slidedeck.co

NVIDIA investor presentation, October 2025 non-deal roadshow

Cover slide of the NVIDIA 2025 investor presentation
Cover of NVIDIA's 2025 investor presentation.

The deck opens with a title slide and a forward-looking statements disclaimer, then a slide tracing NVIDIA's shift from chips to an AI infrastructure company. The core argument is built around a repeated headline, "$3-4 Trillion AI Infrastructure Spend by 2030", which carries through the middle section. It covers the end of Moore's Law and accelerated computing, the CUDA-X platform portfolio shown as an icon row, and the three AI scaling laws (pre-training, post-training and test-time scaling) with a token-growth summary. A run of three near-identical slides on physical AI (robots, autonomous cars and factories) each pair a training computer with a digital-twin computer. The later section covers an annual hardware roadmap from Blackwell through Rubin and Feynman, the co-design argument, MLPerf benchmark wins and a closing slide on the OpenAI partnership and its gigawatt build-out commitment. The deck closes on that partnership slide, with no separate summary. Slides worth studying include the roadmap chart on slide 10, the MLPerf line chart on slide 12 and the repeated three-computer diagram on slide 8.

Slides worth studying

  • Slide 5: A dense icon grid of CUDA-X libraries uses one consistent visual per library and a single headline to make the platform breadth legible at a glance.
  • Slide 10: The multi-year roadmap is laid out as a timeline with tiered rows for compute, scale-up and networking, which makes a complex product plan readable.
  • Slide 12: A simple cumulative line chart with three colour-coded series and quoted press headlines beside it makes the benchmark lead easy to read at a distance.

Details

  • Company: NVIDIA
  • Year: 2025
  • Type: Investor presentation
  • Industry: Semiconductors
  • Pages: 15

Review score: 6 out of 10, Needs substantial work

Presentation quality, judged with the SlideDeck Co review heuristic. Ask, evidence and idea merit are not rated here.

The deck has a strong visual identity and several strong individual pages: the CUDA-X image row on slide 5, the roadmap on slide 10 and the MLPerf chart on slide 12 are clear and well made. The route is weaker. The same headline, '$3-4 Trillion AI Infrastructure Spend by 2030', sits over unrelated content on slides 4 to 9, so titles stop saying what each page proves, and the three near-identical physical AI slides 7 to 9 repeat one diagram. Slides 4, 13 and 14 use very small text and sparse diagrams that would be hard to read in a room, and the deck ends on a dense partnership slide and a logo page with no summary.

Worth keeping: Keep the consistent green and black identity, the full-width CUDA-X image row on slide 5 and the labeled roadmap on slide 10, which are legible and give the deck a recognisable look.

Most useful repair: Replace the repeated headline with action titles that state each slide's own point, merge slides 7 to 9 into one diagram with three examples, and enlarge the text on slides 4, 13 and 14.

What would change the score: Add a closing summary slide that restates the argument and the next step, and cut the small-print paragraphs on slides 4 and 14 into fewer, larger points.

Confidence medium. 15 of 15 pages inspected on contact sheets, 4 at readable size. Reviewed 2026-11-10. Model-assisted review (claude-sonnet-5-5).

Full text of the slides (15 slides)

Every slide's words and a description of its charts and images, so the deck can be searched and rebuilt. Also available as Markdown.

Slide 1

NVIDIA / NVIDIA Non-Deal Roadshow / October 2025

Visuals: Full-bleed abstract illustration covering the whole slide: layered, glossy green concentric arcs rendered in 3D, sweeping from the lower right toward the upper left. The arcs are bright lime green (#8BC41A to #A6DA2A) with deep shadowed green (#0F4D0F) recesses at the centre-right, where a dark interior opens up. Thin pale highlight lines run across the surface. The left and top are fading to white, creating a soft diffuse gradient. The image is cropped at the slide edges, with no border or frame and no text overlay on the artwork. Logo: the NVIDIA logo at top-left, consisting of a green (#76B900) eye-shaped emblem inside a square, followed by the black NVIDIA wordmark in bold sans-serif capitals. A second small image sits at the bottom-right corner and is not clearly legible against the green artwork.

Slide 2

Certain statements in this presentation including, but not limited to, statements as to: expectations with respect to growth, performance and benefits of / NVIDIA’s products, services and technologies, and related trends and drivers; expectations with respect to supply and demand for NVIDIA’s products, services / and technologies, and related matters; expectations with respect to NVIDIA’s third party arrangements, including with its collaborators and partners such as / OpenAI, Intel and CoreWeave; expectations with respect to NVIDIA’s investments; expectations with respect to energy efficiency; expectations with respect to / NVIDIA’s strategies; expectations with respect to technology developments, and related trends and drivers; projected market growth and trends; expectations / with respect to AI and related industries; and other statements that are not historical facts are forward-looking statements within the meaning of Section 27A / of the Securities Act of 1933, as amended, and Section 21E of the Securities Exchange Act of 1934, as amended, which are subject to the “safe harbor” created / by those sections based on management’s beliefs and assumptions and on information currently available to management and are subject to risks and / uncertainties that could cause results to be materially different than expectations. / Important factors that could cause actual results to differ materially include: global economic and political conditions; NVIDIA’s reliance on third parties to / manufacture, assemble, package and test NVIDIA’s products; the impact of technological development and competition; development of new products and / technologies or enhancements to NVIDIA’s existing product and technologies; market acceptance of NVIDIA’s products or NVIDIA’s partners’ products; design, / manufacturing or software defects; changes in consumer preferences or demands; changes in industry standards and interfaces; unexpected loss of / performance of NVIDIA’s products or technologies when integrated into systems; and changes in applicable laws and regulations, as well as other factors / detailed from time to time in the most recent reports NVIDIA files with the Securities and Exchange Commission, or SEC, including, but not limited to, its annual / report on Form 10-K and quarterly reports on Form 10-Q. Copies of reports filed with the SEC are posted on the company’s 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, and, except as / required by law, NVIDIA disclaims any obligation to update these forward-looking statements to reflect future events or circumstances. / Many of the products and features described herein remain in various stages and will be offered on a when-and-if-available basis. The statements within are not / intended to be, and should not be interpreted as a commitment, promise, or legal obligation, and the development, release, and timing of any features or / functionalities described for our products is subject to change and remains at the sole discretion of NVIDIA. NVIDIA will have no liability for failure to deliver or / delay in the delivery of any of the products, features or functions set forth herein.

Visuals: A full-bleed background photograph-like abstract of light green swooshing streaks, with thin bright white-lime highlight lines curving from the lower left toward the upper right. The top-left is nearly white and the lower-right is saturated lime green. Dominant colours are pale mint #D8F5C0 and lime green #8CD03A, with the brightest highlights near white #FFFFFF. The image is not cropped and is not boxed; it covers the full slide with no overlay. A small image at bottom right (387 by 150 pixels, about 4.8% by 3.4% of the slide) is a logo, presumably the NVIDIA wordmark; it is not legible against the green background.

Slide 3

NVIDIA’s Evolution From Chips to an AI Infrastructure Company

ACCELERATOR

ACCELERATED COMPUTING

AI

AI SUPERCOMPUTER

2016

DGX-1

NVL8

1999

GPU

2006

CUDA

2012

cuDNN

AI

DATACENTER

2019

MELLANOX

2020

SELENE

2024

NVL72 SCALE-UP

3-SCALING LAWS

AI INFRASTRUCTURE

SPECTRUM-X

SCALE-OUT AND

SCALE-ACROSS

Visuals: Hand-drawn white line sketches with gold-olive shading, arranged as a rising timeline. From left to right: a small GPU card (bottom-left, about 4 percent wide) labelled 1999 GPU, with an ACCELERATOR label above it and a 2006 CUDA label; a short DGX-1 style server box (bottom-centre-left) labelled 2016 DGX-1 NVL8; a wide DGX-style rack chassis (centre-right, about 14 percent wide, shaded in olive-gold with a grid of vertical fins) labelled 2019 MELLANOX and 2020 SELENE; a tall vertical NVL72 rack (about 3 percent wide, 18 percent tall, outlined in white with a gold-olive inner panel) labelled 2024 NVL72 SCALE-UP and 3-SCALING LAWS; and at top-right four flat rack-mount switch units arranged in a 2x2 isometric cluster, drawn in white outline with gold-olive highlights, labelled AI INFRASTRUCTURE and SPECTRUM-X SCALE-OUT AND SCALE-ACROSS. Short white hand-drawn connector strokes join successive labels and sketches, and a long diagonal stroke runs from the NVL72 rack up to the switch cluster. No charts, tables or photographs. Lighting is flat and even, with soft gold glow on the shaded surfaces.

Slide 4

$3-4 Trillion AI Infrastructure

Spend by 2030

Key TAM Growth Drivers

End of Moore’s Law drives fundamental shift from general-purpose

to accelerated computing

For example, Data Processing: NVIDIA cuDF, cuML, cuVS accelerates

structured and unstructured data processing 10-100X over CPU-based

methods; Computational Lithography: NVIDIA cuLitho accelerates computation

lithography tasks, such as creating photomasks, by as much as 40-60X over

CPU-based methods; Genome Sequencing: deciphEHR was able to achieve

>5X faster alignment runtimes and >10X faster variant calling runtimes,

using NVIDIA Parabricks

Hyperscale shift to Generative AI

For example, Ad Generation: Google AI-powered video campaigns on YouTube

deliver 17% higher ROAS (return on ad spend) than manual campaigns;

Recommender Systems: Pinterest was able to move to 100X larger recommender

models by adopting NVIDIA GPUs which increased engagement by 16%; Search

and User-Generated-Content are moving to adopt LLM-powered generative AI

Model Makers — A new industry

OpenAI, Google, Anthropic, xAI, Meta are building the foundations of AI

Enterprise — Agentic AI enters the labor market

Coding: Developers using AI coding assistants could complete tasks up to 55%

faster, according to an MIT study; Vibe Coding: Lovable opens coding to new

users like product designers, creatives, marketing, IT, teachers; Legal: STARA,

an AI designed for statutory research, conducted a task that would take two

humans 8 to13.5 hours and cost ~$3,000 in 20 minutes at a cost of ~$0.86

Robotics, AV, Factories, and Edge powered by Physical AI

Labor shortages drive everything that moves in $50T industrials to

be autonomous

4,000

3,000

2,000

1,000

0

($ Billions)

2022

2023

2024

2025

2026

2027

2028

2029

2030

NVIDIA AI Revenue

Data Center Capex

Global Public Cloud, Private Cloud, Enterprise On-Prem, Edge

~40% 5-Year CAGR

(2025-2030)

Visuals: One clustered column chart on the right panel, with no chart title. Y-axis: labelled ($ Billions), range 0 to 4,000, tick labels 0, 1,000, 2,000, 3,000, 4,000, no gridlines. X-axis: years 2022 to 2030, one category per year. Series 1 NVIDIA AI Revenue in green #76B900, with values about 20 (2022), about 50 (2023), about 110 (2024), and no visible bar for 2025 to 2030 at this scale. Series 2 Data Center Capex in grey #8C8C8C, with values about 270 (2022), about 300 (2023), about 440 (2024), about 690 (2025), about 960 (2026), about 1,290 (2027), about 1,800 (2028), about 2,420 (2029), about 3,500 (2030); no value labels are printed, so all are estimates. Legend below the axis: green square then NVIDIA AI Revenue, grey square then Data Center Capex. Caption under the legend: Global Public Cloud, Private Cloud, Enterprise On-Prem, Edge. A thin black diagonal arrow with an arrowhead at its upper-right end runs from about the 2025 Data Center Capex bar top to just above the 2030 bar, annotated with the text ~40% 5-Year CAGR (2025-2030) centred above the arrow. No highlighted points. Bars are flat with no outlines. The NVIDIA logo at bottom-left is a green and black wordmark.

Slide 5

$3-4 Trillion AI Infrastructure Spend by 2030

NVIDIA CUDA-X Platforms Enable Shift From CPU to GPU Accelerated Computing

cuLitho

Computational

Lithography

cuDSS

cuSPARSE

cuFFT

AmgX

Computer Aided

Engineering

cuOpt

Decision

Optimization

Warp

Physics

cuDF

cuML

Data Science

and Processing

Megatron

Dynamo

NIXL

cuDNN

CUTLASS

Deep

Learning

cuEquivariance

cuTensor

Quantum

Chemistry

cuQuantum

CUDA-Q

Quantum

Computing

Earth-2

Weather

Analytics

MONAI

Medical

Imaging

Parabricks

Genomics

Aerial

Sionna

5G/6G

Signal Processing

cuPyNumeric

Numerical

Computing

Visuals: Thirteen portrait screenshot tiles arranged left to right, each a rounded rectangle with a thin dark bevel border and slight perspective tilt, overlapping neighbours and staggered vertically. Tile 1 (cuLitho): dark tall panel with glowing cyan lithography machine structure and rows of bright cyan lit modules, dominant colours black, cyan, teal. Tile 2 (cuDSS, cuSPARSE, cuFFT, AmgX): grey studio-style car in a white-grey environment with a splash of yellow and gold fluid, dominant colours grey, black, gold. Tile 3 (cuOpt): dark blue field with a glowing green-yellow node at the centre and a web of connected white and blue points and lines, dominant colours navy, green, blue. Tile 4 (Warp): swirling multicolour fluid simulation in magenta, green, blue and white on black. Tile 5 (cuDF, cuML): black field with grey-violet and teal particle clouds forming abstract shapes. Tile 6 (Megatron, Dynamo, NIXL, cuDNN, CUTLASS): blue-to-black gradient with a grey circuit-like diagram of stacked rectangles and connector lines. Tile 7 (cuEquivariance, cuTensor): black background with a ball-and-stick molecule in cyan, blue, red and white atoms. Tile 8 (cuQuantum, CUDA-Q): gold-and-copper cryostat quantum computer tower on dark background. Tile 9 (Earth-2): Earth from orbit with white clouds over blue oceans and coloured weather patterns. Tile 10 (MONAI): dark background with a 3D segmented medical volume in green, red, pink, yellow and purple blobs. Tile 11 (Parabricks): blue molecular surface of a protein with an orange-and-red DNA helix on dark background. Tile 12 (Aerial, Sionna): aerial photo of city buildings with bright green ray-tracing lines radiating across it. Tile 13 (cuPyNumeric): colourful nebula in pink, red, blue and yellow on black. All tiles are full-bleed within their frames, not cropped to a circle; no overlays are visible.

Slide 6

$3-4 Trillion AI Infrastructure Spend by 2030

3 AI Scaling Laws Driving Exponential Computing Demand

NVIDIA Excels at Pre-Training, Post-Training, and Inference

Pre-Training Scaling

Test-Time Scaling

“Long Thinking”

Post-Training Scaling

Perception AI

Generative AI

Agentic AI

Physical AI

Token Generation is Doubling Every Two Months

•

ChatGPT is at ~700M WAUs, with usage up ~4X y/y

•

OpenAI now counts 5M paying business users,

up from 3M in June

•

Microsoft processed over 500 trillion tokens served by

Foundry APIs in FY2025, up 7X y/y

•

Alphabet processed over 980 trillion tokens in the month

of June across its AI services, up from a monthly run-rate

of 480 trillion tokens in May

•

The Gemini app had more than 450M MAUs as of the end

of July with daily requests up >50% in Q2 vs. Q1

Visuals: Schematic line diagram, not a data chart: no numeric axes, scale, ticks or legend. A black L-shaped axis: vertical line at about x 6.9% running from about y 32% to 74%, and a horizontal line at about y 74% running from x 6.9% to about x 45.6%, with black arrowheads on the segments between stage labels. Stage labels along the horizontal axis, black: Perception AI, Generative AI, Agentic AI, Physical AI, separated by black arrows. Three green (#76B900) stroked curves rise diagonally from lower left to upper right, each drawn as a smooth line with a kink: (1) the Pre-Training Scaling curve starts near the bottom-left of the plot at about x 8.6%, y 72% and rises to about x 18%, y 61.6%, continuing as a straight segment ending in a filled green arrowhead at about x 21.6%, y 58.2%; (2) the Post-Training Scaling curve branches from the first curve at about x 18%, y 61.6% and rises to about x 28.4%, y 46.5%, ending without an arrowhead; (3) the Test-Time Scaling curve branches from the second at about x 23.4%, y 51% and rises steeply to about x 34.8%, y 32.3%, ending at the top edge of the plot. Three green text labels sit beside the curves. No highlighted points. A small NVIDIA logo (green eye-like mark with the word NVIDIA in black) appears bottom right at about x 93.1%, y 94.8%.

Slide 7

$3-4 Trillion AI Infrastructure Spend by 2030

NVIDIA 3-Computers Enable Physical AI – the Next Wave of AI

Manipulating Robot

Jetson AGX Thor

Training

Robot AI

Practicing in

Digital Twin

GB200

RTX PRO Server

NVIDIA

Visuals: Diagram: a hub-and-branch flow. Top centre, a black circular photo of a humanoid robot (labelled Manipulating Robot) with a green outline sits over a small Jetson AGX Thor module. From the circle's left and right sides, thin green (#76B900) horizontal lines run outward then turn down; each ends in a downward arrowhead above its label (left to Training Robot AI, right to Practicing in Digital Twin). Below the left label is a photo panel: a robotics lab with a grey floor, a robot hand, a white spray bottle on a wooden table, a grey robot arm, and translucent pink, blue and green motion-trajectory arcs overlaid. Below the right label is a rendered digital-twin kitchen with a humanoid robot. Under each photo panel sits a green icon and a server: under the left panel a green hexagonal icon with a network-graph symbol above a black GB200 server rack; under the right panel a green ring icon above a black RTX PRO Server rack. A green line runs from beneath the GB200 rack, horizontally to beneath the RTX PRO Server, with upward-pointing arrowheads at both ends. No chart or table.

Slide 8

$3-4 Trillion AI Infrastructure Spend by 2030 / NVIDIA 3-Computers Enable Physical AI – the Next Wave of AI / Driving Car / Training / AV AI / Practicing in / Digital Twin / DRIVE AGX Thor / GB200 / RTX PRO Server

Visuals: Diagram with three nodes and connecting arrows. Top node: circular photo (about 8.9% wide) of a translucent cyan wireframe car on dark background, with the NVIDIA DRIVE AGX Thor computer board (dark green-black PCB with heatsink and cyan connector strip) overlapping its bottom edge, labelled 'DRIVE AGX Thor' below. Circle has a thin #76B900 outline. Left node: rectangular photo (about 24.9% by 30.2%, 876 by 600 px) of a white sports sedan on a bridge road with pink and cyan light trails and a suspension bridge in the background, bright daylight tones, boxed with a 2-point #76B900 rounded border, labelled 'Training / AV AI' above; a green-and-black stylised GPU/neural network icon (cube with node graph) overlaps its bottom edge, and below it a black server rack photo labelled 'GB200'. Right node: rectangular screenshot (about 24.9% by 30.2%, 1622 by 1080 px) of a digital-twin city street with a grey car, orange excavators, orange and white barriers and cones, and a left-side asset library panel with thumbnail grid, boxed with a #76B900 rounded border, labelled 'Practicing in / Digital Twin' above; a green ring icon overlaps its bottom edge, and below it a dark server rack photo labelled 'RTX PRO Server'. Connectors: a #76B900 line 0.2% thick runs from the left edge of the Driving Car circle horizontally to x about 27.2%, then down with an arrowhead pointing down onto the Training label; a matching line runs from the right edge of the circle to x about 73%, then down with an arrowhead onto the Practicing label. A bottom #76B900 line runs from x about 27.2% to 72.8% at y about 91.5%, with upward arrowheads at both ends pointing to GB200 and RTX PRO Server.

Slide 9

$3-4 Trillion AI Infrastructure Spend by 2030 / NVIDIA 3-Computers Enable Physical AI – the Next Wave of AI / Running Factories / Training / Factory AI / Jetson AGX Thor / Practicing in / Digital Twin / GB200 / RTX PRO Server

Visuals: Diagram and photographs. (1) Circular photo of an aerial view of a factory floor, bordered by a thin green ring, labelled Running Factories above; below it is a small black Jetson AGX Thor module (a black PCB with a square chip and a green-tinted heatsink-like die) labelled Jetson AGX Thor. (2) Green elbow connectors, with arrowheads, run from the circle's left and right sides to the top of the two image panels. (3) Left rounded box containing a 3D digital-twin render of a factory floor from above, with white machinery, grey walkways and small blue-and-grey figures, plus a strip of four smaller interior camera views along its bottom and a vertical strip on its right showing a grid of coloured tracking dots and paths labelled RTLS and people; the box has a black sky area at top. Below it sits a green 3D cube icon with a black neural-network node pattern overlapping the box's bottom edge. (4) Right rounded box containing a grey-toned interior render of a long, empty, brightly lit factory hall with rows of grey machines along both sides, a small figure in a blue cap near the left, and a yellow stripe on the floor; the box has a green border. Below it sits a green ring icon with a black centre. (5) Black GB200 server rack photo below the left image, labelled GB200. (6) Black RTX PRO Server rack photo below the right image, labelled RTX PRO Server. (7) A green elbow connector with upward arrowheads runs along the bottom between the GB200 and RTX PRO Server racks. Lighting is flat and even; no overlays or tints.

Slide 10

Annual Rhythm and Extreme Co-Design for Sustained Leadership / Full-Stack | One Architecture | CUDA Everywhere / Blackwell / Rubin / Feynman / COMPUTE / Blackwell 8S HBM3e / Blackwell Ultra 8S HBM3e / Grace CPU / BlueField-3 / NVLINK (SCALE-UP) / NVLink 5 Switch 1800 GB/s / NETWORKING (SCALE-OUT) / Spectrum5 51T / CX8 800G / Rubin 8S HBM4 / Rubin CPX GDDR7 / Rubin Ultra 16S HBM4e / Feynman Next–Gen HBM / Vera CPU / BlueField-4 / BlueField-5 / Vera CPU / NVLink 6 Switch 3600 GB/s / NVLink 7 Switch 3600 GB/s / NVLink 8 Switch / Spectrum6 102T, CPO / CX9 1600G / Spectrum7 204T, CPO / CX10 / SYSTEM / Oberon / NVL72 / Liquid Cooled / Kyber / NVL576 / Liquid Cooled / 2025 / 2026 / 2027 / 2028

Visuals: Diagram/roadmap in three generation columns. Blackwell panel (2025 to 2026 span) holds Blackwell 8S HBM3e, Blackwell Ultra 8S HBM3e, Grace CPU, BlueField-3, NVLink 5 Switch 1800 GB/s, Spectrum5 51T and CX8 800G chip images with labels, plus a tall liquid-cooled rack labelled Oberon NVL72 Liquid Cooled. Rubin panel (2027) holds Rubin 8S HBM4, Rubin CPX GDDR7, Rubin Ultra 16S HBM4e, Vera CPU, BlueField-4, NVLink 6 Switch 3600 GB/s, NVLink 7 Switch 3600 GB/s, Spectrum6 102T CPO and CX9 1600G, plus a tall liquid-cooled rack labelled Kyber NVL576 Liquid Cooled. Feynman panel (2028) holds sketch-style placeholder illustrations for Feynman Next-Gen HBM, Vera CPU, BlueField-5, NVLink 8 Switch, Spectrum7 204T CPO and CX10. Chip images are small gold-framed square photographs of processor dies and packages with multicoloured circuit patterns; the Feynman items are white hand-drawn sketch outlines with gold shading. No charts or tables.

Slide 11

Annual Rhythm and Extreme Co-Design for Sustained Leadership / Combining simultaneous breakthroughs in GPU, CPU, NIC, NVLink scale-up fabric, Spectrum-X Ethernet scale-out and scale-across network, system architecture, and a mountain of software and algorithms, we are delivering leaps in performance and cost efficiency never seen before. / NVIDIA extreme co-design has delivered 1,000,000X performance gain in past 10 years – versus 100X of traditional Moore’s law. / GB200 NVL72 SuperPOD / Grace Blackwell MGX Node / NVLink Switch / Quantum Switch / Spectrum-X Switch / Chips Purpose-Built for AI Supercomputing / GPU | CPU | DPU | NIC | NVLink Switch | IB Switch | ENET Switch / Chips to Systems to NVLink to networking to CUDA-X, DOCA, NCCL, TRT-LLM, NIXL, to Dynamo

Visuals: One photograph-style rendered image, boxed (not full-bleed), at x 13.2% y 40.4% w 73.5% h 47.5% with sharp square corners. Subject: NVIDIA rack-scale hardware on a dark studio backdrop. Left to right: a Grace Blackwell MGX Node (a dark-grey 1U-style server chassis, open to show boards, lying upright), an NVLink Switch (a similar open chassis with blue-grey cabling loops on top), ten tall dark rack cabinets labelled GB200 NVL72 SuperPOD at the top centre (four racks, two central darker racks with fan panels, then four racks, all with green-lit cable loops), a Quantum Switch (a dark-grey chassis with a heatsink and two processor packages), and a Spectrum-X Switch (a taller chassis with a large central heatsink). Below the racks is a row of seven chips, left to right: GPU, CPU, DPU, NIC, NVLink Switch, IB Switch, ENET Switch, each a square package with gold, silver or dark-grey casing, under the caption Chips Purpose-Built for AI Supercomputing and the list GPU | CPU | DPU | NIC | NVLink Switch | IB Switch | ENET Switch. Composition is a frontal, centred, straight-on view with the racks in a single row. Lighting is soft and even from above, with spotlight highlights on the chips. Dominant colours are charcoal #2B2B2B to near-black #1A1A1A, with gold #C8A24A and silver-grey #9A9A9A accents and green #76B900 cable highlights. Not cropped. Overlay: none. Labels inside the image are white or light grey small sans-serif text.

Slide 12

Annual Rhythm and Extreme Co-Design for Sustained Leadership / NVIDIA Leads MLPerf With Hundreds of Training and Inference Wins / MLPerf Training and Inference Wins / NVIDIA GPU / Google TPU / All Others / 300 / 250 / 200 / 150 / 100 / 50 / 0 / Cumulative Wins / 2018 / 2019 / 2020 / 2021 / 2022 / 2023 / 2024 / 2025 / “NVIDIA Blackwell Ultra Sets Reasoning Records in MLPerf Debut” / – GamesBeat / “NVIDIA Unveils Rubin CPX Amidst Chart-Topping Blackwell Ultra MLPerf Results” / – HotHardware / “Blackwell GPUs Lift NVIDIA to the Top of MLPerf Training Rankings” / – HPC Wire / “NVIDIA’s Blackwell Conquers Largest LLM Training Benchmark” / – IEEE Spectrum / MLPerf Training reported time-to-train and MLPerf Inference: Data Center Reported Total Score / NVIDIA

Visuals: One line chart titled 'MLPerf Training and Inference Wins' with a legend above the plot (NVIDIA GPU, Google TPU, All Others). The vertical axis is titled 'Cumulative Wins' and runs from 0 to 300 in steps of 50, with no gridlines. The horizontal axis shows the years 2018 to 2025 as category labels. Series values are not printed on the slide, so the following are estimates: NVIDIA GPU (green line) about 5 in 2018, about 15 in 2019, about 39 in 2020, about 81 in 2021, about 119 in 2022, about 175 in 2023, about 230 in 2024 and about 267 in 2025; Google TPU (blue line) about 0 in 2018, about 5 in 2019, about 5 in 2020, about 9 in 2021, about 13 in 2022 and flat at about 13 through 2025; All Others (grey line) about 0 from 2018 to 2020, about 2 in 2021, about 8 from 2022 to 2024 and about 14 in 2025. No points are highlighted. Lines are about 2 px wide with no markers. The chart is a flat vector graphic with no shadows. Also present is the NVIDIA logo (green and black wordmark, bottom right), which is the only logo; no photographs or icons.

Slide 13

Annual Rhythm and Extreme Co-Design for Sustained Leadership

GB NVL72 for Training to Inference

Order of Magnitude Leap in Inference Throughput and Cost Reduction

DEEPSEEK R1

12,500

GB200

15X TPS/GPU

VS. H200

10,000

7,500

5,000

2,500

0

H200

1

50

THROUGHPUT

TPS PER GPU

(HIGHER AI FACTORY OUTPUT)

INTERACTIVITY

TPS PER USER

(HIGHER IS BETTER)

NVIDIA

Visuals: Hand-drawn line chart titled DEEPSEEK R1 in sketch-style capitals. Type: line chart with a dashed vertical reference line at x = 50. Y axis labelled THROUGHPUT, TPS PER GPU (HIGHER AI FACTORY OUTPUT), with tick labels 0, 2,500, 5,000, 7,500, 10,000 and 12,500 (units TPS per GPU). X axis labelled INTERACTIVITY, TPS PER USER (HIGHER IS BETTER), with tick labels 1 and 50 (units TPS per user), range about 1 to about 100 or more. Series 1, GB200: thick green (#76B900) line starting about 11,800 at x about 1.5, falling to about 10,300 at x about 10, about 9,500 at x = 50 (the crossing with the dashed line), about 8,800 at x about 65, and ending about 6,800 at x about 95. Series 2, H200: thin grey (#8C8C8C) line starting about 1,300 at x about 1.5, about 900 at x about 10, about 600 at x = 50, and ending about 300 at x about 85. Highlighted point: the GB200 curve where it meets the dashed line at x = 50, with a grey leader line from that point up to the annotation 15X TPS/GPU VS. H200. Labels GB200 and H200 sit at the start of their curves. Black axes are hand-drawn with rough strokes. Thick black up arrow at left of the Y axis label; thick black right arrow beside the X axis label. No legend; series are identified by direct labels. Chart is monochrome black and grey with green for GB200. The NVIDIA logo is green and black at bottom right.

Slide 14

NVIDIA and OpenAI Partnership / Multi-Year, Multi-Generational Build Out of / at Least 10 Gigawatts / • The OpenAI partnership is a powerful demonstration of / NVIDIA’s ability as an AI infrastructure partner – delivering / architecture, chips, systems, networking, data centers, / software, operations, and financing as one integrated solution / • The partnership extends existing collaboration to scale / multi-giga-watts of infrastructure at Microsoft, OCI, / and CoreWeave / • In addition to CSP capacity, OpenAI and NVIDIA will build / at least 10 gigawatts – millions of GPUs – of infrastructure / to be operated by OpenAI; The first gigawatt launches in / 2026 on the Vera Rubin platform / • For the first time, OpenAI will buy directly from NVIDIA, / secure multi-cycle supply, forging close engineering / collaboration to build AI factories / • NVIDIA intends to invest in increments over time up to / $100 billion in equity; Every 1GW build-out will require / $50-60 billion in total spend. OpenAI will need to re-invest / future revenue and/or secure other sources of financing to / cover the total cost of their build outs.

Visuals: Two logos on the right panel. (1) OpenAI wordmark: black lowercase-and-capital sans-serif wordmark reading OpenAI on a light grey background, unboxed, not cropped, centred horizontally in the right panel at about 67.6 percent from the left and 27.6 percent from the top. (2) NVIDIA logo, horizontal lockup: a green (#76B900) eye-shaped emblem to the left of the black bold wordmark NVIDIA, placed at about 65.4 percent from the left and 64.2 percent from the top, with the emblem about 7.0 by 8.2 percent of the slide and the wordmark about 16.2 by 5.4 percent. A small NVIDIA logo (green emblem and black wordmark) sits in the bottom-left footer at about 0.9 percent from the left and 94.8 percent from the top. A grey horizontal rule about 1 pixel thick (#8C8C8C estimated) runs at about 28.9 percent from the top across from the left edge to about 47 percent of the width. A green vertical bar (#76B900 estimated) runs at the bottom-left edge from about 94.8 to 100 percent of the height.

Slide 15

NVIDIA

Visuals: Full-bleed abstract 3D illustration: smooth, glossy lime-green stacked curved bands sweep from the lower-left to the right edge, forming nested arch-like rings. The rings are lit from above, with soft highlights along their edges. Their centre opens into dark green recesses on the right. Thin pale light streaks fan out from the upper-right toward the top-left, fading into white. The composition is wide and diagonal, with the artwork cropped at the right and bottom edges. Dominant colours are white #FFFFFF, lime green #8CC800 and deep green #1E4D00 in the shadows. One logo: the NVIDIA logo at top-left, a green eye-shaped emblem in a square (#76B900) beside a black bold geometric sans-serif wordmark 'NVIDIA' (#000000). A second small image sits at the bottom-right, about 4.8% wide and 3.4% high; it is not clearly distinguishable in the rendering and appears to blend into the green artwork.

Download this deck

The hosted PDF of every library deck is free for readers of the SlideDeck Co newsletter. Join with your email and the viewer and download appear here.

Preparing a results presentation or business update of your own? The SlideDeck Co Standard Deck is a slide library for business reviews, updates and plans.

More Reporting decks

Deck reviews in your inbox.

See what works in real company decks, what we’d change, and get the latest SDC releases.

Expanded slide preview