Intel and Google are doubling down on AI CPUs with an expanded partnership

Intel Corporation has unveiled plans to enhance its longstanding collaboration with Google Cloud, honing in on AI infrastructure developments. This strategic partnership aims to create custom Infrastructure Processing Units (IPUs) and integrate Intel’s Xeon 6 processors into Google’s data centers, marking a significant step in reversing Intel's recent struggles in a market increasingly dominated by NVIDIA. In a move towards a more comprehensive approach to AI, Intel is prioritizing 'heterogeneous' systems that utilize both specialized accelerators and CPUs, positioning these components as essential 'brains' capable of managing extensive AI workloads efficiently. According to a press release from Intel, this partnership will broaden the application of Xeon 6 processors within Google Cloud’s infrastructure. These processors will enhance various cloud instances, specifically designed to tackle demanding AI tasks, including rapid AI predictions and the training of sophisticated AI models. Moreover, the collaboration will see both companies develop specialized custom chips known as ASIC-based IPUs, which will facilitate essential functions such as networking, storage management, and security tasks. These IPUs will take over responsibilities traditionally handled by CPUs, streamlining infrastructure management. "CPUs and infrastructure acceleration are fundamental to AI systems, overseeing everything from training orchestration to inference and deployment," stated Amin Vahdat, SVP & Chief Technologist for AI Infrastructure at Google. "AI is fundamentally transforming how infrastructure is constructed and scaled." Intel CEO Lip-Bu Tan echoed these sentiments, emphasizing that expanding AI capabilities necessitates balanced systems. "It’s not just about accelerators; achieving effective AI performance and flexibility hinges on the integration of CPUs and IPUs to meet the demands of modern workloads." In tandem with strengthening its partnership with Google, Intel has also inked a monumental $25 billion agreement with Elon Musk to establish the 'Terafab'—a state-of-the-art chip manufacturing plant in Texas. This facility is set to produce cutting-edge 2-nanometer chips at unprecedented rates, aiming for an output of 100,000 wafers monthly. Such advancements are expected to bolster computing capabilities for xAI, Tesla’s self-driving technology, and SpaceX operations.