Intel and Foxconn partner on rack-scale AI infrastructure to compete with NVIDIA
Intel and Foxconn announced a strategic partnership at Computex 2026 to jointly develop next-generation AI infrastructure spanning silicon, racks, systems, and applications. The collaboration pairs Intel's Xeon processor architecture and silicon technology with Foxconn's manufacturing scale and system integration expertise. The companies plan to build rack-scale AI infrastructure powered by Intel Xeon 6+ processors paired with SambaNova SN-50 Reconfigurable Dataflow Units, targeting inference and agentic AI workloads across data centers and edge environments.
The reference configuration delivered by the partnership is a single liquid-cooled rack providing 36,864 cores in 32U at approximately 100 kW — positioning it as Intel's competitive response to NVIDIA's GB200 NVL racks and AMD's MI-series platforms. Foxconn will serve as system integrator and manufacturer, with plans to produce CPU-dense variants for workloads that don't require dedicated accelerators. The companies will also explore custom chip design and specialized system integration for edge AI, robotics, smart manufacturing, and automotive applications.
For Intel, the deal addresses a persistent challenge: translating its data center silicon roadmap into deployed, integrated systems and competing beyond commodity CPU sales. For Foxconn, the partnership diversifies its manufacturing portfolio beyond NVIDIA GPUs and deepens its positioning as a full-stack AI infrastructure provider. No financial terms were disclosed, but the collaboration signals that system-level integration and end-to-end platforms are becoming critical competitive vectors against NVIDIA's vertically integrated approach.