Intel and Google Expand AI Alliance with Focus on CPUs and Custom Chips

Intel and Google have expanded their partnership to supply chips for Google’s AI data centers, with a focus on Intel’s Xeon CPUs and jointly developed custom processors.

The multiyear agreement deepens a relationship that stretches back decades, with Google continuing to rely on Intel processors as a foundation of its global infrastructure. Under the new terms, Google will deploy multiple generations of Xeon chips, including its latest Xeon 6 processors, across cloud systems that support AI and cloud workloads.

The deal reflects a trend in how AI infrastructure is being built: while pricey GPUs have dominated the development of AI models, running those models at scale places greater demands on other parts of the system, particularly CPUs. In Google’s data centers, Xeon processors coordinate workloads and support AI inference, elevating the role of CPUs in overall system efficiency.

The companies are also expanding their joint work on Infrastructure Processing Units, or IPUs. These custom chips are designed to handle background tasks such as networking, storage management, and security, functions that would otherwise consume CPU resources.

“While news of the latest GPUs gets more than its share of headlines, Intel’s expanded partnership with Google emphasizes that general-purpose CPUs remain foundational in AI-era infrastructure,” said Mitch Ashley, VP Practice Lead at Futurum Group. “By committing to multiple future generations of Xeon and co-developing infrastructure silicon, Google is effectively betting that heterogeneous architectures—CPUs, accelerators, and IPUs working as a system—are the only sustainable way to scale AI workloads while keeping flexibility and control over its platform roadmap.” 

Distributed Architecture

The expanded agreement also signals a long-term commitment. By aligning with Intel’s product roadmap across multiple chip generations, Google is making forward-looking infrastructure decisions that extend beyond near-term deployment cycles.

For Intel, the deal reinforces its position in a market where it has been laboring to gain greater traction. While companies like NVIDIA continue to dominate AI training hardware, Intel is focusing on the broader system architecture, where CPUs and supporting chips remain essential.

For Google, the deepened partnership is part of a larger effort to diversify its hardware strategy. The company has developed its own AI accelerators and explored alternative processor designs, reflecting an industry-wide effort to reduce dependence on any single vendor or architecture.

“Google gets better CPUs partnering with Intel, as well as customized infrastructure components that can make them more competitive and efficiently operate the datacenters,” Jack Gold, President of J. Gold Associates, told Techstrong. “Intel gets to understand the optimum components to put into their high-end CPU chips for the needs of cloud hyperscalers through Google feedback, and it gets to increase revenues through sale of custom chips (IPU) while also filling its fabs.”

Across the sector, companies are reevaluating how AI systems are structured. As workloads scale, bottlenecks are emerging not only in compute power but in how systems move and manage data. That shift is driving renewed investment in CPUs and infrastructure-focused silicon.

The Intel-Google partnership is a clear example of this transition. Rather than competing solely on raw processing power, the companies are focusing on how different components work together to deliver performance and scalability. As AI develops, that system-level approach is likely to shape the next phase of infrastructure development.