Neocloud GPU capacity (AI factory build-and-lease) is tracked as a supplier bottleneck across Physical AI, primarily within AI Factory & Cloud Training Infrastructure. The current investment map connects it to 1 public stock, including AMD.
Neocloud GPU capacity (AI factory build-and-lease): technology bottleneck and company exposure
Neocloud GPU capacity (AI factory build-and-lease) is the operator layer of the AI factory: companies procure NVIDIA racks such as GB300 NVL72, HGX B300/B200, and H100/H200, add high-density power and liquid cooling, and lease dedicated single-tenant clusters to AI labs on contracted capacity.
Read full summaryClose summary
This note covers the business—utilisation, placement economics, and take-or-pay contracts—rather than the component parts.
What is the Neocloud GPU capacity (AI factory build-and-lease) bottleneck?
Neocloud GPU capacity (AI factory build-and-lease) matters because Physical AI can scale only as quickly as its limiting infrastructure. Its role across AI Factory & Cloud Training Infrastructure can constrain how reliably intelligence is trained, connected, sensed, manufactured or translated into movement.
Research build in progressSource review pending
Tier01 02 03 04
The full Neocloud GPU capacity (AI factory build-and-lease) research brief is being built.
PXS Research has mapped this technology into Physical AI and the AI Factory & Cloud Training Infrastructure layer, but the source note is not yet ready to publish as finished research. We are validating the engineering bottleneck, adoption evidence, investable supply chain and primary sources before assigning durable conviction.
Define the constraint
Explain the physics, engineering or regulatory bottleneck in plain language.
Validate adoption
Separate measurable deployment evidence from narrative and forecasts.
Map public exposure
Identify the public companies supplying the constraint and how directly they benefit.
Source the thesis
Attach primary filings, research papers and credible industry references.
01
Mapped companies
Public and private companies referenced in this research, linked to their canonical PXS Research profiles.
| Company | Market | Research role | Current state |
|---|---|---|---|
| Lambda, Inc. | Private | Referenced in this research | Company profile |
| TensorWave | Private | Referenced in this research | Company profile |
| Parasail | Private | Referenced in this research | Company profile |
| Advanced Micro Devices, Inc. (AMD) | Public | AI GPU accelerators (MI300/MI400) | POOL · reviewed |
Primary and supporting material
Sources
4 cited sources used to define and evaluate this capability.