More than LLMS

: the future of GPU-based parallel computing and Nvidia’s role

Abstract

The role of parallel GPU computing has transformed fields dependent on heavy computational workloads. Its strength lies in decomposing tasks into thousands of smaller operations that can be executed simultaneously, achieving major efficiency gains compared to serial processing. Nvidia established itself as the pivotal actor by recognizing early the potential of GPUs beyond graphics. By combining hardware performance, developer lock-in, and ecosystem effects, the company turned itself into an infrastructural backbone of today’s AI and high-performance computing economy. With Moore’s Law slowing down, physical and economic limits to transistor scaling raise the question whether Nvidia can sustain this dominance or whether alternative paradigms might eventually challenge it. To address this, the study applies a mixed-method design, quantitative modeling of transistor trajectories (exponential vs. logistic fits) is combined with twelve expert interviews covering technological, strategic, and geopolitical aspects. This triangulation provides a broad perspective on both opportunities and vulnerabilities. Findings indicate that the slowdown of Moore’s Law does not automatically weaken Nvidia’s position. Architectural innovations, software ecosystems, and supply chain alliances compensate for scaling limits and reinforce the company’s strength. At the same time, risks remain in rising costs, fragile geopolitical dependencies, and the disruptive potential of alternative paradigms.

Date of Award	21 Oct 2025
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Peter V. Rajsingh (Supervisor)