Nvidia

Nvidia Corporation
	File:Nvidia logo.svg Logo since 2023
Company type	Public
Industry	Technology, Semiconductors
Founded	April 5, 1993; 33 years ago in Santa Clara, California, U.S.
Founder	Jensen Huang, Chris Malachowsky, Curtis Priem
Headquarters	Santa Clara, California, United States
Area served	Worldwide
Key people	Jensen Huang (CEO)
Products	Graphics processing units (GPUs), system on a chips (SoCs), application programming interfaces (APIs), gaming hardware, Tegra processors, GeForce Now cloud gaming platform
Revenue	US$130.5 billion (FY2025)
Operating income	US$81.5 billion (FY2025)
Net income	US$72.9 billion (FY2025)
Total assets	US$111.6 billion (FY2025)
Total equity	US$79.3 billion (FY2025)
Number of employees	36000 (2025)
Website	www.nvidia.com

Nvidia Corporation (presented as nVIDIA) is a U.S.-based tech firm located in Santa Clara, California. Established in 1993 by Jensen Huang, Chris Malachowsky, and Curtis Priem, Nvidia specializes in the creation of graphics processing units (GPUs), system on a chips (SoCs), and application programming interfaces (APIs) serving the fields of data science, high-performance computing, as well as mobile and automotive sectors.

Originally concentrating on gaming GPUs, Nvidia broadened its scope to include artificial intelligence (AI), professional visualization, and supercomputing technologies. Its portfolio features GeForce GPUs tailored for gaming and creative tasks, alongside professional GPUs designed for edge computing, scientific exploration, and industrial uses. By the first quarter of 2025, Nvidia held a dominant 92% portion of the discrete desktop GPU market.

During the early 2000s, Nvidia committed over $1 billion toward developing CUDA, a platform that allows GPUs to execute massively parallel programs suited for compute-heavy workloads. By 2025, the company commanded more than 80% of the GPU sector dedicated to AI model training and deployment and provided chips powering upwards of 75% of the world's TOP500 supercomputers. Additionally, Nvidia markets gaming hardware such as its Shield series and manages the GeForce Now cloud gaming platform. Its Tegra processors cater to smartphones, tablets, and automotive infotainment systems.

Nvidia achieved a valuation of US$1 trillion in 2023, becoming the seventh American enterprise to reach this milestone. In 2025, it was the first company to exceed a US$4 trillion market cap amid surging demand for data center equipment fueled by AI advancements. This expansion underscores Nvidia’s robust business approach, visionary leadership under CEO Jensen Huang, and growth supported by subsidiaries like Bright Computing, Cumulus Networks, DeepMap, Mellanox Technologies, and Mental Images.

For fiscal year 2025, Nvidia reported revenues amounting to US$130.5 billion, operating profits of US$81.5 billion, net income reaching US$72.9 billion, total assets valued at US$111.6 billion, equity standing at US$79.3 billion, and a global workforce near 36,000 employees.

Founding

Nvidia was established on April 5, 1993, by Jensen Huang, a Taiwanese-American electrical engineer previously employed at LSI Logic and AMD who continues as CEO; Chris Malachowsky, formerly with Sun Microsystems; and Curtis Priem, who had been a senior engineer at both IBM and Sun Microsystems. The trio decided to form the company following a meeting at a Denny’s restaurant in East San Jose during late 1992.

At that time, Malachowsky and Priem were planning to depart from Sun Microsystems while Huang was managing a division within LSI. They collectively envisioned graphics processors as a solution to computational problems that general-purpose CPUs could not efficiently handle. Huang recognized video games as a "killer application" due to their intensive computational requirements coupled with significant market size—offering a viable model to finance research and development.

The founders orchestrated their resignations strategically: Priem exited Sun Microsystems first in December 1992, which was soon followed by Huang and Malachowsky stepping down. By early 1993, operations began from Priem’s townhouse in Fremont with an initial capital injection of $40,000. Subsequently, the firm secured $20 million in venture capital investments from Sequoia Capital, Sutter Hill Ventures, among others.

Throughout the fiercely competitive graphics industry of the late 1990s, Nvidia along with ATI Technologies were the only startup companies that endured; ATI eventually merged into AMD. The name "Nvidia" was derived from the Latin term meaning "envy," selected after discarding alternatives based on founders’ names or internal product codenames. The company initially operated out of Sunnyvale, California before expanding its facilities.

Initial graphics accelerator

Nvidia's debut graphics accelerator, the NV1, utilized quadrilateral primitives through forward texture mapping, setting it apart from rivals who predominantly used triangles. However, with the introduction of Microsoft's DirectX platform, which exclusively supported triangle primitives via the Direct3D API, the NV1 struggled to gain traction in the market.

Collaborating with Sega, Nvidia supplied the graphics processor for the Dreamcast console, but its technology fell behind competing solutions. This situation compelled Nvidia to confront the dilemma of either persisting with a project unlikely to succeed or risking financial insolvency.

During this period, Sega's president, Shoichiro Irimajiri, met with Huang and disclosed that Sega intended to select an alternative supplier. Nonetheless, Sega invested $5 million in Nvidia, a critical infusion of capital that Huang later acknowledged as pivotal in sustaining the company's operations.

In 1996, Huang downsized Nvidia’s staff from 100 to 40 employees and shifted focus toward creating a graphics accelerator optimized for triangle primitives: the RIVA 128. By its release in August 1997, Nvidia had funds sufficient for only one month’s payroll, a financial strain that inspired the company’s unofficial motto: "Our company is thirty days from going out of business."

The RIVA 128 achieved sales of approximately one million units within four months, generating revenue that enabled development of subsequent products. The 1998 launch of the RIVA TNT solidified Nvidia’s status as a pioneer in graphics technology.

Transition to public company

Nvidia became publicly traded on January 22, 1999. For Sega's president Irimajiri, investing in Nvidia proved advantageous; after departing Sega in 2000, he sold his shares for $15 million.

In late 1999, Nvidia introduced the GeForce 256 (NV10), its inaugural GPU aimed at consumer-grade 3D hardware markets. This product featured onboard transformation and lighting (T&L) capabilities. Operating at 120 MHz with four pixel pipelines, it offered superior video acceleration and significantly outperformed competitors.

This success secured Nvidia a $200 million contract to develop graphics components for Microsoft's Xbox console, although it temporarily diverted some of its top engineering talent. The GeForce 2 GTS was released in summer 2000. Later that year, in December, Nvidia agreed to acquire essential intellectual property from rival 3dfx Interactive, completing the acquisition by April 2002 and thereby enhancing its leadership in graphics technology.

In 2001, Nvidia replaced Enron in the S&P 500 index, which increased demand among institutional investors.

Nvidia expanded through strategic acquisitions: Exluna in July 2002 for software-rendering tools; MediaQ in August 2003 for roughly $70 million, leading to the GoForce product line; and iReady in April 2004 for high-performance TCP offload engines and iSCSI controllers. In December 2004, Nvidia collaborated with Sony to develop the RSX graphics processor for the PlayStation 3. On December 14, 2005, Nvidia acquired ULI Electronics, a chipset southbridge supplier to competitor ATI. In March 2006, Nvidia purchased Hybrid Graphics. Later that year, both Nvidia and AMD (which had acquired ATI) faced subpoenas from the U.S. Department of Justice concerning potential antitrust issues within the graphics card sector.

2007–2014

In 2007, Forbes honored Nvidia as Company of the Year, acknowledging its significant progress over the preceding five years. On January 5 of that year, Nvidia finalized the acquisition of PortalPlayer, Inc.. The following year, in February 2008, Nvidia purchased Ageia, the creator of the PhysX physics engine, incorporating this technology into its upcoming GPU designs.

By July 2008, Nvidia recorded a $200 million write-down due to manufacturing flaws that caused unusually high failure rates in certain mobile chipsets and GPUs. These defects impacted laptops from manufacturers such as Apple Inc., Dell, and HP, resulting in a class action lawsuit that was resolved in 2010 with compensation for repairs or device replacements. On January 10, 2011, Nvidia and Intel settled patent disputes by entering into a six-year cross-licensing agreement worth $1.5 billion.

November 2011 marked the release of the ARM-based Tegra 3 system-on-chip (SoC), notable for being the first quad-core mobile CPU. Earlier that year, in May, Nvidia acquired UK-based baseband chipmaker Icera for $367 million. The company introduced Tegra 4 and the Nvidia Shield—an Android-powered handheld gaming device utilizing the new SoC—in January 2013. To bolster its high-performance computing (HPC) capabilities, Nvidia acquired PGI from STMicroelectronics in July 2013.

In February 2013, Nvidia revealed plans for a new headquarters featuring two triangular structures designed to symbolize "the fundamental building block of computer graphics," reflecting its commitment to innovation and design leadership.

During 2014, Nvidia ported Valve’s titles Portal and Half-Life 2 to its Shield Tablet through Lightspeed Studio. Since then, the company has strategically diversified its focus into three primary sectors: gaming, automotive electronics, and mobile devices.

That same year, Nvidia prevailed in litigation related to its 2000 purchase of intellectual property from bankrupt 3dfx Interactive. The U.S. Court of Appeals confirmed that Nvidia paid fair market value for these assets, reinforcing its position in intellectual property acquisitions.

2016–2018

On May 6, 2016, Nvidia introduced the initial GPUs of the GeForce 10 series, specifically the GTX 1080 and GTX 1070 models, built on the new Pascal platform. These GPUs surpassed the performance of the Maxwell-based Titan X and utilized GDDR5X and GDDR5 memory fabricated on a 16 nm process. Pascal architecture also debuted simultaneous multi-projection (SMP) technology to improve rendering across multiple monitors and enhance virtual reality (VR) experiences. By late 2017, laptops equipped with these GPUs met Nvidia’s "Max-Q" design criteria for thin and efficient gaming notebooks.

In July 2016, Nvidia resolved a false advertising lawsuit concerning the GTX 970, which was unable to fully leverage its advertised 4 GB VRAM due to hardware limitations. In May 2017, a partnership with Toyota was announced to integrate Nvidia’s Drive PX AI platform into autonomous vehicle systems. July of that year saw an extensive AI collaboration with Chinese technology leader Baidu, encompassing cloud computing, autonomous driving initiatives, consumer electronics, and Baidu’s PaddlePaddle AI framework; notably, Nvidia’s Drive PX 2 became Baidu’s foundational autonomous driving platform.

The Titan V GPU was launched on December 7, 2017, followed by the Quadro GV100 on March 27, 2018. Later that year, on September 27, Nvidia released the RTX 2080 GPUs. Additionally, Google incorporated Nvidia’s Tesla P4 GPUs into its Cloud AI offerings during 2018.

Mid-2018 brought challenges regarding macOS Mojave compatibility for Nvidia GPUs installed in legacy Mac Pro machines. Apple mandated 'Metal compatible' graphics cards for Mojave support but excluded certain Nvidia models that had functioned under earlier macOS versions using Nvidia’s web drivers. Nvidia indicated that driver updates depended on Apple’s authorization, highlighting tensions between the two companies. Reports from early 2019 suggested Apple’s management was hesitant to support Nvidia within macOS due to prior strained relations and Apple's emphasis on developing proprietary GPU technology. Consequently, Mac users were advised to consider alternative supported graphics solutions such as AMD Radeon.

2019 acquisition of Mellanox Technologies

On March 11, 2019, Nvidia revealed its plan to acquire Mellanox Technologies for $6.9 billion, aiming to strengthen its foothold in the high-performance computing sector. By May 2019, Nvidia introduced RTX Studio laptops, promoting performance gains of up to seven times compared to leading MacBook Pro models in professional software such as Maya and RedCine-X Pro. In August 2019, Nvidia launched Minecraft RTX, which incorporated real-time DXR ray tracing exclusively on the Windows 10 platform, significantly enhancing lighting and reflection effects through path tracing technology.

2020–2023

In May 2020, Nvidia expanded its networking capabilities by acquiring Cumulus Networks and integrating it with Mellanox’s operations. During the COVID-19 crisis, Nvidia contributed by developing an open-source ventilator design to help alleviate shortages of critical medical equipment. On May 14, 2020, the company unveiled the Ampere GPU microarchitecture alongside the A100 GPU accelerator, designed to elevate data center performance.

By July 2020, Nvidia entered negotiations to purchase UK-based chip designer Arm from SoftBank for $32 billion. The GeForce 30 series GPUs built on the Ampere architecture were launched on September 1, 2020. Shortly thereafter, on September 13, Nvidia announced a $40 billion deal to acquire Arm, with SoftBank retaining a minority stake of 10%.

In October 2020, Nvidia pledged $100 million towards Cambridge-1, the United Kingdom’s most powerful supercomputer dedicated to AI-driven healthcare research. That same month marked the discontinuation of the Quadro workstation GPU brand in favor of unifying under the Nvidia RTX branding aligned with Ampere technology.

Regulatory scrutiny intensified in August and October 2021 when both the UK's Competition and Markets Authority and the European Commission initiated investigations into the Arm acquisition due to antitrust concerns. These regulatory obstacles prompted Nvidia and SoftBank to terminate the transaction in February 2022. Additionally, Nvidia faced a cyberattack toward the end of 2021.

In early 2022, CEO Jensen Huang expressed willingness to collaborate with Intel for semiconductor manufacturing, signaling a strategic move toward diversifying foundry partnerships. In April of that year, Nvidia announced plans to establish a new research facility in Yerevan, Armenia.

May 2022 marked the inauguration of Voyager, a significant new structure at Nvidia’s expanded headquarters featuring cutting-edge architectural elements. In September 2022, Nvidia unveiled Drive Thor, an advanced automotive-grade chip consolidating multiple vehicle functions into a single integrated system.

That same month, Nvidia partnered with the Broad Institute of MIT and Harvard to embed its AI-driven healthcare software suite Clara within biomedical research platforms.

Following U.S. export restrictions imposed in October 2022 on advanced chips destined for China, Nvidia introduced the A800 GPU specifically designed to comply with these regulatory requirements.

In September 2023, Getty Images collaborated with Nvidia to launch Generative AI by Getty Images powered by Nvidia’s Edify model from its Picasso AI collection.

On September 26, 2023, Denny's CEO Kelli Valade joined Jensen Huang at Nvidia’s original founding location in East San Jose to celebrate the company’s growth into a $1 trillion enterprise. Demand for Nvidia's Hopper (H100) GPUs remained robust among industry leaders such as Oracle’s Larry Ellison and Tesla’s Elon Musk.

In October 2023, reports surfaced that Nvidia had commenced development of ARM-based CPUs tailored for Microsoft’s Windows platform with an anticipated release by 2025, marking a strategic expansion into CPU markets traditionally dominated by Intel.

2024–2025

At the start of 2024, Nvidia intensified its lobbying activities in Washington, D.C., recruiting government relations experts from influential agencies amid ongoing AI regulatory debates. Although the company allocated $350,000 to lobbying efforts in 2023—a figure modest relative to other AI technology firms—it deliberately enhanced its governmental engagement.

Analysts at Raymond James Financial projected the H100 GPU’s price to range between $25,000 and $30,000, while secondary market values surpassed $40,000. Leading technology corporations invested billions in capital expenditures acquiring these GPUs to support generative AI initiatives.

By February 2024, Nvidia had become a highly desirable employer within Silicon Valley, offering competitive salaries with half of its workforce earning in excess of $228,000 annually. Due to the substantial value of Nvidia GPUs, transportation security was heightened, including the use of armored vehicles.

On March 1, 2024, Nvidia joined an exclusive category by reaching a market valuation above $2 trillion in just 180 days—achieving this milestone significantly faster than Apple and Microsoft. Later that month, the company introduced its Blackwell AI chip and corresponding microarchitecture, honoring mathematician David Blackwell.

In April 2024, reports emerged indicating that China had obtained Nvidia chips banned for export through servers supplied by Supermicro and Dell. By June, antitrust investigations were initiated by the FTC and DOJ targeting Nvidia, Microsoft, and OpenAI; these inquiries focused on industry dominance rather than corporate mergers, prompted by internal concerns over rapid AI sector expansion.

On June 18, 2024, Nvidia became the most valuable company worldwide with a market capitalization exceeding $3.3 trillion. That same month saw a strategic alliance formed with Trend Micro to develop AI-enhanced cybersecurity solutions leveraging Nvidia’s platforms to improve data protection and threat response.

October 2024 marked the debut of NVLM 1.0, Nvidia’s open-source multimodal large language model featuring 72 billion parameters designed to advance AI text processing capabilities.

In November 2024, Nvidia was added to the Dow Jones Industrial Average. Morgan Stanley reported that all Blackwell chip production for 2025 was fully booked. Additionally, Nvidia acquired 1.2 million shares in Nebius Group and secured the #3 position on Forbes’ "Best Places to Work" list.

By January 7, 2025, Nvidia’s market capitalization reached $3.66 trillion—more than twice the combined valuations of AMD, ARM, Broadcom, and Intel. However, later that month the company experienced an unprecedented single-day market value decline of $600 billion due to competition from Chinese AI startup DeepSeek.

In April 2025, Nvidia launched the Llama-3.1-Nemotron-Ultra-253B-v1 reasoning language model under its Open Model License in three configurations: Nano, Super, and Ultra.

Despite forecasting second-quarter revenue below expectations in May—affected by U.S. export restrictions on AI chips destined for China—Nvidia’s stock price climbed as investors maintained confidence in sustained AI demand.

During July 2025, Nvidia acquired Canadian artificial intelligence firm CentL for an estimated $400 million. On July 10, the company closed trading with a historic market capitalization surpassing $4 trillion—the first corporation ever to reach this threshold—outvaluing all publicly listed companies in the United Kingdom combined.

Fabless manufacturing

Nvidia operates under a fabless manufacturing strategy, delegating all wafer production, assembly, testing, and packaging processes to third-party vendors. This methodology reduces the need for substantial capital expenditure and lowers manufacturing risks. Internally, Nvidia dedicates its efforts to product innovation, quality control, marketing initiatives, and customer service, while occasionally sourcing critical elements such as memory and substrates directly.

Corporate affairs

Sales by business unit (2023)
Business unit	Sales (billion $)	Share
Compute & networking	47.4	77.8%
Graphics	13.5	22.2%

Sales by region (2023)
Region	Sales (billion $)	Share
United States	27.0	44.3%
Taiwan	13.4	22.0%
China	10.3	16.9%
Other countries	10.2	16.8%

Leadership

As of March 2024, Nvidia's senior management team consists of:

Jensen Huang, founder, president, and chief executive officer
Chris Malachowsky, founder and Nvidia fellow
Colette Kress, executive vice president and chief financial officer
Jay Puri, executive vice president of global field operations
Debora Shoquist, executive vice president of operations
Tim Teter, executive vice president, general counsel, and corporate secretary

Board of directors

As of November 2024, the board of directors at Nvidia includes:

Rob Burgess (former CEO of Macromedia Inc.)
Tench Coxe (ex-managing director at Sutter Hill Ventures)
John Dabiri (engineer and professor at Caltech)
Persis Drell (physicist and Stanford University professor)
Jensen Huang (co-founder, CEO, and president)
Dawn Hudson (former chief marketing officer of the National Football League)
Harvey C. Jones (managing partner at Square Wave Ventures)
Melissa B. Lora (former president of Taco Bell International)
Stephen Neal (lead independent director; former CEO and Chairman Emeritus of Cooley LLP)
Ellen Ochoa (former director of NASA Johnson Space Center)
Brooke Seawell (venture partner at New Enterprise Associates)
Aarti Shah (former senior vice president and chief information officer at Eli Lilly)
Mark Stevens (managing partner at S-Cubed Capital)

Finances

10-year financials (2016–2025)
Year	Revenue (mn. US$)	Net income (mn. US$)	Employees
2016	5,010	614	9,227
2017	6,910	1,666	10,299
2018	9,714	3,047	11,528
2019	11,716	4,141	13,277
2020	10,918	2,796	13,775
2021	16,675	4,332	18,975
2022	26,914	9,752	22,473
2023	26,974	4,368	26,000
2024	60,922	29,760	29,600
2025	130,497	72,880	36,000

During fiscal year 2020, Nvidia posted a net profit of US$2.796 billion on revenues totaling US$10.918 billion—a 6.8% decrease compared to the previous year. Early in 2021, shares traded above $531 each with a market capitalization surpassing US$328.7 billion. By the third quarter of 2024, the company’s market value had escalated to nearly US$2.98 trillion.

In the second quarter of 2020 alone, revenue hit $3.87 billion—a 50% rise year-over-year—propelled by increased demand for computing solutions amid the global pandemic. CFO Colette Kress emphasized the growing importance of enterprise technologies such as Nvidia laptops and virtual workstation products that facilitate remote work and collaboration.

By May 2023, Nvidia’s market capitalization exceeded $1 trillion during trading hours and expanded further to $1.2 trillion by November. Due to its size and market influence, Nvidia is recognized as part of Bloomberg’s "Magnificent Seven," denoting the largest and most impactful technology firms listed on public exchanges.

Ownership

By early 2024, Nvidia's principal shareholders comprised prominent institutional investors and company insiders: The Vanguard Group held 8.28%, BlackRock owned 5.62%, Fidelity Investments possessed 5.16%, State Street Corporation controlled 3.71%, and founder Jensen Huang maintained a 3.51% stake, among others. These key stakeholders play a crucial role in shaping Nvidia's governance and strategic decisions.

GPU Technology Conference

The GPU Technology Conference (GTC), initiated by Nvidia in 2009 in San Jose, California, originally served as a venue to highlight GPU-based computing innovations. Over the years, it has transformed into a premier event emphasizing fields such as artificial intelligence, deep learning, autonomous driving, healthcare advancements, high-performance computing, and educational programs through the Nvidia Deep Learning Institute (DLI). Attendance figures have surged markedly, with more than 8,400 attendees recorded in 2018 and roughly 59,000 registrants participating in the virtual GTC held in 2020. Following several online-only editions, GTC resumed its physical format in March 2024, reaffirming Nvidia’s leadership position and dedication to promoting innovation globally.

Product families

Nvidia's extensive product range includes graphics processing units, wireless communication components, and automotive-related hardware and software solutions, such as:

GeForce, consumer-oriented graphics cards
RTX, professional GPUs succeeding GTX and Quadro lines for advanced visual computing
NVS, multi-display graphics processors designed for business applications
Tegra, a system-on-chip lineup tailored for mobile devices
Tesla, GPUs focused on professional and scientific image processing
nForce, motherboard chipsets compatible with Intel and AMD CPUs
GRID, technology facilitating graphics virtualization through hardware and services
Shield gaming devices including Shield Portable, Shield Tablet, and Shield TV
Drive, an integrated hardware and software platform for autonomous driving, featuring the high-performance Drive PX system and Driveworks operating system for self-driving vehicles
BlueField, data processing units acquired via Mellanox Technologies
The datacenter/server CPU known as Grace, introduced in 2023 to support AI workloads and high-performance computing
DGX, an enterprise-grade platform optimized for deep learning tasks
Maxine, a suite of AI-powered conferencing tools designed for developers

This broad array of products strengthens Nvidia’s competitive standing across graphics processing, artificial intelligence, automotive technology, and data center industries.

Open-source software support

Prior to September 23, 2013, Nvidia had not released detailed documentation for its advanced hardware components, which hindered development of free and open-source drivers without resorting to reverse engineering.

Nvidia provides proprietary binary GeForce drivers compatible with X.Org servers as well as an open-source library that interfaces with Linux, FreeBSD, or Solaris kernels alongside its closed-source components. The company formerly supported a limited-functionality open-source driver focused on 2D acceleration but has since discontinued it.

The proprietary licensing of Nvidia’s drivers has been criticized by advocates of free software. In 2012, Linus Torvalds openly criticized Nvidia’s approach to Linux support. Many users of Linux and BSD prefer open-source drivers and consider Nvidia’s binary-only policy insufficient when compared to competitors such as Intel and AMD, which supply hardware documentation and contribute to open-source driver development.

Nvidia’s proprietary drivers are available only for x86/x64 and ARMv7-A architectures; this restricts capabilities like CUDA on other platforms. Some users have claimed that Linux drivers impose artificial limitations—for example, restricting the number of monitors that can be used simultaneously—though Nvidia has not publicly responded to these allegations.

Starting with the Maxwell GPU architecture in 2014, Nvidia implemented a requirement for signed firmware to enable full feature access on its graphics cards. On May 12, 2022, the company announced it would open-source its GPU kernel modules—a notable shift toward embracing open-source strategies. In 2023, support for Nvidia firmware was incorporated into the Nouveau open-source driver project, enhancing power management and enabling GPU reclocking on Turing-generation GPUs and newer models.

On July 21, 2025, Nvidia expanded CUDA compatibility to include the RISC-V architecture. This extension broadens its artificial intelligence ecosystem and aligns with global movements toward semiconductor self-reliance—particularly within China—thereby increasing Nvidia’s influence by extending its offerings to emerging processor platforms.

Nvidia Open-Source Initiatives

Deep Learning

Nvidia's graphics processing units (GPUs) play a crucial role in deep learning and accelerated data analytics through the company's proprietary CUDA platform, which facilitates the parallel execution of BLAS operations essential for machine learning algorithms. These GPUs have been widely adopted by researchers, laboratories, technology companies, and corporations worldwide. Nvidia was instrumental in the 2009 breakthrough of deep learning when the integration of neural networks with its GPUs dramatically enhanced computational performance. The Google Brain team notably utilized Nvidia GPUs to develop deep neural networks, with Andrew Ng reporting a 100-fold acceleration in deep learning processes. Although Tesla initially relied on Nvidia hardware for autonomous vehicle development, it transitioned to proprietary systems by 2019. Additionally, Nvidia’s GPUs power cloud platforms such as Google Cloud, advancing AI and deep learning capabilities across various sectors.

DGX

Nvidia’s DGX series represents a collection of supercomputers engineered to expedite deep learning and artificial intelligence workloads.

In April 2016, Nvidia unveiled the DGX-1, an 8-GPU system integrating both hardware and specialized deep learning software to optimize AI computations. The first DGX-1 unit was donated to OpenAI in August 2016, drastically shortening AI training durations from six days to just two hours. Nvidia also introduced virtual machines equipped with Tesla K80 and P100 GPUs, accessible via Google Cloud starting November 2016. Microsoft previewed GPU-powered servers in its N series featuring Nvidia Tesla K80s, while AWS launched P2 instances capable of supporting up to 16 Tesla K80 GPUs. During that period, Nvidia collaborated with IBM to develop IBM PowerAI, a software toolkit designed to enhance Watson’s AI functionality. Nvidia also provides its own Deep Learning SDK. In 2017, these GPU technologies were deployed at Japan’s Riken Center for Advanced Intelligence Project in partnership with Fujitsu. This advancement in deep learning technology contributed significantly to Nvidia’s robust financial results in 2017.

In May 2018, Nvidia researchers showcased a system enabling robots to acquire skills by observing human actions, marking progress toward universal robotic control for future generations. Beyond manufacturing GPUs, Nvidia supports high-performance parallel computing environments that empower scientists and researchers to efficiently execute complex computational tasks.

Robotics

In 2020, Nvidia introduced Omniverse, a virtual platform designed for engineers, alongside open-sourcing Isaac Sim—a simulation tool utilizing Omniverse for training robots through physics-accurate virtual environments.

By 2024, CEO Jensen Huang redirected Nvidia’s strategic focus toward humanoid robotics and autonomous vehicles, identifying these sectors as primary avenues for growth.

In 2025, Nvidia launched Isaac GR00T N1, an open-source foundational model intended to accelerate development in humanoid robotics. Early adopters include Neura Robotics, 1X Technologies, and Vention. This initiative enhances Nvidia’s competitive position in robotics by merging artificial intelligence with simulation technologies to drive innovation and operational capabilities worldwide.

Inception Program

Nvidia's Inception Program is designed to assist start-ups that are innovating in artificial intelligence and data science fields. Winners of the program are revealed at Nvidia's GTC Conference. The initiative expanded from 1,300 members in May 2017 to 2,800 by March 2018, eventually surpassing 8,500 companies across 90 nations by August 2021, with total funding amounting to US$60 billion.

GTX 970 hardware specifications

It was observed by users that although the GeForce GTX 970 was equipped with 4 GB of memory, it seldom utilized more than 3.5 GB effectively. Nvidia later acknowledged that the specifications had been modified prior to launch, which affected performance when memory usage exceeded the 3.5 GB mark. Compared to the GTX 980, the GTX 970 featured reduced L2 cache (1.75 MB versus 2 MB) and fewer ROPs (56 instead of 64), with its memory divided into a faster 3.5 GB segment and a slower 0.5 GB segment.

Initially, Nvidia pledged a driver update to resolve these performance concerns but subsequently stated that no dedicated update would be provided. The company extended refund options to customers impacted by the issue. In February 2015, CEO Jensen Huang issued a public apology regarding the matter. Following this, a class-action lawsuit accusing Nvidia and Gigabyte Technology of false advertising was filed.

The architectural design of the GTX 970 permits disabling units containing L2 cache and ROPs without turning off entire memory controllers, resulting in memory buses operating at different speeds—3.5 GB on a 224-bit bus and an additional 0.5 GB on a slower 32-bit bus.

On July 27, 2016, Nvidia reached a preliminary agreement to settle the class-action lawsuit, offering $30 refunds to purchasers of the GTX 970 as compensation for discrepancies between expected and actual memory performance.

GeForce Partner Program

The Nvidia GeForce Partner Program was a marketing strategy that provided partners with advantages such as PR support, bundled video games, and marketing development funds. The initiative attracted criticism over concerns of anti-competitive behavior. Launched on March 1, 2018, it was discontinued by May 4, 2018 following negative feedback.

Hardware Unboxed

On December 10, 2020, Nvidia notified Steven Walton of the YouTube channel Hardware Unboxed that it would cease supplying GeForce Founders Edition graphics card review samples. Nvidia justified this decision by citing Walton’s emphasis on rasterization rather than ray tracing in his reviews, which they claimed did not align with their vision for gamers and the industry.

This action provoked backlash from technology reviewers and raised questions about editorial independence. Prominent figures such as Linus Sebastian from Linus Tech Tips and the Gamers Nexus review team publicly criticized Nvidia’s decision as harmful to impartial product assessment.

Two days later, Nvidia reversed its position and issued an apology. Hardware Unboxed confirmed the retraction and shared an additional apology from Nvidia’s Senior PR Manager Bryan Del Rizzo, who acknowledged that withholding samples due to disagreement with review content was inappropriate. This episode underscored the difficulties in maintaining unbiased evaluations amid corporate influence within the tech sector.

Inadequate disclosures concerning cryptomining

During 2018, Nvidia's processors became highly sought after for use in cryptomining, a method that rewards the validation of blockchain transactions. The U.S. Securities and Exchange Commission (SEC) asserted that Nvidia did not adequately reveal that cryptomining represented a "significant factor" contributing to the revenue increase from its gaming chip sales. This lack of disclosure resulted in investors being misinformed about how cryptomining influenced Nvidia’s financial outcomes. Without admitting or denying the allegations, Nvidia consented to pay $5.5 million in May 2022 to resolve SEC civil charges.

Investigation by the French Competition Authority

On September 26, 2023, the French Competition Authority conducted a search of Nvidia’s offices in France as part of an inquiry into alleged anti-competitive conduct within the graphics card sector. Nvidia has yet to issue any official statement regarding this investigation.

Planned Shanghai manufacturing site

In May 2025, U.S. Senators Jim Banks and Elizabeth Warren expressed concerns over Nvidia’s proposed facility in Shanghai, highlighting issues related to national security and economic implications that warrant comprehensive evaluation.

External links

Nvidia Developer website