Nvidia Corporation · NASDAQ:NVDA
Nvidia — GTC 2026
76 slides from Nvidia's conference presentation, Mar 16, 2026
Slide 1: NVIDIA
GI NVIDIA
Slide 2: Slide 2
Slide 3: NVIDIA
NVIDIA
Slide 4: Slide 4
Slide 5: Customer and partner logos
ABB above —accenture. = Adobe. autwunntux eni2— nivazes — Gamat = Jmrar —_ CDAbaba loud rimerk WS @ AME ANTHROMNC = rateck SUS (Armin axiom Foasoten sbioty DPm BLUE ORIGIN >BNY @ BoscH o CB —IByrovderce cadence Caltech en ae CATERPILAR ae ELOUDIAN Sconere #8 Cognition compa, tinwiunion «Cooly CM Cor
Slide 6: Breakthroughs
BREAKTHROUGHS ( 20 YEARS \ ECOSYSTEMS CUDA DEVELOPERS \ FLYWH _/ INSTALLED BASE
Slide 7: GeForce GTX 1080
GeForce GTX 1080
Slide 8: Slide 8
Slide 9: Announcing DLSS 5 3D-Guided Neural Rendering
Announcing DLSS 5 3D-Guided Neural Rendering
Slide 10: Structured Data is the Ground Truth of AI
Structured Data is the Ground Truth of AI $120B Structured Data Ecosystem – The Ground Truth of Enterprises Source: Gartner DBMS, 2025
Slide 11: Unstructured Data is the Context of AI
Unstructured Data is the Context of AI 100’s of Zettabytes Per Year of Unstructured Data – Growing Exponentially Source: IDC Global DataSphere, 2025
Slide 12: IBM and NVIDIA with Nestlé data refresh
“Working with IBM and NVIDIA, early results show we can refresh global operations data in minutes at reduced cost—making a capability that can be turned into tangible business impact in areas such as manufacturing or warehousing.” ― Chris Wright, Chief Information and Digital Officer of Nestlé Order
Slide 13: Slide 13
Slide 14: Dell AI Data Platform with NVIDIA, NTT DATA
“Dell AI Data Platform with NVIDIA helps us build scalable and repeatable data pipelines that drive automation—processing massive data volumes in minutes instead of hours and delivering transformative value for our clients.” ― Abhijit Dubey, CEO of NTT DATA Enterprise Data Processing Dell AI Data Pl
Slide 15: NVIDIA and Google Cloud with Snap
“Our collaboration with NVIDIA and Google Cloud helps us innovate faster for more than a billion Snapchatters worldwide. By lowering costs and scaling experiments across petabytes of data, we’re delivering AI-powered experiences more quickly and efficiently.” ― Saral Jain, CIO of Snap A/B Experiment
Slide 16: System Builders and OEMs: Google Cloud
AI Optimized AI Data Science Open Software, Infrastructure Platforms and Analytics Frameworks Liquid Storage Networking Google Vertex AI Google Cloud Run Google BigQuery Dataproc Dataflow Bigtable JAX XLA vLLM Cooling Distributed Kubernetes Compute Cloud Engine Engine System Builders and OEMs NVIDIA
Slide 17: System Builders and OEMs: AWS
Amazon Amazon AWS AWS Amazon Amazon Amazon Amazon Amazon AWS Braket EKS Nitro Enclaves Batch EMR Bedrock SageMaker AI EC2 EFA ParallelCluster System Builders and OEMs NVIDIA AI Platform
Slide 18: System Builders and OEMs: Microsoft
Microsoft Windows Microsoft Azure Azure Foundry Microsoft 365 Microsoft Foundry 365 Fabric Container Apps Kubernetes Service Local Copilot Bing System Builders and OEMs NVIDIA AI Platform
Slide 19: System Builders and OEMs: Oracle
OCI Oracle Oracle Autonomous Oracle AI OCI AI OCI OCI OCI Superclusters AI Database AI Lakehouse Data Platform Accelerator Packs Generative AI Data Science OKE System Builders and OEMs NVIDIA AI Platform
Slide 20: System Builders and OEMs: CoreWeave
W&B W&B W&B W&B CoreWeave CoreWeave CoreWeave CoreWeave CoreWeave CoreWeave Models Training Inference Weave Inference Slurm on Kubernetes Compute Storage Mission Control Kubernetes Service System Builders and OEMs NVIDIA AI Platform
Slide 21: Dell
Rubix Ontology Pipeline Builder Fusion Agent Studio Model Studio AIP Logic Model Catalog Workshop AI FDE Dell NVIDIA AI Platform
Slide 22: Industries
AUTOMOTIVE NVIDIA GTC FINANCIAL SERVICES HEALTHCARE AND. LIFE SCIENCES INDUSTRIAL MEDIA AND. ENTERTAINMENT QUANTUM = RETAIL AND CPG ROBOTICS TELCO
Slide 23: Slide 23
Slide 24: AI Natives
AI Natives AI for Auto AI for Customer Support AI for Engineering AI for Healthcare AI for Robotics AI for Search AI for Software Development DL Frameworks Inference Frameworks Agent Frameworks/Protocols Frontier Model Builders Model to Production NVIDIA AI Platform
Slide 25: Inference Inflection Arrives
Inference Inflection Arrives 10,000X ChatGPT Compute
Slide 26: Inference Inflection Drives Strong Growth
Inference Inflection Drives Strong Growth NVIDIA Full-Stack Expanding AI to All Regions and Industries
Slide 27: NVIDIA Extreme Co-Design Revolutionized Token Cost
NVIDIA Extreme Co-Design Revolutionized Token Cost “GB NVL72 Inference King” Tokens per Watt Drives Factory Revenue Performance Drives Token Cost 9,000,000 $4.50 8,000,000 $4.00 H200 NVL8 7,000,000 $3.50 Competition 6,000,000 GB300 NVL72 $3.00 35X Lower 5,000,000 50X $2.50 Cost 4,000,000 Higher $2.0
Slide 28: NVIDIA is the Global Standard for AI Inference at Scale
NVIDIA is the Global Standard for AI Inference at Scale All Leading Inference End-Points Run on NVIDIA Kimi K2.5 Reasoning
Slide 29: AI Factories are the Industrial Infrastructure of the AI Era
AI Factories are the Industrial Infrastructure of the AI Era Inference is the Workload Tokens are the New Commodity Compute is Revenue
Slide 30: Vera Rubin POD
Vera Rubin POD ALFLOPS 60,000 PF Memory Sandwidth 92 T8/s All-to-All Bandwidth 260 TB/s Scale-Out Radix 131,072 a
Slide 31: Inference Performance and Efficiency Drive Company Results
Inference Performance and Efficiency Drive Company Results
Slide 32: Inference Performance and Efficiency Drive Company Results
1.8 M 1.6 M 1.4 M Inference Performance 1.2 M and Efficiency Drive Throughput Company Results TPS/MW 1.0 M 0.8 M 0.6 M 0.4 M 0.2 M 0.0 M 50 100 200 400 600 800 1,000+ Interactivity TPS/User Free Medium High Premium Ultra Qwen 3 Kimi K2.5 GPT MoE GPT MoE GPT MoE 235 Billion 1 Trillion 2 Trillion 2 Tr
Slide 33: Inference Performance and Efficiency Drive Company Results
1.8 M 1.6 M 1.4 M Inference Performance 1.2 M and Efficiency Drive Throughput Company Results TPS/MW 1.0 M 0.8 M 0.6 M 0.4 M 0.2 M Hopper 0.0 M 50 100 200 Interactivity TPS/User Free Medium Qwen 3 Kimi K2.5 235 Billion 1 Trillion 32K 128K $0 $3
Slide 34: Inference Performance and Efficiency Drive Company Results
1.8 M 1.6 M 1.4 M Inference Performance 1.2 M and Efficiency Drive Throughput Company Results TPS/MW 1.0 M Blackwell 0.8 M NVL72 0.6 M 0.4 M 35X 0.2 M Hopper 0.0 M 50 200 Interactivity TPS/User Free Medium High Qwen 3 Kimi K2.5 GPT MoE 235 Billion 1 Trillion 2 Trillion 32K 128K 128K $0 $3 $6
Slide 35: Inference Performance and Efficiency Drive Company Results
1.8 M 1.6 M 1.4 M Inference Performance 1.2 M and Efficiency Drive Throughput Company Results TPS/MW 1.0 M Blackwell 0.8 M NVL72 0.6 M 0.4 M 35X 0.2 M Hopper 0.0 M 50 100 200 400 Interactivity TPS/User Free Medium High Premium Qwen 3 Kimi K2.5 GPT MoE GPT MoE 235 Billion 1 Trillion 2 Trillion 2 Tril
Slide 36: Inference Performance and Efficiency Drive Company Results
1.8 M 1.6 M Rubin NVL72 1.4 M Inference Performance 1.2 M and Efficiency Drive 2X Throughput 2X Company Results TPS/MW 1.0 M Blackwell 0.8 M NVL72 0.6 M 0.4 M 3X 0.2 M Hopper 10X 0.0 M 50 100 200 400 Interactivity TPS/User Free Medium High Premium Qwen 3 Kimi K2.5 GPT MoE GPT MoE 235 Billion 1 Trill
Slide 37: NVIDIA Vera Rubin Unlocks $150B Revenue Opportunity
NVIDIA Vera Rubin Unlocks $150B Revenue Opportunity Blackwell Rubin $150B $150B $100B 5X Annual Revenue Per Gigawatt $50B $30B $0B Free Medium High Premium Total Annual Qwen 3 Kimi K2.5 GPT MoE GPT MoE Revenue 235 Billion 1 Trillion 2 Trillion 2 Trillion 32K 128K 128K 400K $0 $3 $6 $45
Slide 38: Inference Performance and Efficiency Drive Company Results
1.8 M 1.6 M Rubin NVL72 1.4 M Inference Performance 1.2 M and Efficiency Drive 2X Throughput 2X Company Results TPS/MW 1.0 M Blackwell 0.8 M NVL72 0.6 M 0.4 M 3X 0.2 M Hopper 10X 0.0 M 50 100 200 400 Interactivity TPS/User Free Medium High Premium Qwen 3 Kimi K2.5 GPT MoE GPT MoE 235 Billion 1 Trill
Slide 39: Inference Performance and Efficiency Drive Company Results
1.8 M 1.6 M Rubin NVL72 1.4 M Inference Performance 1.2 M and Efficiency Drive 2X Throughput 2X Company Results TPS/MW 1.0 M Blackwell 0.8 M NVL72 0.6 M 0.4 M 3X 35X 0.2 M Hopper 10X 0.0 M 50 100 200 400 Interactivity TPS/User Free Medium High Premium Qwen 3 Kimi K2.5 GPT MoE GPT MoE 235 Billion 1 T
Slide 40: 35X Performance Efficiency Drives Company Results
1.8 M 1.6 M Rubin NVL72 1.4 M 35X Performance 1.2 M 2X Efficiency Drives Throughput 2X Company Results TPS/MW 1.0 M Blackwell 0.8 M NVL72 Rubin + LPX 0.6 M 3X 0.4 M Hopper 35X 35X 0.2 M 0.0 M 50 100 200 400 600 800 1,000+ Interactivity TPS/User Free Medium High Premium Ultra Qwen 3 Kimi K2.5 GPT MoE
Slide 41: NVIDIA Vera Rubin + LPX Unlocks $300B Revenue Opportunity
NVIDIA Vera Rubin + LPX Unlocks $300B Revenue Opportunity Blackwell Rubin VR+LPX $300B $300B $250B 10X $200B Annual Revenue Per Gigawatt $150B $150B $100B $50B $30B $0B Free Medium High Premium Ultra Total Annual Qwen 3 Kimi K2.5 GPT MoE GPT MoE GPT MoE Revenue 235 Billion 1 Trillion 2 Trillion 2 Tr
Slide 42: Uniting Processors of Extreme Performances
Rubin GPU Groq 3 LPU 288 GB HBM4 500 MB SRAM 22 TB/s 150 TB/s SRAM Bandwidth 50 PFLOPs (NVFP4) 1.2 PFLOPs (FP8) 336B Transistors 98B Transistors + 2.5T (HBM4) Uniting Processors of Extreme Performances
Slide 43: Uniting Processors of Extreme Performances
Rubin GPU Groq 3 LPU 288 GB HBM4 4 GB SRAM 22 TB/s 1,200 TB/s SRAM Bandwidth 55X 50 PFLOPs (NVFP4) 9.6 PFLOPs (FP8) 336B Transistors 784B Transistors + 2.5T (HBM4) Uniting Processors of Extreme Performances
Slide 44: Uniting Processors of Extreme FLOPS and Bandwidth
NVIDIA Dynamo Vera Rubin NVL72 Groq 3 LPX Prefill Input 1st Remaining Token Tokens KV Cache Decode Uniting Processors of Extreme FLOPS and Bandwidth
Slide 45: NVIDIA Groq 3 LPX
NVIDIA Groq 3 LPX AI Inference Compute 315 PFLOPS SRAM Capacity 128 GB Memory Bandwidth 40 PB/s Scale-Up Density 256 Chips Scale-Up Bandwidth 640 TB/s LPU C2C Spine Connectors 8 LPUs FPGA Host CPU BF4 LPU C2C Links Available 2H26
Slide 46: Announcing NVIDIA Vera Rubin NVL72 Launch Partners
Announcing NVIDIA Vera Rubin NVL72 Launch Partners AI Labs Cloud OEMs and ODMs 10X Perf/W | 3.6 EF NVFP4 | 1.6 PB/s HBM4 | 260 TB/s NVLink6
Slide 47: Announcing NVIDIA Vera CPU Launch Partners
Announcing NVIDIA Vera CPU Launch Partners Cloud OEMs and ODMs 256 Vera CPUs | 300 TB/s LPDDR5X | ETL Spine | 6.5X Throughput
Slide 48: Announcing NVIDIA BlueField-4 STX Launch Partners
Announcing NVIDIA BlueField-4 STX Launch Partners Cloud OEMs 5x Tokens/sec | 50 Tb/s Networking BW | 16TB Shared Context / GPU
Slide 49: NVIDIA Vera Rubin Chips – Rack Systems
NVIDIA Vera Rubin 7 Chips – 5 Rack Systems AI Factory for the Agentic AI Frontier 1 GW AI Factory X86 + Hopper Vera Rubin # of GPUs 600K 300K AI FLOPS 1.2 ZFLOPS 16 ZFLOPS All-to-All Scale-up 7.2 TB/s 260 TB/s Memory BW-per-Domain 2 EB/s 100 EB/s (GROQ SRAM) Tokens per Second 2M 700M Rubin GPU Vera
Slide 50: NVIDIA Extreme Co-Design Delivering X-Factors Every Year
NVIDIA Extreme Co-Design Delivering X-Factors Every Year From Chips to Racks to AI Factories Feynman Rubin Feynman Blackwell Die Stacking Custom HBM Rubin Rubin Ultra HBM4 HBM4e Blackwell Blackwell Ultra HBM3e HBM3e LP40 LP30 LP35 NVLink NVFP4 Grace CPU Vera CPU Rosa CPU BlueField-3 BlueField-4 NVLi
Slide 51: NVIDIA DSX AI Factory Platform
NVIDIA DSX AI Factory Platform Extreme Co-Design at Infrastructure Scale Rubin DSX DSX Omniverse Reference Designs Ecosystem DSX Blueprint Reference Designs, Best Practices, Methodologies DSX Max-Q DSX Flex DSX Exchange DSX Sim DSX Libraries, APIs, Software Power Liquid Cooling Chips, Systems, Facil
Slide 52: Slide 52
Slide 53: NVIDIA DSX AI Factory Platform
NVIDIA DSX AI Factory Platform Accelerates Scalable, Energy-Efficient AI Factory Deployment DSX Flex DSX Max-Q DSX Sim DSX Exchange
Slide 54: Announcing NVIDIA Space-1 Vera Rubin Module
Announcing NVIDIA Space-1 Vera Rubin Module
Slide 55: Inference Inflection Arrives
Inference Inflection Arrives
Slide 56: Announcing NVIDIA NemoClaw for OpenClaw
Announcing NVIDIA NemoClaw for OpenClaw
Slide 57: NemoClaw running on OpenClaw demo
©@@ ubuntu@brev-aeGah8vzm: ~ NIM running LASLae mMy-GSSLStant CLOCALMOST. SWWU) Run: nemoclaw my-assistant connect Status: nemoclaw my-assistant status Logs: nemoclaw my-assistant logs --follow % OpenClaw 2026.3.13 (61d171a) I'm not saying your workflow is chaotic... I'm just bringing a linter and a
Slide 58: Agents – A New Computing Platform
Agents – A New Computing Platform
Slide 59: Enterprise IT Renaissance from SaaS to Agent-as-a-Service
Enterprise IT Renaissance from SaaS to Agent-as-a-Service
Slide 60: Enterprise IT Renaissance from SaaS to Agent-as-a-Service
Enterprise IT Renaissance from SaaS to Agent-as-a-Service
Slide 61: Announcing NVIDIA NemoClaw Reference OpenClaw
Announcing NVIDIA NemoClaw Reference OpenClaw NVIDIA Agent Toolkit for Building Specialized Agents
Slide 62: NVIDIA Leaderboard Topping Open Models
NVIDIA Leaderboard Topping Open Models BioNeMo Earth-2 GR00T Alpamayo Biology AI AI Physics Robotics Autonomous Vehicles Nemotron Cosmos Agentic AI Physical AI
Slide 63: Earth-2 Medium Range “E”
Earth-2 Medium Range “E”
Slide 64: Nemotron and open model benchmarks
AA Intelligence vs Efficiency PinchBench DeepResearch Bench II ViDoRe, MMTEB, BRIGHT Nemotron 3 Super Nemotron 3 Super Nemotron 3 Super (AI-Q) Nemotron RAG OCRBench v2 AA Conversational Dynamics Physical AI Bench Physical Reasoning from Video NemotronVL vs Reasoning Cosmos Predict Cosmos Reason Nemo
Slide 65: NVIDIA Nemotron 3 Super – Best Open Model for OpenClaw
NVIDIA Nemotron 3 Super – Best Open Model for OpenClaw
Slide 66: NVIDIA Nemotron 3 Ultra – Best Open Base Model
NVIDIA Nemotron 3 Ultra – Best Open Base Model 5X Efficiency and Highest Reasoning Accuracy on NVIDIA GB200 NVL72 GLM Kimi K2 Nemotron 3 Ultra 5X 90% 4X 80% 3X 2X 70% 1X X 60% Peak Understanding Code Math Multilingual Throughput MMLU Pro HumanEval, MBPP GSM8K, Minerva Math Global MMLU, MGSM
Slide 67: The World Building Regional AI With NVIDIA Nemotron
The World Building Regional AI With NVIDIA Nemotron Bielik.ai Linagora NAVER Trillion Labs SOOFI TII United States EuroLLM Viettel Institute of Science Dicta Tokyo Stockmark HUMAIN BharatGen Gnani.ai Sarvam AI AI Singapore WideLabs Indosat
Slide 68: Global AI Leaders Join NVIDIA Nemotron Coalition
Announcing Global AI Leaders Join NVIDIA Nemotron Coalition to Advance Open Frontier Models
Slide 69: o vtfnctfne
o vtfnctfne r A ATLASSIAN ——\ ,-——— cadence , cisco fe pe | x ‘a DS Sarees 7 -— ———. -— synapsys* y x / a /) a QAlibaba LX fei) Foaseter hil ByteDance Q@cvounian crewal 8 Cognition OW CoreWeave DéAL fechroicgies Oddn ‘O€verpure a ractory Google Cloud Occore leon Honeywell harbor $e Harmonic HPE @B w
Slide 70: NEXT AI
© CodeRabbit ——. _——. COHESITY ———-._ -———. 8 Edison & ;— ABRIDGE ——— _ 4 MIM Security NEXTAL n Schneider Rockwel Automation ) la Besseie perplexity enterprise ==h6 N y . Hitiaea eee QAlibaba LNs fei) Foaseter hil ByteDance Q@cvounian crewal 8 Cognition OW CoreWeave DéAL fechroicgies Oddn ‘O€verpure
Slide 71: Announcing NVIDIA NemoClaw Reference OpenClaw
Announcing NVIDIA NemoClaw Reference OpenClaw NVIDIA Agent Toolkit for Building Specialized Agents
Slide 72: Slide 72
Slide 73: Slide 73
Slide 74: Slide 74
Slide 75: Slide 75
Slide 76: NVIDIA
GI NVIDIA