Foundations of Artificial Intelligence: Big Data, Processors and Large Language Models

This chapter reviews the interdependencies among AI-specialized microprocessors, big dataecosystems, and large-scale foundation models, including large language models (LLMs) andlarge multimodal models (LMMs). The primary objective is to analyze how the convergence ofdata availability, algorithmic advancements, and hardware innovation has reshaped theartificial intelligence (AI) landscape and accelerated the development of generative AIsystems. Drawing on industry reports, technical documentation, and recent literaturepublished between 2024 and 2025, this study examines three critical dimensions: (i) theevolution of processing architectures from general-purpose CPUs to specialized acceleratorssuch as GPUs, TPUs, NPUs, and emerging paradigms (i.e., neuromorphic and photoniccomputing), (ii) the role of large-scale and multimodal big data in enabling foundation modeltraining, and (iii) the growing energy consumption and sustainability challenges associatedwith large AI systems. In addition, national and global semiconductor strategies, includinginvestments in exascale supercomputing and sovereign AI infrastructures, are analyzed tounderstand geopolitical and technological dynamics. The findings highlight three key insights.

First, the co-evolution of big data, processors, and model architectures forms the foundationaldriver of modern AI capabilities. Second, the rapid scaling of LLMs and LMMs has significantlyincreased computational and energy demands, making hardware efficiency and infrastructuredesign critical bottlenecks. Third, global competition in semiconductor technologies and AIinfrastructure is intensifying, with nations prioritizing technological sovereignty and secureaccess to advanced computing resources. In conclusion, the future of AI depends on a balancedintegration of scalable data ecosystems, energy-efficient hardware, and advanced modelarchitectures. Sustainable and secure AI development will require coordinated efforts inhardware–software co-design, algorithmic optimization, renewable energy integration, andpolicy-driven governance to ensure responsible and inclusive technological progress. ModernAI relies on three key interconnected pillars: LLMs for advanced language processing, LMMsfor handling diverse data types (i.e., text, images, audio, video) beyond language alone, and AI-specialized microprocessors (i.e., GPUs, TPUs, custom accelerators) that supply the massivecomputational power needed for training and running these models efficiently. Together,these enable foundation models capable of broad, generalized learning and real-worldapplications while posing significant challenges in energy use, scalability, and deployment asobserved in research and industry trends from 2024–2025.