NVIDIA Corporation

Overview

In July 2026, the chief executive of the world’s most valuable company flew to Tokyo to thank a retired video-game executive for a decision made thirty years earlier — a roughly $5 million lifeline, approved around 1995 or 1996, that kept NVIDIA alive long enough to build the chip that saved it (Techstartups; Inc). The reunion, held at a private NVIDIA event in Tokyo’s Akihabara district on July 15, 2026, brought Jensen Huang together with former Sega president Shoichiro Irimajiri and game designer Yu Suzuki (Techstartups). It closed a loop that most company profiles never open: before NVIDIA became the backbone of the AI boom and the first company to reach a $5 trillion market value, it was about a month of payroll away from vanishing (Reuters; Moneywise).

The five-million-dollar lifeline

NVIDIA was founded in April 1993 by three computer scientists: Jen-Hsun (“Jensen”) Huang, previously a director at LSI Logic and a microprocessor designer at AMD; Chris Malachowsky, an engineer at Sun Microsystems; and Curtis Priem, a senior staff engineer and graphics chip designer at IBM and Sun (Britannica). The name NVIDIA combines invidia, the Latin word for envy, with the abbreviation NV — short for “next vision” — which the founders had used to label their early files (Britannica). In its founding year the company received $20 million in venture funding from firms including Sequoia Capital, support that carried it toward its 1999 IPO (Britannica).

The early technology bet was wrong in a way that nearly ended the company. NVIDIA built its first chips around quadrilateral-based rendering, while the industry — led by Microsoft’s DirectX platform — standardized on triangles (Techstartups). Huang had landed a contract to build graphics technology for Sega’s next-generation console, and the failing architecture put both the contract and the startup at risk (Techstartups). He flew to Japan, told Sega the chip could not be delivered as planned, asked to be released from the project — and then asked Sega to provide the remaining $5 million anyway (Techstartups).

Irimajiri said he was no longer betting on the product; he was betting on Huang (Inc). He described Huang as “a genuinely honest, straightforward individual who never played games and never misrepresented what he believed,” and said it would have been “a tremendous loss to see Nvidia fail with someone of his caliber at the helm” (Inc). Sega confirmed to Inc. that it approved the roughly $5 million around 1995 or 1996, though precise records no longer exist and it is unclear whether the money came through contract payments, an equity investment, or both (Inc). The money bought about six months to abandon the old architecture and build the triangle-based RIVA 128 (Inc).

What followed was a gamble Huang later called the defining moment that saved the company (Moneywise). To fund the RIVA 128 effort, NVIDIA cut its staff from about 100 to about 40 (Moneywise). Unable to afford a prototype, Huang bought the inventory of a struggling emulation company on the spot — with about half of the company’s remaining cash — and used software emulation to virtually test the chip before committing to production at TSMC (Moneywise). When the RIVA 128 shipped in August 1997, NVIDIA had one month of payroll left (Moneywise). It sold roughly a million units in its first four months, a win that pushed the company toward its 1999 IPO (Moneywise). The desperation hardened into the company’s unofficial motto — “Our company is thirty days from going out of business” — which Huang used for years as his opener for internal presentations (Moneywise).

Sega later sold its NVIDIA stake for roughly $15 million, a respectable return at the time (Techstartups). Huang has since estimated that, measured against NVIDIA’s multi-trillion-dollar market value, the stake could have been worth nearly $1 trillion had Sega held it — his own estimate, not a computed figure (Techstartups).

From RIVA to CUDA

The RIVA series gave NVIDIA its foothold: the RIVA line launched in 1997 and the RIVA TNT followed in 1998, expanding the company’s reputation in PC graphics (Britannica). In 1999, its IPO year, NVIDIA released the GeForce 256, which brought onboard transformation and lighting to consumer-grade hardware, and the success of these products won it the contract to develop the hardware for Microsoft’s Xbox, with a $200 million advance (Britannica). Sony later selected NVIDIA to help design the RSX Reality Synthesizer graphics processor for the PlayStation 3 (Britannica).

Growth came through acquisition from early on: NVIDIA bought graphics rival 3dfx Interactive in 2000, mobile communications maker Icera in 2011, and compiler technology maker PGI in 2013, beginning a decade in which it completed more than a dozen deals (Britannica). The decisive product bet of the 2000s arrived in 2006 with CUDA, the Compute Unified Device Architecture platform and programming model that turned GPUs into general-purpose parallel computers (Britannica). The same year, the company received its first U.S. Department of Justice subpoena over possible antitrust violations in the graphics industry (Britannica).

The decade’s end brought both recognition and trouble. Forbes named NVIDIA its “Company of the Year” for 2007 (Britannica). Then reports of abnormal failure rates in certain mobile chips and GPUs — manufacturing defects the company would not fully disclose — led to a class-action lawsuit that was settled in September 2010 (Britannica). In January 2011 NVIDIA signed a $1.5 billion cross-licensing deal with IBM that ended all pending litigation between the two companies, and it released its first-generation Tegra mobile processor the same year (Britannica). Later in the decade it pushed into vehicles — announcing a 2017 partnership with Toyota for its Drive PX autonomous-driving platform — and into the metaverse with the 2020 beta of Omniverse (Britannica).

The AI era and the trillion-dollar run

CUDA’s payoff arrived with the deep-learning boom. OpenAI developed the technology behind ChatGPT using a supercomputer powered by 10,000 NVIDIA GPUs, and when ChatGPT reached the public in 2022, rivals scrambled for NVIDIA hardware and investors for its shares (Britannica). The company reinforced its position with successive data-center GPU generations: the Blackwell architecture, introduced in 2024, and its successor Rubin, unveiled in early 2026, both designed for large-scale AI training and inference and widely adopted by major cloud providers (Britannica).

The market milestones followed in quick succession. In July 2025 NVIDIA became the first company to reach a $4 trillion market capitalization (Britannica). On October 29, 2025, it became the first to reach $5 trillion, closing at $207.04 for a $5.03 trillion valuation after announcing $500 billion in AI chip orders (Reuters). Since the launch of ChatGPT in 2022, its shares had climbed roughly twelvefold (Reuters). At those prices, Huang’s stake was worth about $179.2 billion, making him the world’s eighth-richest person on the Forbes list (Reuters). Britannica’s data page listed a $5.53 trillion market capitalization and $302.97 billion in annual revenue at the September 28, 2026 market close (Britannica).

The company kept expanding beyond chips. In September 2025 it agreed with OpenAI to build at least 10 gigawatts of computing capacity, providing up to $100 billion in hardware and data-center capacity as phases completed (Britannica). In November 2025 it bought about $2 billion in Synopsys stock alongside an expanded collaboration (Britannica). In December 2025 it agreed to pay about $20 billion for most of the assets of Groq, an AI chip startup specializing in fast AI-model inference, licensing Groq’s technology and hiring its founder Jonathan Ross while Groq stayed independent (Britannica). In September 2026 it agreed to acquire Hugging Face, the developer platform for sharing AI models, for $12.9 billion, saying the site would continue to operate as an open platform (Britannica).

The deals that broke and the fine nobody mentions

The $40 billion bid for Arm, announced in September 2020, would have been the largest semiconductor merger in history (Silicon UK). It drew investigations in the United Kingdom — where the government ordered a full Phase 2 probe — the European Union, and the United States, where the Federal Trade Commission sued in December 2021 to block it (Silicon UK). On February 8, 2022, NVIDIA and SoftBank terminated the agreement, citing significant regulatory challenges (Silicon UK). The cancellation had a price: SoftBank kept the $1.25 billion NVIDIA had prepaid, recording it as fourth-quarter profit, while NVIDIA retained a 20-year Arm license (Silicon UK). Arm’s then-CEO Simon Segars stepped down and was replaced by Rene Haas, a former NVIDIA executive (Silicon UK).

A quieter regulatory episode came three months later. On May 6, 2022, the SEC announced settled charges against NVIDIA for inadequate disclosures about cryptomining’s impact on its gaming business: during fiscal 2018, the company had failed to disclose that cryptomining was a significant element of its material revenue growth from gaming GPUs, even as it told investors about crypto-driven demand elsewhere in the business (SEC). Without admitting or denying the findings, NVIDIA agreed to a cease-and-desist order and a $5.5 million penalty (SEC). Scrutiny continued: in September 2023, French authorities raided NVIDIA’s offices in a cloud-computing antitrust inquiry (Britannica).

The least-told part of the growth story may be the networking business built on Mellanox. NVIDIA announced the acquisition of the Israeli company in March 2019 for about $6.9 billion in cash and completed it on April 27, 2020 (Calcalistech). Networking revenue rose from $8.57 billion in fiscal 2024 to $13.0 billion in fiscal 2025, then jumped 140% to $31.38 billion in fiscal 2026, which ended in January 2026 (Calcalistech). In the first half of fiscal 2027 alone, networking revenue reached $32.3 billion — already more than the entire previous year (Calcalistech). As of 2025, the company held a 92% share of the discrete graphics segment (Britannica).

What this profile does not claim

It does not give the exact founding day — this profile uses Britannica’s “April 1993” because no fully-read source verified a specific date. It does not state an employee count, which could not be verified from a fully-read source. It does not repeat the Denny’s founding-meeting story, which is widely covered elsewhere and is the lead of this site’s Jensen Huang profile. The “nearly $1 trillion” figure for Sega’s foregone stake is Huang’s own estimate, not a computed valuation. The Hugging Face acquisition is reported as agreed in September 2026; its completion is not asserted here.