Supply chain / Accelerators
NVIDIA (Blackwell data centre GPUs)
NVIDIA designs the GPUs and rack-scale systems that many cloud and neocloud AI clusters are built around. Its Blackwell generation joins two large dies in one package and is sold as boards, racks and DGX systems.
- Type
- Company
- Where it sits
- Designer of the accelerator hardware you are most likely to see listed by GPU clouds, so its product names are the usual vocabulary for capacity listings.
- Constraint
- Blackwell parts are built on a TSMC process (NVIDIA names 4NP) and depend on the foundry, packaging and memory layers below.
What we checked
- NVIDIA states that Blackwell GPUs use two reticle-limited dies joined by a 10 TB/s chip-to-chip link and are made on TSMC 4NP. (source)
- NVIDIA lists GB200 NVL72 and GB300 NVL72 rack-scale systems, HGX B300 and DGX SuperPOD among its Blackwell-based products. (source)
Written by ComputeFinder from public pages on 2 October 2026. Company statements are the company's own, not independently checked, and reported items are marked as such. Not yet reviewed line by line by a second person. If something is wrong, tell us.