Google Cloud and Intel have strengthened their multi-year AI infrastructure partnership, focusing on CPU deployment and custom chip co-development. Google Cloud will implement Intel’s Xeon 6 processors globally for C4 and N4 instances, while both companies enhance their development of custom Infrastructure Processing Units to transfer networking, storage, and security responsibilities away from host CPUs in hyperscale AI settings. This comes as Intel’s stock rose 33% within a week and shortly after being chosen as Tesla’s foundry partner for the Terafab megaproject.
“Balanced systems”: Intel and Google argue that modern AI infrastructure requires more than just GPU accelerators. Intel’s CEO, Lip-Bu Tan, emphasized the need for balanced systems including CPUs and IPUs for AI workload efficiency and flexibility. Intel has shifted focus toward the role of CPUs and custom silicon in AI, challenging GPU-centric views. Google’s Amin Vahdat highlighted the importance of CPUs in AI systems and expressed trust in Intel’s roadmap for continued performance and efficiency, positioning the partnership as a long-term strategy rather than a single procurement.
Xeon 6 in Google Cloud: Intel’s Xeon 6 processor family is central to this partnership, deployed in Google Cloud’s optimized C4 and N4 instances. These instances reportedly deliver over twice the total cost of ownership benefit compared to previous versions, thanks to enhanced performance and power efficiency. The partnership commits Google to Intel’s future CPU releases, integrating Intel’s trajectory into its infrastructure plans. Concurrently, Google strengthens its custom silicon ties, providing Anthropic with TPU capacity through Broadcom.
The CPU context reveals a shift in AI workloads from training to inference, which places more demand on CPU resources. Google’s investment in Xeon 6 reflects a belief in the growing importance of CPU efficiency for inference economics.
Custom IPU Program: The partnership emphasizes the extended co-development of Infrastructure Processing Units, designed to manage networking, storage, and security tasks typically handled by host CPUs. Offloading these tasks enhances utilization rates, energy efficiency, and workload performance consistency in large-scale environments. The partnership’s expansion signifies a broader research scope, though specific technical details remain undisclosed.
Intel and Google aim to challenge Nvidia’s dominance by promoting efficiency gains from systems surrounding GPU accelerators. With Google’s infrastructure scale, the partnership has the capacity to empirically support these claims and move away from a single-vendor dependency.
Intel’s Strategic Moment: The partnership with Google aligns with Intel’s evolving industrial role. Prior to Google’s announcement, Intel became the main foundry partner for Terafab, a significant venture by Tesla, SpaceX, and xAI, using its 18A technology. This dual-strategy approach seeks to enhance hyperscale cloud partnerships and develop a foundry business for custom AI silicon. The week’s news led to a 33% increase in Intel’s stock, a notable spike in recent years.
Intel’s future depends on its ability to execute these plans, hinged on its 18A technology. The AI infrastructure market has seen major investments, highlighting Intel’s competitive potential and its re-emergence in the AI infrastructural segment.