NVIDIA — research brief

Researched: 2026-09-20 (CT)
Primary focus: Full-stack accelerated computing platform — CUDA software moat → Data Center GPUs (Blackwell / Vera Rubin) → Networking (NVLink, Spectrum, InfiniBand) → NVIDIA AI Enterprise / NIM / NeMo → DGX Cloud & partner AI factories.

Labels: FACT = NVIDIA IR / newsroom / docs / nvidia.com; SECONDARY = press wire or third-party; INFERENCE = synthesis from facts.

Company & mission

FACT — NVIDIA Corporation (NASDAQ: NVDA) positions itself as the leader in accelerated computing / AI infrastructure, selling GPUs, systems, networking, and enterprise AI software. Source: Q2 FY2027 IR press release, About / product pages.
FACT — 10-K / company narrative (prior filings): Compute & Networking includes Data Center platforms, networking, automotive, Jetson/embedded, and DGX Cloud; paid software includes NVIDIA AI Enterprise and vGPU; DGX Cloud is a fully managed AI-training-as-a-service platform. Source: FY2025 Form 10-K (SEC).

Products (what they sell)

FACT — CUDA and CUDA-X libraries are the developer software foundation that makes NVIDIA GPUs programmable for AI/HPC; free via NVIDIA Developer Program (separate from paid AI Enterprise). Source: Software licensing FAQ, AI Enterprise product.
FACT — Data Center GPUs / rack-scale systems: Blackwell platform (GB200/GB300 NVL72 era), transitioning to Vera Rubin rack-scale (full production at partners including CoreWeave, Google Cloud, Microsoft Azure, OCI, Nebius as of Q2 FY2027 PR). Source: Q2 FY2027 IR.
FACT — Networking: NVLink / NVLink Switch fabric, InfiniBand, Spectrum Ethernet for AI; Q2 FY2026 CFO commentary broke Data Center into Compute $33.8B and Networking $7.3B. Source: Q2 FY2026 newsroom, CFO commentary summaries.
FACT — NVIDIA AI Enterprise: production software suite (NIM microservices, NeMo, Blueprints, orchestration / Run:ai, drivers, frameworks) with enterprise support. List price $4,500 / GPU / year (1-year subscription, Business Standard support); CSP marketplace ~$1 / hour / GPU + instance cost; free Developer Program NIM for research/dev on up to 16 GPUs; 90-day evaluation license for production trial. Source: Licensing Guide — Pricing, NIM product FAQ, AI Enterprise.
FACT — DGX Cloud: NVIDIA’s AI factory in the cloud / managed training platform offered with AWS, Google Cloud, Microsoft Azure, OCI (private-offer / marketplace pricing). Source: DGX Cloud, 10-K.
FACT — Adjacent lines (smaller than Data Center but material): Gaming (GeForce), Professional Visualization (RTX PRO), Automotive / Robotics (DRIVE, Jetson) — Edge Computing called out at $7.2B in Q2 FY2027. Source: Q2 FY2027 IR.

Scale metrics (company disclosures — public company)

FACT — Q2 FY2027 (quarter ended July 26, 2026): Revenue $96.2B (+106% YoY, +18% QoQ); Data Center revenue $89.0B (+117% YoY); GAAP gross margin 75.0%; GAAP diluted EPS $2.46; returned ~$26.0B to shareholders in the quarter via buybacks + dividends. Q3 FY2027 revenue outlook $108.0B ±2% (no assumed China Data Center compute). Source: IR press release Aug 26, 2026.
FACT — Q2 FY2026 (ended July 27, 2025) for architecture of the DC mix: Revenue $46.7B; Data Center $41.1B; Compute $33.8B; Networking $7.3B; Gaming $4.3B; large cloud providers ~50% of Data Center revenue (CFO commentary). Source: Newsroom Q2 FY2026, SEC EX-99.1.

Partners & GTM

FACT — Manufacturing outsourced to foundry / packaging / assembly partners (company risk factor; TSMC ecosystem commonly cited — treat foundry dependence as FACT from 10-K risk language, specific partner shares as SECONDARY unless NVIDIA discloses). Source: 10-K risk factors.
FACT — Hyperscalers and AI clouds are primary Data Center buyers / hosts (Azure, OCI, AWS, Google; NVIDIA Cloud Partners such as CoreWeave, Nebius named in Q2 FY2027 Vera Rubin ramp). Source: Q2 FY2027 IR.
FACT — Systems OEMs / NVIDIA-Certified Systems and NPN channel sell servers and AI Enterprise licenses. Source: AI Enterprise / Certified Systems messaging, Licensing Guide.
INFERENCE — Dual GTM: developer PLG via free CUDA / Developer Program / build.nvidia.com NIM → enterprise hardware + AI Enterprise subscriptions via NPN / cloud marketplaces → hyperscaler / CSP multi-year GPU purchases + DGX Cloud / AI factory reference designs.

Metrics we are NOT inventing

Do not invent CUDA developer headcount, “installed base” GPU counts, or software ARR unless NVIDIA discloses them.
Do not invent AI Enterprise attach rates or DGX Cloud revenue breakouts (company does not publish a clean public software ARR in the citations above).
Do not invent China H20 / export volumes beyond stated “no assumed China DC compute” outlook language.
Gaming / Auto figures only when cited from the same IR release.

Teaching arc (for canvas)

CUDA / developer ecosystem lock-in → Data Center GPU platforms (Blackwell → Vera Rubin) → Networking fabric (NVLink / Spectrum / InfiniBand) → Software & platforms (AI Enterprise, NIM, NeMo, Omniverse) → DGX Cloud + CSP / OEM / SI partners deepen the stack.

# NVIDIA — research brief Researched: 2026-09-20 (CT) Primary focus: Full-stack accelerated computing platform — CUDA software moat → Data Center GPUs (Blackwell / Vera Rubin) → Networking (NVLink, Spectrum, InfiniBand) → NVIDIA AI Enterprise / NIM / NeMo → DGX Cloud & partner AI factories. Labels: **FACT** = NVIDIA IR / newsroom / docs / nvidia.com; **SECONDARY** = press wire or third-party; **INFERENCE** = synthesis from facts. --- ## Company & mission - **FACT** — NVIDIA Corporation (NASDAQ: NVDA) positions itself as the leader in accelerated computing / AI infrastructure, selling GPUs, systems, networking, and enterprise AI software. Source: [Q2 FY2027 IR press release](https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Second-Quarter-Fiscal-2027/default.aspx), [About / product pages](https://www.nvidia.com/en-us/data-center/). - **FACT** — 10-K / company narrative (prior filings): Compute & Networking includes Data Center platforms, networking, automotive, Jetson/embedded, and DGX Cloud; paid software includes NVIDIA AI Enterprise and vGPU; DGX Cloud is a fully managed AI-training-as-a-service platform. Source: [FY2025 Form 10-K (SEC)](https://www.sec.gov/Archives/edgar/data/1045810/000104581025000023/nvda-20250126.htm). ## Products (what they sell) - **FACT** — **CUDA** and CUDA-X libraries are the developer software foundation that makes NVIDIA GPUs programmable for AI/HPC; free via NVIDIA Developer Program (separate from paid AI Enterprise). Source: [Software licensing FAQ](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-software-licensing-overview-and-faq/), [AI Enterprise product](https://www.nvidia.com/en-us/data-center/products/ai-enterprise/). - **FACT** — **Data Center GPUs / rack-scale systems**: Blackwell platform (GB200/GB300 NVL72 era), transitioning to **Vera Rubin** rack-scale (full production at partners including CoreWeave, Google Cloud, Microsoft Azure, OCI, Nebius as of Q2 FY2027 PR). Source: [Q2 FY2027 IR](https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Second-Quarter-Fiscal-2027/default.aspx). - **FACT** — **Networking**: NVLink / NVLink Switch fabric, InfiniBand, Spectrum Ethernet for AI; Q2 FY2026 CFO commentary broke Data Center into **Compute $33.8B** and **Networking $7.3B**. Source: [Q2 FY2026 newsroom](https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-second-quarter-fiscal-2026), CFO commentary summaries. - **FACT** — **NVIDIA AI Enterprise**: production software suite (NIM microservices, NeMo, Blueprints, orchestration / Run:ai, drivers, frameworks) with enterprise support. List price **$4,500 / GPU / year** (1-year subscription, Business Standard support); CSP marketplace **~$1 / hour / GPU** + instance cost; free Developer Program NIM for research/dev on up to **16 GPUs**; **90-day** evaluation license for production trial. Source: [Licensing Guide — Pricing](https://docs.nvidia.com/ai-enterprise/planning-resource/licensing-guide/latest/pricing.html), [NIM product FAQ](https://docs.api.nvidia.com/nim/docs/product), [AI Enterprise](https://www.nvidia.com/en-us/data-center/products/ai-enterprise/). - **FACT** — **DGX Cloud**: NVIDIA’s AI factory in the cloud / managed training platform offered with AWS, Google Cloud, Microsoft Azure, OCI (private-offer / marketplace pricing). Source: [DGX Cloud](https://www.nvidia.com/en-us/data-center/dgx-cloud/), [10-K](https://www.sec.gov/Archives/edgar/data/1045810/000104581025000023/nvda-20250126.htm). - **FACT** — Adjacent lines (smaller than Data Center but material): Gaming (GeForce), Professional Visualization (RTX PRO), Automotive / Robotics (DRIVE, Jetson) — Edge Computing called out at **$7.2B** in Q2 FY2027. Source: [Q2 FY2027 IR](https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Second-Quarter-Fiscal-2027/default.aspx). ## Scale metrics (company disclosures — public company) - **FACT** — **Q2 FY2027** (quarter ended July 26, 2026): **Revenue $96.2B** (**+106%** YoY, **+18%** QoQ); **Data Center revenue $89.0B** (**+117%** YoY); GAAP gross margin **75.0%**; GAAP diluted EPS **$2.46**; returned ~**$26.0B** to shareholders in the quarter via buybacks + dividends. Q3 FY2027 revenue outlook **$108.0B ±2%** (no assumed China Data Center compute). Source: [IR press release Aug 26, 2026](https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Second-Quarter-Fiscal-2027/default.aspx). - **FACT** — **Q2 FY2026** (ended July 27, 2025) for architecture of the DC mix: Revenue **$46.7B**; Data Center **$41.1B**; Compute **$33.8B**; Networking **$7.3B**; Gaming **$4.3B**; large cloud providers ~**50%** of Data Center revenue (CFO commentary). Source: [Newsroom Q2 FY2026](https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-second-quarter-fiscal-2026), [SEC EX-99.1](https://www.sec.gov/Archives/edgar/data/1045810/000104581025000207/q2fy26pr.htm). ## Partners & GTM - **FACT** — Manufacturing outsourced to foundry / packaging / assembly partners (company risk factor; TSMC ecosystem commonly cited — treat foundry dependence as FACT from 10-K risk language, specific partner shares as SECONDARY unless NVIDIA discloses). Source: [10-K risk factors](https://www.sec.gov/Archives/edgar/data/1045810/000104581025000023/nvda-20250126.htm). - **FACT** — Hyperscalers and AI clouds are primary Data Center buyers / hosts (Azure, OCI, AWS, Google; NVIDIA Cloud Partners such as CoreWeave, Nebius named in Q2 FY2027 Vera Rubin ramp). Source: [Q2 FY2027 IR](https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Second-Quarter-Fiscal-2027/default.aspx). - **FACT** — Systems OEMs / NVIDIA-Certified Systems and NPN channel sell servers and AI Enterprise licenses. Source: [AI Enterprise / Certified Systems messaging](https://www.nvidia.com/en-us/data-center/products/ai-enterprise/), [Licensing Guide](https://docs.nvidia.com/ai-enterprise/planning-resource/licensing-guide/latest/licensing.html). - **INFERENCE** — Dual GTM: developer PLG via free CUDA / Developer Program / build.nvidia.com NIM → enterprise hardware + AI Enterprise subscriptions via NPN / cloud marketplaces → hyperscaler / CSP multi-year GPU purchases + DGX Cloud / AI factory reference designs. ## Metrics we are NOT inventing - Do **not** invent CUDA developer headcount, “installed base” GPU counts, or software ARR unless NVIDIA discloses them. - Do **not** invent AI Enterprise attach rates or DGX Cloud revenue breakouts (company does not publish a clean public software ARR in the citations above). - Do **not** invent China H20 / export volumes beyond stated “no assumed China DC compute” outlook language. - Gaming / Auto figures only when cited from the same IR release. ## Teaching arc (for canvas) **CUDA / developer ecosystem lock-in → Data Center GPU platforms (Blackwell → Vera Rubin) → Networking fabric (NVLink / Spectrum / InfiniBand) → Software & platforms (AI Enterprise, NIM, NeMo, Omniverse) → DGX Cloud + CSP / OEM / SI partners deepen the stack.**