NVIDIA is a US technology company founded in 1993 by Jensen Huang, Chris Malachowsky and Curtis Priem, headquartered in Santa Clara, California, and widely regarded as the defining company of the current AI wave. Originally a maker of graphics processing units (GPUs) for computer gaming, it now develops graphics processors, systems on chips and application programming interfaces for data science, high-performance computing, artificial intelligence and automotive applications.
Business model
The majority of revenue comes from the data-center business, reported as the "Compute & Networking" segment, which delivered about 89 percent of group revenue in fiscal 2026; the "Graphics" segment contributed only about 11 percent. NVIDIA's Blackwell-generation accelerators — including the rack-scale GB200 NVL72 and GB300 NVL72 systems — together with the in-house Grace CPU and the NVLink interconnect form the compute infrastructure behind much of the world's AI training and inference workloads. This is complemented by the CUDA software platform (released 2007, extended through CUDA-X), a de-facto standard for GPU programming that drives strong customer lock-in, because CUDA software, per Wikipedia, still runs almost exclusively on NVIDIA GPUs.
AI relevance
NVIDIA is both an AI stock and an AI-chip stock: its accelerators are the central compute tool for providers of generative AI, cloud hyperscalers and research labs. It also supplies pre-configured systems such as DGX SuperPOD/BasePOD and DGX Cloud, networking technology such as Spectrum-X and BlueField DPUs (after the Mellanox acquisition), and software stacks for robotics (Isaac) and autonomous vehicles (Drive).
Key figures
For fiscal 2026 (ended January 25, 2026), NVIDIA reported record revenue of roughly $215.9 billion (up 65 percent), of which $193.7 billion came from the data-center business (up 68 percent), with about 42,000 employees. In Q1 fiscal 2027 (ended April 26, 2026), revenue reached a further record of $81.6 billion (up 85 percent), of which $75.2 billion came from data centers (up 92 percent); NVIDIA also announced the Vera Rubin platform with the Vera CPU for agentic AI. The shares trade on Nasdaq under the ticker NVDA; in July 2026 market capitalization briefly touched roughly $5.1 trillion — the world's most valuable listed company.
From gaming to AI infrastructure
NVIDIA's path from graphics cards for gaming to AI data-center supplier is tied to using highly parallel graphics processors for general-purpose computing. Its origins trace to research by Ian Buck at Stanford, which NVIDIA picked up from 2004 and released as CUDA in 2007; from around 2015, CUDA development shifted toward machine learning. CUDA remains an important competitive advantage, since much of the world's AI software is built around this proprietary environment; open alternatives such as AMD's ROCm or Intel's oneAPI have so far captured only a niche.
Ecosystem and competition
The business is closely linked to cloud providers' investment cycles; Blackwell instances are available, per the company, on AWS, Google Cloud, Azure and Oracle Cloud Infrastructure. Some of these customers are also developing their own accelerators, and AMD and Intel compete with NVIDIA in graphics and AI chips. As a fabless company, NVIDIA has its chips made externally, mainly by TSMC. Export restrictions have a direct impact: in Q1 fiscal 2026, a new US licensing requirement for the China-specific H20 chip cost $4.5 billion.
This profile is a neutral description of the business model and is expressly not investment advice or a price-target forecast.