investor update...this presentation contains historical revenue amounts for certain of our market...
TRANSCRIPT
September 2018
INVESTOR UPDATE
SAFE HARBOR
Forward-Looking Statements
Except for the historical information contained herein, certain matters in this presentation including, but not limited to, statements as to: the world’s demand for computing power
growing exponentially; NVIDIA’s ability to advance computing, including by optimizing across the entire stack, and the rate of advancing computing; NVIDIA’s growth into multiple
markets, including through the use of its unified architecture; the availability of NVIDIA GPUs and Jetson AGX Xavier DevKit; NVIDIA GPUs accelerating deep learning frameworks and
applications; the benefits, impact, performance and abilities of: NVIDIA’s computing platform, the Turing architecture, RT Core, ray tracing, Tensor Core, GeForce, GeForce RTX,
Quadro RTX, RTX Servers, NVIDIA’s GPUs and accelerated computing platform, DGX, inference performance, TensorRT, DGX-2, HGX-2, Xavier, AGX, DRIVE AGX, Jetson AGX, Jetson
AGX Xavier, Clara AGX, DRIVE, DRIVE AGX, DRIVE AGX DevKit, NVIDIA Isaac, Clara AGX Xavier, GPU Servers, and our products, technologies, services, platforms and programs; our
market and industry opportunities and TAM; DGX being the world’s most advanced AI platform; shorter training time equaling faster progress in advancing AI; NVIDIA’s ability to
master inference; every industry tapping into the power of AI; DGX-2 being the world’s most powerful AI system for data center deployments; the value of accelerated datacenter
performance; Xavier enabling automation in many large industries; NVIDIA’s full stack enabling market leaders to take advantage of our platform; the entities developing on DRIVE
and creating the future of transportation; the next chapter of AI being autonomous machines; NVIDIA tackling big opportunities in some of the largest markets; accelerated
computing being the way forward as Moore’s Law has ended; NVIDIA’s unified architecture driving growth and margin expansion across markets; our growth and growth drivers; our
strategies; market trends; future financial results, estimates and forecasts; how NVIDIA innovates; our production roadmap and schedules; and other predictions and estimates are
forward-looking statements within the meaning of the Private Securities Litigation Reform Act of 1995. These forward-looking statements and any other forward-looking statements
that go beyond historical facts that are made in this presentation are subject to risks and uncertainties that may cause actual results to differ materially. Important factors that
could cause actual results to differ materially include: global economic conditions; our reliance on third parties to manufacture, assemble, package and test our products; the
impact of technological development and competition; development of new products and technologies or enhancements to our existing products and technologies; market
acceptance of our products or our partners’ products; design, manufacturing or software defects; changes in consumer preferences and demands; changes in industry standards and
interfaces; unexpected loss of performance of our products or technologies when integrated into systems and other factors. For a complete discussion of factors that could
materially affect our financial results and operations, please refer to the reports we file from time to time with the SEC, including our annual report on Form 10-K and quarterly
reports on Form 10-Q. Copies of reports we file with the SEC are posted on our website and are available from NVIDIA without charge. These forward-looking statements are not
guarantees of future performance and speak only as of the date hereof, based on information currently available to us. Except as required by law, NVIDIA disclaims any obligation to
update these forward-looking statements to reflect future events or circumstances.
Financial Measures
This presentation contains historical revenue amounts for certain of our market platforms and businesses which provides investors with additional information to supplement the
segment reporting information contained in our Form 10-K for the fiscal year ended January 28, 2018. In addition to U.S. GAAP financials, this presentation includes certain non-
GAAP financial measures. These non-GAAP financial measures are in addition to, and not a substitute for or superior to, measures of financial performance prepared in accordance
with U.S. GAAP. See our website for a reconciliation between each non-GAAP measure and the most comparable GAAP measure. Where we present non-GAAP financial measures,
including non-GAAP gross margin, non-GAAP operating margin, and free cash flow we generally exclude stock-based compensation, product warranty charge, restructuring costs, and
other charges, where applicable. Free cash flow is calculated as GAAP net cash provided by operating activities less purchases of property and equipment and intangible assets.
The world’s demand for computing power continues to grow exponentially, yet CPUs are no longer keeping up as Moore’s Law has ended.
NVIDIA pioneered accelerated computing to solve this challenge.
Optimizing across the entire stack —from silicon to software — allows NVIDIA to advance computing in the post-Moore’s Law era for large and important markets:
Gaming, Pro Viz, High Performance Computing (HPC), AI, Cloud, Transportation, Healthcare, Robotics, AI IOT
ACCELERATED COMPUTING 1000X EVERY 10 YEARS
1980 1990 2000 2010 2020
103
105
107
GPU PERFORMANCECPU PERFORMANCE
+
GROWING INTO MULTIPLE LARGE MARKETS
Revenue in millions. 1H FY18 and 1H FY19 represents the first six months of each fiscal year.
Gaming+59%
Auto+9%
$0
$1,000
$2,000
$3,000
$4,000
1H FY18 1H FY19
$0
$100
$200
$300
$400
1H FY18 1H FY19
$3.5B
Pro Viz+21%
$0
$150
$300
$450
$600
1H FY18 1H FY19
$532M
Datacenter(HPC, Cloud, AI)
+77%
$0
$400
$800
$1,200
$1,600
1H FY18 1H FY19
$1.5B $306M
$0
$2
$4
$6
$8
$10
$12
$14
FY16 Annualized FY19
$5.0B
3 YEARS OF GROWTH AND MARGIN EXPANSION
Gross margin, operating margin, and FCF are non-GAAP measures. 1H FY19 represents first six months of the fiscal year.Annualized FY19 revenue and FCF is calculated based on Q2 FY19 results multiplied by four.
50%
55%
60%
65%
70%
FY16 1H FY19
0%
10%
20%
30%
40%
50%
FY16 1H FY19
$0
$1
$2
$3
$4
FY16 Annualized FY19
$1.1B
Gross MarginUp ~700 bps
Operating MarginUp ~2000 bps
Free Cash FlowUp 2.9X
22%57%
RevenueUp 2.5X
$12.5B 43%
64%
$3.1B
Three types of chips — GPU, SOC, and NVSwitch.
Three types of systems — GTX/RTX for graphics cards, DGX for AI/HPC, AGX for AI machines.
Accelerated computing requires an end-to-end stack, vertical market expertise, and ecosystem partners.
One architecture leverages scale and investment to address large, growing markets.
ONE ARCHITECTURE, EIGHT LARGE MARKETS
ACCELERATION STACKS
GAMING HPC TRANSPORTATION HEALTHCARE
PRO VIZ AI ROBOTICS AI IOT
DEVELOPERS
USERS
Best platform > more developers > more apps > more customers.
NVIDIA GPUs are available in every cloud, from every systems maker.
Accelerate every deep learning framework; 600 applications, including top 15 HPC apps.
1M DEVELOPERS ON NVIDIA COMPUTING PLATFORM
20182013
25K
20182013
8M
CUDA Downloads — 5X in 5 YrsGTC Attendees — 7X in 5 Yrs
TENSOR CORE RT CORE
SHADER | COMPUTE
NVIDIA invents RTX hybrid rendering model that combines rasterization, ray tracing, and AI.
Two new processors make it possible —RT Core and Tensor Core.
New RT Core achieves holy grail of computer graphics — real-time ray tracing.
New multi-precision Tensor Core is a giant leap for AI inference — up to 12X Pascal.
Advanced shaders deliver up to 2X speed-up for traditional rasterized games.
TURING — A GIANT LEAP FOR GRAPHICS AND AI
0
20
40
60
80
100
120
TURING 9X PEAK FLOPS
PASCALTITAN Xp
TURING2080 Ti
114
14.214.212.9
FP32 INT32 TC FP32 INT32 TC
MAXWELL PASCAL TURING
GTX 980 Ti
GTX 980
GTX 1080 Ti
GTX 1080
RTX 2080 Ti
RTX 2080
4K 60FPS
DEEP LEARNINGIMAGING
MAXWELL PASCAL TURING
GTX 980 Ti
GTX 980
GTX 1080 Ti
GTX 1080
4K 60FPSRTX 2080
RTX 2080 Ti
GeForce RTX is the biggest GPU leap in 10 years — first real-time ray tracing GPU, first 100+ TFLOPS GPU, first neural graphics processing.
First to enable 60Hz gaming on 4K monitors.
GeForce RTX delivers 2X performance boost with Deep Learning Super Sampling (DLSS, a new AI-enabled anti-aliasing technique), and up to 6X for ray-traced graphics.
Over 30 games support RTX including Battlefield V, Shadow of the Tomb Raider, and Metro Exodus.
RTX and rising GPU requirements help outgrow gaming market.
GEFORCE RTX RESETS $100B GAMING MARKET
FASTEST GAMING GPUPLAYS AT 4K 60FPS
DEEP LEARNING IMAGING WITH 114 TFLOPS TENSOR CORE
Quadro RTX delivers photoreal graphics that creators didn’t expect for another 5-10 years.
GPUs can now accelerate photoreal rendering for large industries that previously only used CPU server farms — 1.5M CPU servers worldwide.
RTX Server with 8 GPUs can replace 60 dual-CPU servers — reducing cost, space, and power.
Millions of servers, workstations, and PCs — estimated $50B of computing —are in use today for photoreal rendering.
RTX OPENS $250B VISUAL EFFECTS INDUSTRY
DESIGN AEC
VISUALIZATION FILM & TELEVISION
RTX-ACCELERATED RENDERERSPHOTOREAL VFX
RTX-ACCELERATED RENDER FARM: 1/4 THE COST, 1/10 THE SPACE, 1/11 THE POWER
11%
25%
56%
2015Tesla K80
2017Tesla P100
2018Tesla V100
DELIVERED 56% OF NEW TOP500 PERFORMANCE IN 2018
17 OF TOP 20 MOST ENERGY-EFFICIENT
5 OF WORLD’S TOP 7 MOST POWERFUL
#1 Summit: 122 PF#3 Sierra: 72 PF#5 ABCI: 20 PF#6 Piz Daint: 20 PF #7 Titan: 18 PF
NVIDIA powers five of top seven supercomputers, 17 of top 20 most energy-efficient.
Five of the six 2018 Gordon Bell prize finalists did their work on the new NVIDIA GPU-accelerated Summit and Sierra systems.
Delivered 56% of new performance on the latest Top500 list.
Accelerate ~100 of the Top500 —4X expansion opportunity.
NVIDIA GPUS POWER NEW SUPERCOMPUTERS IN $10B HPC MARKET
FASTESTSINGLE NODE
FASTESTSINGLE CHIP
FASTESTSINGLE NODE
FASTESTAT SCALE
FASTESTAT SCALE
108minutes
24hours
5hours
32minutes
6.6minutes
IMAGES TRANSLATION
DGX-2 WORLD’S LARGEST GPU: 2 PFLOPS, 512GB HBM2
Images: Single Chip: Resnet-50 V1 Training on Tesla V100 with NVIDIA NGC MXNet Container 18.09 Pre-release by NVIDIA. Single Node: Resnet-50 V1 Training on NVIDIA DGX-2 with NVIDIA NGC MXNet Container 18.09 by NVIDIA. At scale: Resnet-50 Training with 2048 P40 GPUs by Tencent. Translation: Single Node Strong NMT (Transformer) Training on WMT ‘14 English-German Translation with DGX-1 by Facebook Research. At Scale Strong NMT (Transformer) Training on WMT ‘14 English-German Translation with 16 DGX-1 Systems by Facebook Research.
AI training is the ultimate HPC challenge.
NVIDIA DGX is a reference full-stack-optimized system designed for AI, HPC, and big data analytics.
Shorter time to train equals faster progress in advancing better and more accurate AI.
WORLD’S MOST ADVANCED AI PLATFORM
12,000+ NVIDIA TENSORRT DOWNLOADS FROM 4,000 COMPANIES
CAMBRIAN EXPLOSION OF NETWORKS
Speech
NLP
ImageMaps
Video
Search
Programmability: Increasing diversity of neural networks
Latency: The right answer, right now
Accuracy: If the answer isn’t right, speed doesn’t matter
Size of network: Exploding neural network complexity
Throughput: Concurrently run many AI models
Efficiency: Deliver performance within power budget
Rate of learning: Constant retraining and deployment, sometimes multiple times per day
MASTERING THE COMPLEX CHALLENGE OF INFERENCE
Images: Fastest Inference: Resnet-50 Inference on Tesla V100 with TensorRT, Batch size: 1, Int8 optimized. Inference Throughput: Resnet-50 Inference on Tesla V100 with TensorRT, Batch size: 128, Int8 Optimized. Inference Efficiency: Resnet-50 Inference on Tesla T4, Int8 optimized, batch size = 32. Translation: GMNT Inference on Newstest2015 test dataset on Tesla V100.
13,160words/second
IMAGES TRANSLATION
6,250images/second
1millisecond
56images/second/watt
HIGHEST INFERENCE EFFICIENCY
HIGHEST INFERENCE THROUGHPUT
FASTEST INFERENCE
HIGHEST INFERENCE THROUGHPUT
Convolutional Networks Recurrent Networks
Generative Adversarial Networks Reinforcement Learning
GIANT LEAPS IN INFERENCE PERF.TENSORRT HYPERSCALE
21XASR
Deep Speech 2
36XNLPGNMT
20XRECOM
NCF
8XTTS
WaveNet
27XVIDEO/IMAGE
ResNet-50
DNN Models
NV DL SDK
Kubernetes
NV Docker
TensorRT
Inference Server
TENSORRT HYPERSCALE PLATFORM BOOSTS SERVER UTILIZATION
Tesla T4 is a universal inference accelerator designed for hyperscale datacenters.
TensorRT Hyperscale is a fully optimized Kubernetes stack and an inference server that enables multiple types of models from different frameworks to run concurrently on one GPU node.
TensorRT Hyperscale enables higher utilization, scalability, and resilience for large-scale deep learning operations.
TENSOR CORE OPS16X PASCAL FP INFERENCE
5.52265
130
260
0
50
100
150
200
250
300
FLOAT INT8 INT4 FLOAT INT8 INT4
P4 T4
65
130
260
5.522
FLOAT INT8 INT4 FLOAT INT8 INT4
0
5,000
10,000
15,000
20,000
2013 2014 2015 2016 2017 2018 2019
Maxwell FP32
Kepler FP32
Pascal FP32
Volta FP16
TuringINT8
TuringINT4
ResN
et-
50 I
nfe
rence (
i/s)
$20B TAM IN 2023 INDUSTRY LEADER ADOPTION
HGX FOR CLOUDDGX FOR DATACENTERS
Transportation
Healthcare
Public Sector
Manufacturing
Oil & Gas
Cloud/Other
Every industry is tapping into the power of AI; our full-system expertise enables enterprise adoption.
DGX-2: World’s most powerful AI system for datacenter deployments; $399K system can replace $3M of CPU servers.
HGX-2: Cloud server platform for OEM and ODM partners; Single HGX-2 can replace 300 dual-CPU server nodes.
AI IS REVOLUTIONIZING THE ENTERPRISE
WORKLOAD BASELINECPU-Only
HPC (Amber, LAMMPS, NAMD, VASP)
VISUALIZATION(Rendering, Ray Tracing)
AI TRAINING(Caffe2, TensorFlow, MXNet)
AI INFERENCE (Image, Speech, Translation)
Speed Up 1X 20X 60X >100X 200X
Servers 5,000 250 84 <50 25
Capex $45M $11M $10.5M $7.5M $1.5M
3-Year Opex
(Power + Cooling)$19.5M $2.5M $2M $1M $0.5M
TCO Savings N/A 79% 81% 86% 97%
“THE MORE YOU BUY, THE MORE YOU SAVE”
ACCELERATED DATACENTER VALUE
Note(s): CPU Baselined to 5000 Servers for each workload | Capex Costs: CPU node with 2x Skylake CPU’s ~$9K; | Opex Costs: Power & cooling is $180/kW/month | Power: CPU server + n/w = 0.6 KW; GPU server + n/w = 1.6 KW; DGX-1V/HGX-1 Server = 3.2KW | Visualization: GPU node with 8x RTX 8000 GPU’s compared to 12 core CPU server | HPC: GPU node with 4xV100 compared to 2xCPU Server | DL Training: DGX-1V compared to a 2xCPU server | DL Inference: GPU Node with 16 xT4 Compared to 2x CPU Server |numbers rounded to nearest $0.5M
Trajectory planning
Vehicle dynamics
LQR
F/I kinematics
Visualization
Compression
3D-2D transform
Image registration
Classification
Route/lane/path planning
Env prediction
A*/D*
Collision testing
Language translation
Tree search
Beam-forming
Iterative recon
Odometry
Tracking
SLAM
Deep Learning (CNN, RNN, GAN, U-Net, MLP, etc.)
EXAMPLE AI COMPUTATION PIPELINE
INPUT DATA PROCESSING
PERCEPTION &INFERENCE
REASONING ACTION
Tensor Core
CUDA
CPU
Tensor Core
CUDA
CPU
CUDA
GPU
Video Processor
CPU
“Where can I go for ramen?”
“Let me suggest Jankara Ramen”
ISP
Video Processor
Vision Process
CUDA
CPU
Calibrate sensor
Decompression
Tone mapping
De-warping
ICP
FFT/IFFT
De-noise
Optical flow
Motion comp
Feature point
Xavier, the first chip ever designed for autonomous AI systems, will enable automation in many large industries.
AGX is our line of systems that addresses this opportunity.
DRIVE AGX for autonomous cars, trucks, and robotaxis.
Jetson AGX for robots and drones, as well as AI IOT systems for safe and smart cities.
Clara AGX for next-generation imaging in medical instruments.
For each, we’ve created a full stack to enable market leaders to take advantage of our platform.
AGX FOR AUTONOMOUS MACHINES AND AI IOT
NVIDIA AGX
DRIVE CLARAJETSON
SIMULATINGTRAINING
DRIVING
DRIVE is a full-stack open platform for the transportation industry.
More than a driving computer and software stack, DRIVE is an HPC infrastructure for training, simulation, and testing.
DRIVE AGX car computer delivers the performance needed to run a wide array of deep learning models and algorithms concurrently.
Companies can build their own AV applications with the new DRIVE AGX DevKit.
NVIDIA DRIVE FOR AV REVOLUTION
Pedestrians LightsSignsPathLanesCars
Hundreds of automakers, mobility services, truck makers, tier ones, mapping, and sensor companies are developing on DRIVE to create the future of transportation.
Toyota is using DRIVE AGX Xavier as the AI brain in production cars beginning 2020.
Daimler and Bosch are developing robotaxi fleets with DRIVE.
VW is developing AI-infused cockpit experiences on DRIVE.
$25B TAM opportunity by 2025.
NVIDIA DRIVE ADOPTION ACROSS $10T TRANSPORTATION INDUSTRY
The next chapter of AI is autonomous machines.
Jetson AGX Xavier delivers the energy-efficient computational power needed for embedded systems like robots, drones, and smart cities.
NVIDIA Isaac is a simulation environment and SDK that accelerates the development and deployment of robots of all kinds.
Jetson AGX Xavier DevKit is available for order now.
$5B TAM opportunity by 2022.
JETSON AGX AND ISAAC DELIVER AI TO $250B ROBOTICS AND IOT INDUSTRY
NVIDIA ISAACJETSON AGX XAVIER DEV KIT
Scale Up to 200 TOPS DL Processing
8 GIGA Rays | 200W
AI-ENHANCED ULTRASOUNDCLARA AGX
3D SegmentedVisualization
2DVisualization
Ultrasound, CT, MR, Endoscopy, Sequencing, and Pathology allow us to see the invisible in the human body.
Instruments are large sensors that require supercomputing level of processing for high-speed sensor processing, reconstruction, AI, and visualization.
Traditionally build with FPGA, CPU, and GPU.
Clara AGX Xavier replaces all three into one programmable architecture.
$2B TAM opportunity by 2022.
JETSON AGX TRANSFORMS $100B MEDICAL IMAGING INDUSTRY
Large markets — Gaming, Pro Viz, HPC, AI, Transportation, Healthcare, Robotics, and AI IOT — adopt NVIDIA Accelerated Computing as Moore’s law ends.
1M+ developers
56% of new FLOPS in Top500
12,000+ TensorRT downloads from 4,000 companies
New Turing is a giant leap for GPU — extending leadership in gaming and opening new growth for rendering and hyperscale AI.
New AGX line fuses HPC and AI — opening new growth opportunities in AV, medical instruments, robotics, and AI IOT.
Leveraging one architecture across multiple large markets allows us to drive growth and margin expansion.
INCREASING NVIDIA ADOPTION
NEW PRODUCTS
NEW MARKETS
ANOTHER GIANT LEAP
RECONCILIATION OF NON-GAAP TO GAAP FINANCIAL MEASURES
GROSS MARGIN ($ in millions & margin percentage)
NON-GAAPSTOCK-BASED
COMPENSATION (A)
PRODUCT WARRANTY
(B)GAAP
FY16
$2,846 (15) (20) $2,811
56.8% (0.3) (0.4) 56.1%
1H FY19
$4,059 (16) — $4,043
64.1% (0.2) — 63.9%
RECONCILIATION OF NON-GAAP TO GAAP FINANCIAL MEASURES
A. Stock-based compensation charge was allocated to cost of goods sold
B. Consists of the release of warranty reserve balance and warranty charge associated with a product recall
OPERATING MARGIN ($ in millions & margin percentage)
NON-GAAPSTOCK-BASED
COMPENSATION (A)
RESTRUCTURING
COSTS
OTHER
(B)GAAP
FY16
$1,125 (205) (131) (42) $747
22.5% (4.1) (2.6) (0.9) 14.9%
1H FY19
$2,718 (262) — (4) $2,452
42.9% (4.2) — — 38.7%
RECONCILIATION OF NON-GAAP TO GAAP FINANCIAL MEASURES (CONTD.)
A. Stock-based compensation charge was allocated to cost of goods sold, research and development expense, and sales, general and administrative expense.
B. Consists of the release of warranty reserve balance and warranty charge associated with a product recall, amortization of acquisition-related intangible assets, transaction costs, compensation charges, and other credits related to acquisitions.
($ IN MILLIONS) FY16 Q2 FY19
GAAP net cash flow provided by operating activities $1,175 $913
Purchase of property and equipment
and intangible assets(86) (128)
Free cash flow $1,089 $785
RECONCILIATION OF NON-GAAP TO GAAP FINANCIAL MEASURES (CONTD.)