AMD
Primary GPU challenger with Instinct MI300/MI350 and the ROCm software stack
AMD is Nvidia's most credible data-center GPU competitor with its Instinct line: MI300X/MI300A shipped in volume through 2024-2025, MI325X and the CDNA4-based MI350/MI355X launched in 2025, with MI400 and rack-scale 'Helios' systems targeted for 2026. AMD competes on HBM capacity, price/performance, and an open ROCm software stack versus CUDA. It is fabless on TSMC (N5/N3 chiplets) and also designs EPYC x86 server CPUs that pair with AI accelerators.
AI GPU line
Instinct MI300X/MI325X, MI350/MI355X (CDNA4)
2026 roadmap
MI400 + 'Helios' rack-scale
Software
ROCm (open) vs CUDA
Also designs
EPYC x86 server CPUs
How it fits the stack
AMD with what it depends on (above) and what it feeds (below). The figure renders as a crawlable diagram and upgrades to an interactive 3D graph as it scrolls into view.
AMD in the AI stack. AMD with its immediate upstream dependencies (top) and downstream dependents (bottom) in the AI value chain. Hover a node in 3D, or read the full relationships below.
Graph data (text) — 18 entities, 18 relationships
- AMD —supplies→ AT&S
- AMD —used by→ Cadence Design Systems
- AMD —uses→ High-Bandwidth Memory (HBM)
- AMD —used by→ High-Bandwidth Memory (HBM)
- AMD —supplies→ Samsung Electronics (Memory)
- AMD —supplies→ SK Hynix
- AMD —used by→ Synopsys
- Amazon Web Services (AWS) —uses→ AMD
- Google Cloud (GCP) —uses→ AMD
- HUMAIN —depends on→ AMD
- Hygon (Haiguang) —depends on→ AMD
- Meta (AI Infrastructure) —depends on→ AMD
- Microsoft Azure —uses→ AMD
- OpenAI —supplies→ AMD
- AMD —competes with→ Arm Holdings
- AMD —competes with→ CUDA / software moat
- AMD —competes with→ Intel
- AMD —competes with→ Nvidia
Depends on ↑ · 13
Feeds ↓ · 8
Context — capital, rivals, policy · · 8