ARTIFICIAL INTELLIGENCE

“A Graphics Processing Unit (GPU) is a specialised processor designed for parallel computing tasks, excelling at handling thousands of threads simultaneously, unlike CPUs which prioritise sequential processing. It is widely used for AI.” – GPU

A Graphics Processing Unit (GPU) is a specialised electronic circuit designed to accelerate graphics rendering, image processing, and parallel mathematical computations by executing thousands of simpler operations simultaneously across numerous cores.^1,2,4,6

Core Characteristics and Architecture

GPUs excel at parallel processing, dividing tasks into subsets handled concurrently by hundreds or thousands of smaller, specialised cores, in contrast to CPUs which prioritise sequential execution with fewer, more versatile cores.^1,3,5,7 This architecture includes dedicated high-bandwidth memory (e.g., GDDR6) for rapid data access, enabling efficient handling of compute-intensive workloads like matrix multiplications essential for 3D graphics, video editing, and scientific simulations.^2,5 Originally developed for rendering realistic 3D scenes in games and films, GPUs have evolved into programmable devices supporting general-purpose computing (GPGPU), where they process vector operations far faster than CPUs for suitable applications.^1,6

Historical Evolution and Key Applications

The modern GPU emerged in the 1990s, with Nvidia’s GeForce 256 in 1999 marking the first chip branded as a GPU, transforming fixed-function graphics hardware into flexible processors capable of shaders and custom computations.^1,6 Today, GPUs power:

• Gaming and media: High-resolution rendering and video processing.^4,7
• AI and machine learning: Accelerating neural networks via parallel floating-point operations, outperforming CPUs by orders of magnitude.^1,3,5
• High-performance computing (HPC): Data centres, blockchain, and simulations.^1,2

Unlike neural processing units (NPUs), which optimise for low-latency AI with brain-like efficiency, GPUs prioritise raw parallel throughput for graphics and broad compute tasks.¹

Best Related Strategy Theorist: Jensen Huang

Jensen Huang, co-founder, president, and CEO of Nvidia Corporation, is the preeminent figure linking GPUs to strategic technological dominance, having pioneered their shift from graphics to AI infrastructure.¹

Biography: Born in 1963 in Taiwan, Huang immigrated to the US as a child, earning a BS in electrical engineering from Oregon State University (1984) and an MS from Stanford (1992). In 1993, at age 30, he co-founded Nvidia with Chris Malachowsky and Curtis Priem using $40,000, initially targeting 3D graphics acceleration amid the PC gaming boom. Under his leadership, Nvidia released the GeForce 256 in 1999—the first GPU—revolutionising real-time rendering and establishing market leadership.^1,6 Huang’s strategic foresight extended GPUs beyond gaming via CUDA (2006), a platform enabling GPGPU for general computing, unlocking AI applications like deep learning.^2,6 By 2026, Nvidia’s GPUs dominate AI training (e.g., via H100/H200 chips), propelling its market cap beyond $3 trillion and Huang’s net worth over $100 billion, making him the world’s richest person at times. His “all-in” bets—pivoting to AI during crypto winters and data centre shifts—exemplify visionary strategy, blending hardware innovation with ecosystem control (e.g., cuDNN libraries).^1,5 Huang’s relationship to GPUs is foundational: as Nvidia’s architect, he defined their parallel architecture, foreseeing AI utility decades ahead, positioning GPUs as the “new CPU” for the AI era.³