Artificial Intelligence · Profile from AI Research Companion
Nvidia
The Santa Clara chip designer that began in 1993 making 3D graphics cards for PC games, bet from 2006 that its graphics processors could become general-purpose parallel computers (CUDA), was in the right place when deep learning took off in 2012, and became the main supplier of the computing power behind the generative-AI boom. It sells GPUs, but increasingly whole systems: chips, networking, racks and software. Revenue went from ~$27bn in fiscal 2023 (year to January 2023) to ~$215.9bn in fiscal 2026 (year to January 2026), ~90% of it from data centres (company). It became the first company worth $4tn (July 2025) and $5tn (October 2025), and was valued at ~$5.7–5.8tn in early October 2026 (market data, reported). Its biggest setbacks were political as well as technical: the blocked $40bn purchase of Arm (2022) and four years of US export controls and Chinese counter-measures that took its China data-centre business from most of the market to close to nothing.
At a glance
- Founded
- April 1993, by Jensen Huang, Chris Malachowsky and Curtis Priem (the founding meeting is usually placed at a Denny's restaurant in San Jose)
- Headquarters
- Santa Clara, California
- Ownership / structure
- Public (Nasdaq: NVDA) since January 1999. Widely held; index funds (Vanguard, BlackRock, State Street) are the largest holders; Jensen Huang owns ~3.5% (reported, verify current figure) and has sold shares under pre-arranged 10b5-1 plans
- Leadership (2026)
- Jensen Huang (co-founder, president and CEO since 1993); Colette Kress (CFO since 2013); Jay Puri (EVP, worldwide field operations); Debora Shoquist (EVP, operations); Bill Dally (chief scientist); Ian Buck (VP, hyperscale and HPC; creator of CUDA's precursor)
- Products
- Data-centre GPUs (A100, H100, H200, Blackwell B200/B300, GB200/GB300 "superchips", Rubin from 2026), Grace and Vera Arm-based CPUs, NVLink and NVLink Switch, InfiniBand (Quantum) and Ethernet (Spectrum-X) networking, BlueField DPUs, DGX systems and DGX Cloud, CUDA and libraries (cuDNN, TensorRT, NeMo, NIM microservices), Omniverse, GeForce gaming GPUs, professional visualisation (RTX), automotive (DRIVE), robotics (Jetson, Isaac)
- Scale
- Fiscal 2026 (year to January 25, 2026): revenue ~$215.9bn (+65%), data-centre revenue ~$193.7bn (+68%); Q2 fiscal 2027 (to late July 2026): revenue ~$96.2bn, data centre ~$89.0bn; Q3 fiscal 2027 guidance ~$108bn (company). ~36,000 employees in January 2025 (company; 2026 figure verify)
- Market value
- ~$1tn (May 30, 2023); ~$2tn (March 2024); ~$3tn (June 2024); first company at ~$4tn (July 9, 2025 intraday; first close July 10) and ~$5tn (October 29, 2025); ~$5.7–5.8tn at record closes in early October 2026 (market data, reported)
- Signature moves (incl. failures)
- CUDA (2006); the AlexNet moment and the deep-learning bet (2012–14); DGX and the data-centre pivot (2016); Mellanox (2019–20); the blocked Arm deal (2020–22); China export controls from A800 to H200 (2022–26); the H100 shortage and Blackwell supply (2023–25); the $4tn and $5tn market caps (2025); the Intel stake, OpenAI deals and "circular financing" debate (2025–26)
9 strategic moves analysed
Each one gets a full strategy file in the companion: context, the decision, the financing, what happened, and the lesson.
2006
CUDA and General Purpose GPU Computing
Launch (software platform and chip architecture)
2012
AlexNet and the Deep Learning Bet
Pivot (strategic refocus on deep learning)
2016
DGX and the Data Centre Pivot
Launch / pivot (from components to systems; from gaming to data centres)
2019
Mellanox Acquisition
Acquisition
2020
Arm Deal Blocked
Acquisition (abandoned under regulatory opposition)
2022
China Export Controls A800 to H200
Regulatory action (and Nvidia's product responses to it)
2023
H100 Shortage and Blackwell Supply
Launch / supply (product ramp under capacity constraints)
2025
Four and Five Trillion Market Cap
Market milestone (valuation and capital-markets events)
2025
Intel Stake OpenAI Deals and Circular Financing
Investment / compute deal / licence-and-hire (a programme of strategic stakes in customers and partners)