led by Jensen Huang
NVIDIA profile, products and governance
The GPU and AI infrastructure company behind the Blackwell and Vera Rubin platforms, CUDA and the open Nemotron models.
Hokai's read on NVIDIA
NVIDIA, the Santa Clara chipmaker, sells the Blackwell and Vera Rubin data center platforms, CUDA and the open Nemotron models. Vera Rubin entered full production in 2026, and Google Cloud announced general availability on October 21, 2026, during GTC Berlin week.
NVIDIA products
Models from NVIDIA
| Model | Lab | Context | In / out per 1M | Reads and writes |
|---|---|---|---|---|
| Nemotron Labs Dory 1.0 4BNVIDIA's open 4B looped model: one checkpoint reuses its shared layers 1, 2 or 4 times per token, with a 256K-token context window and a permissive licence. | NVIDIA | 262K | -- | text, code → text |
| Nemotron 3.5 Bolt 74B-A9B 24GBNVIDIA's open 74B-A9B vision-language MoE, quantized to 2 bits to fit 24GB, built for local agents. Released in October 2026 with 256K context. | NVIDIA | 262K | -- | text, image, video → text |
| Nemotron 3.5 Super VL 120B-A12BNVIDIA's open mixture-of-experts vision-language model for document, screenshot and video agents, released in October 2026. | NVIDIA | 1M | $0.10 / $0.40 | text, image, video → text |
| Nemotron 3.5 LightningNVIDIA's open MoE model for AI agents: 3B of 30B parameters active per token, a 1M-token window and quicker agent tasks than a same-size Qwen rival on PinchBench. | NVIDIA | 1M | $0.025 / $0.12 | text, code → text |
| Nemotron 3.5 ASRNVIDIA's open streaming speech recognizer, released 4 June 2026, transcribing dozens of language-locales from a single checkpoint. | NVIDIA | -- | -- | audio → text |
| Nemotron 3 Nano 30B A3BNVIDIA's open Nemotron 3 Nano (30B total, 3.6B active) hits 78.3% MMLU-Pro with 3.3x the throughput of Qwen3-30B-A3B, released December 2025. | NVIDIA | 1M | $0.05 / $0.20 | text → text |
Skills from NVIDIA
| Skill | Works with | Connects by | Checked |
|---|---|---|---|
| NVIDIA BioNeMo Agent ToolkitAgent-callable skills for protein folding, docking, and genomics on NVIDIA NIM | Claude · Codex | Skill file | 23 Oct |
Key products, as the company lists them
| Product | What it is |
|---|---|
| CUDA | GPU compute platform |
| Blackwell / Vera Rubin platforms | AI data center hardware |
| Nemotron | Open reasoning and agentic AI models |
| NIM Microservices | Managed inference API |
| Omniverse / Cosmos | Robotics and world model simulation platform |
| DGX Systems | AI supercomputer hardware |
NVIDIA facts
- Company type
- Bigtech
- Founded
- 1993
- Headquarters
- Santa Clara, CA, USA
- Led by
- Jensen Huang
- Team size
- 40000-45000
- Stock ticker
- NASDAQ:NVDA
- Trust centre
- Published
- Privacy policy
- Published
- Terms of service
- Published
- Funding
- Public company (NASDAQ: NVDA) since January 1999; pre-IPO raised $20M from Sequoia Capital and Sutter Hill Ventures. No private funding rounds since IPO.
- Valuation
- Publicly traded (NASDAQ: NVDA); market capitalization changes daily
Compliance certifications
- SOC 2
- ISO 27001
- ISO 27017
- ISO 27018
- ISO 27701
- ISO 9001
- ISO 14001
- ISO 21434
- ISO 26262
- ISO 50001
- TISAX
- CAIQ
Data regions
- us-east
- us-west
- eu-central
- ap-northeast
- ap-southeast
NVIDIA releases
Every model Hokai holds for this company that carries a release date, newest first.
- Available Nemotron Labs Dory 1.0 4B released
- Available Nemotron 3.5 Bolt 74B-A9B 24GB released
- Available Nemotron 3.5 Super VL 120B-A12B released
- Available Nemotron 3.5 Lightning released
- Available Nemotron 3.5 ASR released
- Available Nemotron 3 Nano 30B A3B released
Not sure which product fits?
Questions about NVIDIA
What does NVIDIA actually sell to AI teams?
It sells in three layers, starting with hardware: Blackwell systems today and Vera Rubin racks now ramping, plus DGX systems and the RTX Spark superchip for desktops. Software: CUDA, CUDA-X libraries, NIM microservices and the Agent Toolkit. Models: open Nemotron, Cosmos world models and Isaac GR00T. Most buyers reach it through a cloud provider that already owns the hardware.
Who runs NVIDIA and when was it founded?
Jensen Huang, Chris Malachowsky and Curtis Priem founded it on April 5, 1993. Huang has been CEO continuously since then and Colette Kress has been CFO since 2013. The company has been listed on NASDAQ since January 1999 and is headquartered in Santa Clara, California. Leadership has not changed hands since founding, so strategy on CUDA, networking and open models has been unusually consistent.
How is NVIDIA financed and how fast is it growing?
It is entirely public-market and cash-flow financed; the only outside money before the 1999 listing was a $20 million round from Sequoia Capital and Sutter Hill Ventures. In its latest report, for the quarter ended July 26, 2026, quarterly revenue more than doubled year over year, with Data Center the dominant segment. Market capitalization moves daily, so check a live quote rather than a stored figure.
Which security and compliance certifications does NVIDIA list?
Its AI Trust Center lists twelve attestations and standards, among them SOC 2 for security controls, the ISO 27000-series privacy and cloud security family, two automotive standards (ISO 21434 and ISO 26262), TISAX and the CSA CAIQ. For US government work it says NVIDIA AI Enterprise Government Ready software is designed to meet the applicable software development controls for FedRAMP High and FISMA High; that is a design alignment, not a blanket certificate for every product. HIPAA eligibility is not stated on that page, so confirm per product with NVIDIA.
Where does NVIDIA have real competition?
On silicon, AMD's Instinct accelerators and the custom chips built by its own biggest customers: Google's TPUs, Amazon's Trainium and Microsoft's Maia. On models, Nemotron competes with other open-weight families such as Qwen and Llama. NVIDIA's defence is the integrated stack of chips, NVLink interconnect and CUDA software, which competitors assemble from separate parts. It has no closed consumer chatbot to compete with the frontier labs, by design.
About NVIDIA
Santa Clara-based NVIDIA reported $96.2 billion of revenue for the quarter ended in July 2026, of which $89.0 billion came from Data Center, at a 75.0% gross margin. Listed on NASDAQ as NVDA, it is the leading supplier of AI training and inference hardware.
In their words
To bring superhuman capabilities to every human, in every industry.
NVIDIA was founded on April 5, 1993 by Jensen Huang, Chris Malachowsky and Curtis Priem, who had worked at LSI Logic, Sun Microsystems and AMD. It started with $40,000 of founder capital plus a $20 million venture round from Sequoia Capital and Sutter Hill Ventures, and it is headquartered in Santa Clara, California. The bet was that graphics processing would become its own programmable computing category, and that bet took about two decades to pay off in AI. The company listed on NASDAQ (NVDA) on January 22, 1999 at $12 a share and has raised no private funding since.
Read the full background
Today the business spans AI data center hardware, networking, software and models. The Blackwell generation has been shipping since 2024, and the Vera Rubin platform, a six-chip system built around the Vera CPU, Rubin GPU, NVLink 6 switch, ConnectX-9 SuperNIC, BlueField-4 DPU and Spectrum-6 Ethernet switch, was described in the August 26, 2026 earnings release as ramping into full production with racks running at CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius. Google Cloud announced general availability of Vera Rubin infrastructure on October 21, 2026, during GTC Berlin. Around the hardware sit the CUDA software platform, NIM microservices, the DGX and RTX Spark product lines, the Omniverse and Cosmos platforms for physical AI, Isaac GR00T for humanoid robots, Alpamayo for autonomous driving, and domain toolkits such as the BioNeMo Agent Toolkit for life sciences.
NVIDIA also publishes its own open models under the Nemotron name, with open weights, training data and recipes. The line now includes Nemotron 3 Nano, a hybrid Mamba-Transformer mixture-of-experts model, Nemotron 3.5 Lightning, Nemotron 3.5 Super VL and Nemotron 3.5 Bolt, plus Nemotron 3 Ultra and separate retrieval, speech and safety families. The developer page lists a 1M-token context for the Nemotron 3 family and more than 12T tokens of open multilingual reasoning, coding and safety data.
The money comes overwhelmingly from selling to cloud providers and AI labs. For the quarter ended July 26, 2026, NVIDIA reported revenue of $96.2 billion, with Data Center at $89.0 billion and a 75.0% gross margin, and guided the following quarter to $108.0 billion without assuming any Data Center compute revenue from China. The same release describes financing partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR intended to mobilize over $500 billion of third-party capital for AI infrastructure, subject to definitive agreements. Jensen Huang remains founder, president and CEO, and Colette Kress is chief financial officer.
The case against treating NVIDIA as the default is concentration and cost. Its largest customers, including Google, Microsoft and Amazon, also build their own accelerators, and a buyer whose workload is a single well-understood inference service can sometimes run it cheaper on custom silicon or on AMD's Instinct line. CUDA's lead is real but it is also the lock-in a cost-sensitive team should price in. Export controls add a second risk: the company's own outlook excludes China Data Center compute. Teams that only need to call a model through an API, for example from Together AI or an OpenAI endpoint, may never touch NVIDIA's stack directly. Rented capacity such as CoreWeave is the usual way smaller teams reach the hardware.
Sources and guides
Official sources
- Company website Official
- Documentation Docs
- Engineering blog Blog
- Trust centre Trust
- Privacy policy Legal
- Terms of service Legal
- Source repositories GitHub
- Careers Hiring
- LinkedIn Social
- Twitter / X Social
Everything on this page (funding, leadership, compliance posture, the product list) is read from NVIDIA's own pages and filings, or from the Hokai records for the products it ships. A field the company does not publish is left off the page rather than guessed at.
Hokai guides that mention NVIDIA
Lists that include NVIDIA
More AI Companies on Hokai
BaiduBeijing AI group behind the ERNIE models and Apollo Go robotaxis
MeituanBeijing local-services platform behind the LongCat AI models and Keeta
ByteDanceChinese consumer-internet giant behind TikTok, CapCut and the Seed AI models.GoogleAlphabet's AI-first subsidiary powering Gemini, Google Cloud, and AI Overviews for billions.
MicrosoftEmpowering organizations and individuals through cloud, AI, and productivity tools.
AmazonThe cloud giant behind Amazon Bedrock, the Nova models, Nova Act agents and Trainium chips.