer shipments by year-end 2024. The company is finalizing production schedules with manufacturers this summer. Early adopters will evaluate the chip against current hardware before placing major orders. No specific quarterly milestones were provided beyond the 2024 launch window.
