Google and Intel deepen their cooperation in the artificial intelligence race

Google and Intel announced that they have expanded their collaboration on artificial intelligence and cloud infrastructure. The two companies’ multi-year agreement includes Google Cloud’s continued use of Intel-based infrastructure while also continuing co-processor development work. This development comes at a time when demand for hardware components needed to scale AI systems is rapidly increasing.

On the Google Cloud side, Intel’s Xeon processor family has been used for a long time. Under the new agreement, the company will continue to use Intel’s server-class processors, including the latest Xeon 6 series, in artificial intelligence, cloud computing and inference processes. This choice responds to the need to run different workloads in a balanced manner, especially in data centers. In addition, the long history of use of Xeon processors allows Google to move forward while preserving its existing infrastructure investments.

The collaboration of the two companies does not only include the use of existing hardware. In addition, the development process for infrastructure processing units (IPUs) in data centers, which improves performance by offloading certain tasks to the CPU, is also being expanded. This joint development program, which starts in 2021, will focus specifically on custom ASIC-based IPU solutions. Such units improve overall system efficiency by optimizing tasks such as data flow management and network operations.

CPU and IPU balance stands out in artificial intelligence infrastructure

The proliferation of artificial intelligence applications is not limited to just increasing the need for GPUs on the hardware side. While GPUs stand out in the training processes, CPUs play a critical role in the running and scaling of the models. Intel CEO Lip-Bu Tan also emphasized in his statement that artificial intelligence systems require a balanced infrastructure and at this point, the combination of CPU and IPU is decisive in terms of performance, efficiency and flexibility.

On the other hand, it is noteworthy that the demand for CPUs in the semiconductor industry has increased recently. Similar to the intense demand on the GPU side, there has also begun to be a shortage in CPU supply for data centers and artificial intelligence applications. This situation leads companies to develop alternative solutions. As a matter of fact, Arm Holdings, a subsidiary of SoftBank, recently entered this field directly by announcing the Arm AGI CPU, the first processor it developed itself.

The expanded collaboration between Google and Intel stands out as one of the concrete examples of this transformation in the industry. While the continuity of the existing infrastructure is ensured, special hardware development processes are accelerating. All this shows that more customized and balanced systems will be decisive in the future of artificial intelligence-focused data centers.