NVIDIA Corp.

NVDA ·Technology, Semiconductors, United States

NVIDIA Corporation (NVDA)

Overview

NVIDIA Corporation is a Santa Clara, California-based semiconductor and computing platform company that designs graphics processing units (GPUs) and the surrounding hardware, software, and networking stack used to run artificial intelligence, high-performance computing, and visual computing workloads. It is a member of the S&P 500 and, as of early September 2026, is the world's most valuable public company, with a market capitalization around $5.25 trillion. The company is now overwhelmingly an AI infrastructure supplier: in fiscal year 2026 (ended roughly January 2026) it generated $215.9 billion in total revenue, up 65% year-over-year, with net income of $120.1 billion and GAAP gross margins above 71%. NVIDIA employs roughly 36,000–40,000 people, a headcount that has grown rapidly alongside its data-center business, though it remains small relative to its revenue and market value because it is a "fabless" designer that outsources chip manufacturing.

What They Do & How They Make Money

NVIDIA's business model centers on designing high-performance chips — GPUs, and increasingly full "AI factory" systems combining GPUs, CPUs, networking, and software — and selling or licensing them to a wide range of customers, from cloud giants to individual gamers. The company does not own semiconductor fabrication plants; it designs chips and contracts manufacturing out to foundries such as TSMC, then captures value through architecture, packaging, and above all software. Its original business was graphics chips for PC gaming, sold both directly to system builders and through add-in-card partners. Over the past decade the company's growth engine shifted to data-center computing: NVIDIA sells GPU accelerators (the H100, H200, Blackwell, and now Vera Rubin platform generations), CPUs (Grace), high-speed networking (Mellanox/NVLink/InfiniBand and Spectrum-X switches), and full rack-scale AI systems to hyperscale cloud providers (Microsoft, Amazon, Google, Oracle), AI labs (OpenAI, Anthropic, xAI), enterprises, and sovereign/national AI initiatives. Layered on top of the hardware is CUDA, NVIDIA's proprietary parallel-computing software platform and the ecosystem of libraries, frameworks, and developer tools built on it over nearly two decades; this software layer is what locks developers and customers into NVIDIA hardware and lets the company charge premium prices and margins rather than compete purely on chip specifications. Beyond gaming and data center, NVIDIA also sells workstation/visualization GPUs to designers and content creators, and automotive/robotics computing platforms (DRIVE, DriveOS, Jetson) to carmakers and robotics companies, typically through a mix of direct sales, OEM partnerships, and long-term platform licensing deals. Revenue is thus a blend of hardware unit sales at very high margins (driven by scarce leading-edge silicon and system-level design), plus a growing halo of software, networking, and services revenue as NVIDIA sells complete data-center "factories" rather than standalone chips.

Business Segments

NVIDIA has historically reported results in four to five segments; for fiscal year 2026 (ended January 2026) it broke out revenue as follows, before a subsequent reorganization of its segment reporting:

Data Center — ~$193.7 billion (FY2026), ~90% of total revenue, +68% YoY. By far the dominant segment: AI training and inference GPUs (Hopper, Blackwell, and the newer Vera Rubin platform), Grace CPUs, NVLink/InfiniBand/Spectrum-X networking, and full AI "factory" systems sold to hyperscalers, cloud providers, AI labs, enterprises, and governments. In the most recent reported quarter (Q2 FY2027, ended roughly July 2026), this segment alone generated $89.0 billion in quarterly revenue, up 117% year-over-year.
Gaming (and AI PC) — ~$16.0 billion (FY2026), roughly 7% of revenue, +41% YoY. GeForce RTX desktop and laptop GPUs, gaming software technologies (DLSS, Reflex), and increasingly AI-PC-capable hardware for consumers and enthusiasts.
Professional Visualization — ~$3.2 billion (FY2026), roughly 1.5% of revenue, +70% YoY. Workstation GPUs and software for content creation, design, engineering simulation, and "personal AI supercomputer" products aimed at professionals and enterprises.
Automotive and Robotics — ~$2.3 billion (FY2026), roughly 1% of revenue, +39% YoY. DRIVE AGX Orin/Thor compute platforms, DriveOS software, and robotics/physical-AI platforms sold to automakers, autonomous-vehicle developers, and robotics companies.
OEM & Other — a smaller, not separately quantified bucket covering legacy and miscellaneous hardware sales.

Notably, starting in fiscal year 2027 NVIDIA restructured its external reporting, consolidating Gaming, Professional Visualization, Automotive, and related consumer/edge products into a single combined "Edge Computing" segment (which generated $7.2 billion in Q2 FY2027, up 27% YoY) alongside a standalone Data Center segment — reflecting how thoroughly the company's financial identity has become that of a data-center/AI infrastructure business, with everything else now a secondary, blended category. Given this shift, Data Center effectively represents around 90–93% of both revenue and (given its high margins) an even larger share of profit.

Competitors

Data Center / AI accelerators (core business):

AMD (Instinct MI300X/MI350X GPU line) — NVIDIA's closest merchant-silicon rival, competing on price and increasingly on raw specifications
Intel (Gaudi accelerators, though a distant player)
Custom silicon from hyperscalers: Google (TPUs), Amazon (Trainium/Inferentia), Microsoft (Maia), Meta (MTIA), and Broadcom as a key ASIC design partner to several of these
Chinese domestic accelerator makers (Huawei Ascend, Cambricon, and others), dominant within China following U.S. export restrictions

Gaming / consumer graphics:

AMD (Radeon GPUs)
Intel (Arc GPUs)

Professional visualization / workstation:

AMD (Radeon Pro)
Intel

Automotive & robotics:

Qualcomm (Snapdragon Ride)
Mobileye (Intel)
Traditional automotive semiconductor suppliers and in-house silicon efforts at some automakers
Tesla (in-house autonomy compute)

Indirect/broader competitive pressure: the major cloud providers (Microsoft, Amazon, Google) are simultaneously NVIDIA's largest customers and, through their custom chip programs, potential long-term substitutes for merchant GPU purchases.

Competitive Position

NVIDIA's dominant position rests on a combination of hardware performance leadership and, more durably, its CUDA software ecosystem — nearly two decades of developer tools, libraries, and framework integrations that make switching away from NVIDIA hardware costly and technically disruptive for AI developers and enterprises. This software moat, paired with NVIDIA's ability to ship full rack-scale systems (compute, networking, and software together) rather than standalone chips, has let it capture an estimated 75–80%+ share of the AI accelerator market and post gross margins above 70%, extraordinary for a hardware company. Demand visibility remains unusually strong: Jensen Huang has described AI infrastructure spending as still in an "inflection point," with the company guiding to continued high double-digit percentage revenue growth and citing over $500 billion in newly mobilized data-center capital commitments from partners as of mid-2026.

That dominance faces real and growing threats, however. AMD's Instinct MI-series GPUs have closed much of the raw hardware gap with NVIDIA's Blackwell/Vera Rubin chips and undercut on price, while software compatibility layers (OpenAI's Triton compiler, improved AMD ROCm tooling) are gradually narrowing — though not eliminating — the CUDA lock-in advantage. A potentially larger long-term risk comes from NVIDIA's own largest customers: hyperscalers including Google, Amazon, and Microsoft are investing heavily in custom AI silicon (TPUs, Trainium, Maia) explicitly to reduce dependence on NVIDIA GPUs and their associated margins, and these custom-chip efforts collectively already represent a low-double-digit share of AI compute deployment. Geopolitics is another significant swing factor: U.S. export controls have effectively pushed NVIDIA's market share in China to near zero (from roughly two-thirds of the Chinese AI chip market in 2024), ceding that market to domestic players like Huawei; Huang himself has argued the policy has "largely backfired" by accelerating China's chip self-sufficiency, and any escalation or de-escalation of export policy is a material variable for future results. Additional risks include supply-chain concentration (nearly total dependence on TSMC's advanced-node manufacturing and CoWoS packaging), the capital intensity and cyclicality of hyperscaler AI infrastructure spending (raising the question of how durable current growth rates are if AI monetization disappoints), and antitrust/regulatory scrutiny given NVIDIA's outsized market position. On balance, NVIDIA enters the current period as the clear category leader with a wide and reinforcing moat, but one being tested simultaneously on price (AMD), on vertical integration (hyperscaler custom silicon), and on geopolitics (China export policy) in ways that will shape whether its extraordinary current margins and share prove durable over the coming years.