NVIDIA vs Nykaa

NVIDIA Market Stance

NVIDIA Corporation occupies a position in the technology industry that has no precise historical parallel. In the span of roughly three years — from 2021 to 2024 — the company transformed from a respected but conventionally sized semiconductor business with approximately $16 billion in annual revenue into one of the largest companies in the world by market capitalization, briefly surpassing $3 trillion in mid-2024 and trading at revenue multiples that reflected investor conviction that NVIDIA had become the essential infrastructure provider for the most consequential technological transition in a generation. The company was founded in 1993 by Jensen Huang, Chris Malachowsky, and Curtis Priem in Sunnyvale, California. Huang, a Taiwanese-American engineer who had previously worked at AMD and LSI Logic, brought a distinctive vision: that visual computing — the specialized processing of graphics — was a fundamentally different computational problem from general-purpose CPU processing, and that dedicated hardware architectures could solve it orders of magnitude more efficiently. The early NVIDIA products were graphics accelerators for the PC gaming market, competing against companies like 3dfx and ATI in a market that was growing rapidly as PC games became more visually sophisticated. The pivotal architectural decision came in 1999 with the GeForce 256, which NVIDIA marketed as the world's first Graphics Processing Unit — a term the company coined to describe a chip that could handle the full geometry and rendering pipeline for 3D graphics without CPU involvement. The GPU concept was not merely a marketing formulation; it described a genuinely different computational architecture. Where CPUs are optimized for sequential task execution — doing one complex thing very fast — GPUs are optimized for parallel task execution — doing thousands of simple things simultaneously. This architectural difference, originally designed to render thousands of independent pixels in parallel, would prove to have implications far beyond graphics that NVIDIA itself did not fully anticipate for more than a decade. The introduction of CUDA (Compute Unified Device Architecture) in 2006 was the strategic inflection point that separated NVIDIA's trajectory from every other GPU company. CUDA was a parallel computing platform and programming model that allowed developers to use NVIDIA GPUs for general-purpose computation — not just graphics — by writing code in a modified version of the C programming language. Before CUDA, using a GPU for non-graphics computation required the developer to frame their problem as a graphics rendering task, a contortion that limited adoption to specialists. CUDA eliminated this barrier, opening NVIDIA's GPU architecture to the entire scientific computing and research community. The consequences of CUDA took years to compound but eventually proved epochal. Researchers in machine learning — a field that had been computationally constrained since its theoretical foundations were established decades earlier — discovered that training neural networks on NVIDIA GPUs with CUDA was orders of magnitude faster than training on CPUs. The landmark 2012 AlexNet paper, which demonstrated that a deep convolutional neural network trained on NVIDIA GPUs could dramatically outperform existing computer vision systems on the ImageNet benchmark, effectively launched the modern deep learning era and cemented NVIDIA's role as the hardware platform of choice for AI research. From 2012 through 2022, NVIDIA's GPU computing platform grew steadily in the data center as machine learning adoption expanded from academic research into production applications at technology companies. Revenue grew from approximately $4 billion in 2013 to $16.7 billion in fiscal year 2022. Then the generative AI wave — catalyzed by the release of ChatGPT in November 2022 and the subsequent explosion of large language model development — triggered demand for NVIDIA's H100 GPU that exceeded the company's manufacturing capacity for multiple consecutive quarters. The H100, manufactured on TSMC's 4nm process and containing 80 billion transistors, is the primary computational tool for training and deploying large language models. Training a frontier AI model like GPT-4 or Gemini requires thousands of H100 GPUs running continuously for weeks. Every major technology company — Microsoft, Google, Amazon, Meta, and Oracle — along with dozens of AI startups and sovereign nations building national AI infrastructure, placed H100 orders that created a backlog measured in billions of dollars. NVIDIA's data center revenue grew from $3.8 billion in fiscal year 2022 to over $47 billion in fiscal year 2024 — a more than tenfold increase in two years. Jensen Huang's leadership through this period has been widely recognized as one of the most successful instances of long-term strategic positioning in technology business history. Huang, who has led NVIDIA continuously since its founding — an extraordinary tenure by Silicon Valley standards — made the foundational investment in CUDA in 2006 when GPU computing for AI was not a visible commercial opportunity. He sustained that investment through a decade of gradual adoption, built the software ecosystem that made NVIDIA GPUs not just the best AI hardware but the only hardware that most AI researchers knew how to use, and positioned the company to capture the demand surge when it arrived with manufacturing relationships, product roadmaps, and software tools already in place. The scale of NVIDIA's current market position is difficult to overstate. The company is estimated to supply approximately 70-80% of the AI training chips used by the global technology industry. Its H100 and the subsequent H200 and Blackwell architecture GPUs are the primary hardware substrate on which the AI models that are reshaping every industry — from healthcare diagnostics to legal research, from software development to drug discovery — are being trained and deployed. In this sense, NVIDIA has become something analogous to what Intel was to the PC era or what TSMC is to semiconductor fabrication: the essential, largely irreplaceable infrastructure provider for a foundational technology platform.