AMD · AI and HPC accelerators · 3D corporate launch video

Corporate Launch Video for AMD Instinct MI300 Series: 2 Chips, 1 Platform, Every Claim Footnoted

AMD needed one launch video for 2 new chips, an 8-accelerator platform and a software release, each with claims that data center buyers would check. Advids built the AMD Instinct MI300 Series corporate launch video in 3D: each product gets its own named block, the camera goes inside the package to the memory, and every superlative carries an endnote marker that the last frames spell out.

Last updated Oct. 20, 2026

AMD · Instinct MI300 Series

Client: AMD
Product: AMD Instinct MI300 Series: MI300A APU, MI300X accelerator, MI300X Platform and ROCm 6 software
Made for: The MI300 Series launch on AMD's YouTube channel, Dec. 6, 2023
Format: 3D corporate launch video with on-screen text
Built from: Converted CAD models of the chips and the platform board
Length: 1:39
Production timeline: 3 to 4 weeks
Budget: $1,200 to $2,000

The problem: 4 launches in one, and every claim gets checked

AMD announced 2 chips at once. MI300A is, in AMD's words, "the world's first data center APU to integrate CPU and GPU on a single package". MI300X is pitched as "the world's most advanced accelerator for generative AI". On top came an 8-accelerator platform and the ROCm 6 software release.

Data center and AI teams read spec sheets line by line and compare them against the market leader. The video had to keep 4 names apart, put each spec on the part that delivers it, and make every superlative traceable. A SaaS company launching a product suite with benchmark claims faces the same test.

How the launch video gives each product its own block, scene by scene

Every claim is on-screen text, so the video works with the sound off. Select a time to play that scene above.

Time	On screen	What it does
	Particle waves under "AI is accelerating", then "And this is what is accelerating AI" and "The AMD Instinct™ MI300 Series".	Sets the market shift first, then names the answer.
	"MI300A" stays as the header while the line under it changes: "The world's first data center APU to integrate CPU and GPU on a single package" over the package, "Sharing 128 GB HBM3 Memory with 5.3 TB/s bandwidth" as the memory stacks light up, then "Designed to power supercomputers that advance research at exascale" in a server corridor.	Gives the APU its claim, its spec and its use, in that order.
	"MI300X" takes over the header: "The world's most advanced accelerator for generative AI" with "MI300-26. See Endnote 1", then "192 GB HBM3 Memory with 5.3 TB/s bandwidth", then "Purpose-built for massive AI workloads", "Run large AI models directly in memory", "Reduce the number of GPUs needed" as data streams through the die.	Turns memory size into the buyer's gain: fewer GPUs per model.
	One accelerator lowers onto a board, then 2, then all 8: "AMD Instinct MI300X Platform", "Rapidly deploy eight MI300X accelerators within existing infrastructure", "Scale your AI training and inference services to larger models and support more jobs per server".	Moves from the chip to what a data center would install.
	Gold and blue energy lines spread from the platform: "AMD ROCm™ 6 Software", "Delivering new capabilities for AI and HPC", then "Accelerating AI performance by 8x on MI300X" with "MI300-33. See Endnote 2".	Ties the software release to a measured gain on the hardware.
	"MI300 and ROCm together", "Delivered through an open ecosystem to accelerate solutions for the toughest challenges facing the world today". Particles form a head for "To discover cures", a planet and galaxy for "To monitor and predict climate change", and a rover for "To drive us further, faster".	Names 3 uses a buyer can repeat to their board.
	"This is the MI300 series" over a server, "This is how to accelerate AI" in a data center aisle, then the AMD logo with "together we advance_HPC" and "together we advance_AI".	Answers the opening line and closes on the brand.
	Endnotes 1 and 2: the basis for the "most advanced" claim and the test setup behind the 8x figure, plus trademark lines.	Backs both starred claims in full.

Storyboard: the frames that carry the AMD Instinct MI300 corporate launch video

Key frames from the finished video, in order. Select a time to play that moment above.

The series is named only after the market shift is set.
MI300A gets its first-of-its-kind claim over the package itself.
The memory spec sits on the memory: the HBM3 stacks light up beside the die.
The biggest claim arrives with its endnote marker on the same frame.
Memory size becomes a gain the buyer counts: fewer GPUs.
8 accelerators on one board: the unit a data center actually buys.
The software claim is measured on the hardware and starred.
The first of 3 outcomes, drawn in the same particles as the opening.
The close answers the opening line about what is accelerating AI.
The endnotes spell out what each starred claim is based on.

4 decisions behind the AMD Instinct MI300 corporate launch video

Decision 1

Hold the product name as a header

"MI300A" stays fixed top left from 0:10 to 0:21 while 3 lines change under it. "MI300X" does the same from 0:24 to 0:37, so the 2 chips never blur together.

Copy it: Launching more than one product? Keep each name on screen for its whole block and change only the line under it.

Decision 2

Star every superlative

"The world's most advanced accelerator for generative AI*" at 0:26 and "Accelerating AI performance by 8x on MI300X*" at 1:00 each carry a marker on the same frame. The endnotes at 1:36 give the basis and the test setup.

Copy it: Put the footnote marker on screen with the claim and the full test setup at the end, so legal review does not cut the claim.

Decision 3

Point the memory spec at the memory

At 0:16 the camera is inside the package and the HBM3 stacks light up with blue edges as "Sharing 128 GB HBM3 Memory with 5.3 TB/s bandwidth" appears.

Copy it: When a spec belongs to one part, light that part the moment the number appears.

Decision 4

Count the platform up from 1 to 8

At 0:40 a single accelerator lowers onto the board, at 0:46 there are 2, and by 0:48 all 8 are in place under "AMD Instinct MI300X Platform".

Copy it: If the selling unit is a set, build it piece by piece on screen so the viewer counts it.

From AMD CAD files to a 1:39 launch reveal

CAD conversion and modeling. The MI300A package, the MI300X chip and the 8-accelerator platform board were converted and modeled for close-ups.
One block per product. The script was split into named blocks: series, MI300A, MI300X, platform, ROCm 6, then outcomes.
Texturing and lighting. Die, memory stacks and heat sinks were textured and lit so the parts each spec refers to read at a glance.
Object animation. Camera moves into each package, memory stacks light up, and accelerators drop onto the board 1 by 1.
Particle and energy effects. Blue and gold particles carry the opening, the outcome shapes and the ROCm energy lines.
Render, composite and text. Photorealistic renders were composited with the headers, spec lines, footnote markers and endnotes.
Sound and color. Sound design and color enhancement finished the 1:39 cut.

Scope of work:

CAD model conversion
3D modeling
High-resolution texturing
Realistic lighting
Object animation
Photorealistic rendering
Visual effects
Compositing
Sound design
Color enhancement

Production took 3 to 4 weeks on a budget of $1,200 to $2,000. See the 3D video production service for how a project like this is scoped.

Questions about the AMD Instinct MI300 corporate launch video

How does the AMD launch video keep MI300A and MI300X apart?

Each chip gets its own block with its name fixed as the header: "MI300A" from 0:10 to 0:21 and "MI300X" from 0:24 to 0:37. Only the line under the name changes.

What does the video say about MI300X memory?

At 0:30 "192 GB HBM3 Memory with 5.3 TB/s bandwidth", then at 0:34 "Run large AI models directly in memory" and "Reduce the number of GPUs needed". The memory claim leads straight to a cost the buyer can count.

How does the video back up "the world's most advanced accelerator for generative AI"?

The claim carries "*MI300-26. See Endnote 1" on screen at 0:26. The endnote at 1:36 defines it as of November 29, 2023, by process technology, "153 billion transistors and 192 GB HBM3 memory", compared with Nvidia H100.

Where does the ROCm 6 "8x" figure come from?

Endnote 2 (MI300-33) at 1:36: text generated with Llama2-70b chat in AMD internal testing as of 11/17/2023, comparing 4x MI300X on ROCm 6.0 pre-release with 4x MI250 on ROCm 5.4.3.

Which outcomes does the MI300 video close on?

3, each drawn in particles: a head for "To discover cures" at 1:08, a planet and galaxy for "To monitor and predict climate change" at 1:12, and a rover for "To drive us further, faster" at 1:16. Then "This is the MI300 series" and "This is how to accelerate AI".

What did the AMD corporate launch video take to make?

CAD model conversion, 3D modeling, texturing, lighting, object animation, photorealistic rendering, visual effects, compositing, sound design and color enhancement. Production took 3 to 4 weeks on a budget of $1,200 to $2,000.

What Advids clients say

"We had a great experience with ADVIDS. We required a technical 3D video to be produced. The project ran smoothly with excellent communication."Emily Gray, Quality Manager at Pacific Hoseflex