Intel, Google deepen AI infra ties

Intel and Google have announced a multiyear collaboration to co-develop and refine the building blocks that power cloud computing and AI systems.

The deal means that Google Cloud will continue to use Intel’s Xeon processors across a wide range of workloads, including AI, inference, and general computing.

Alongside this, the two companies will deepen work on custom IPUs, or infrastructure processing units. These are specialised chips designed to handle tasks such as networking, storage, and security, which would otherwise burden the central processor.

Intel said the aim is to build “more efficient, balanced AI systems”. This aligns with Google Cloud’s existing setup. Its Compute Engine already runs on multiple generations of Intel Xeon chips, suggesting the partnership formalises and extends a relationship that is already in place.

AI systems rely heavily on powerful accelerators such as GPUs, but they also depend on a broader foundation of computing infrastructure. Intel’s processors and IPUs help manage data movement, system coordination, and other background tasks that keep AI workloads running smoothly. By co-developing these components, Google can improve efficiency and control costs while scaling its cloud services.

Google Cloud does not rely on a single supplier. It already uses chips from AMD alongside Intel, and has developed its own custom silicon. It also works closely with NVIDIA on AI infrastructure. Intel’s role is to strengthen the non-GPU layer of AI infrastructure.

Intel has faced intense competition in recent years, particularly in AI where NVIDIA dominates high-end computing. Securing a continued and expanded role within Google Cloud helps ensure that Intel’s Xeon processors remain relevant in one of the most important markets for computing. It also supports Intel’s argument that CPUs still play a central role in AI systems, handling orchestration, data flow, and system-level operations.

Intel has been working with companies such as NVIDIA on joint infrastructure development, with Cisco on edge computing systems, and with Amazon Web Services on cloud instances powered by its latest chips. It has also collaborated with AI firm SambaNova Systems and cybersecurity group CrowdStrike.

Earlier this week, Intel said it would join Elon Musk’s chipmaking plan with SpaceX, Tesla and xAI, and that its role would help speed up a project aimed at producing vast volumes of advanced compute for AI and robotics. Last month, Musk said Tesla, working with SpaceX, would build two advanced chip factories in Austin, one for cars and humanoid robots and the other for AI data centres in space.

Google, meanwhile, continues to deepen ties with NVIDIA for advanced AI workloads and with AMD for secure computing. Its approach reflects a broader industry shift.

That shift is towards what engineers call a heterogeneous system. This means using different types of chips for different tasks. CPUs handle general operations, IPUs manage infrastructure functions, and GPUs or other accelerators perform the heavy AI calculations. By combining these, companies aim to balance performance, cost, and energy use.