Nvidia Denies Claims of Supply Constraints for Cloud Access to H100, H200, and Blackwell GPUs

Nvidia denies supply constraints claims, stating there are no limitations for cloud access to its H100, H200, or Blackwell GPUs. The company emphasizes that reports of such constraints are "categorically false."

Nvidia (NASDAQ: NVDA) has responded to recent media reports, categorically denying any supply constraints for cloud access to its H100, H200, and Blackwell GPUs. The company issued a statement asserting that these claims are "categorically false" and that it can fulfill all orders without delay.

Nvidia emphasized that its cloud partners can rent every H100/H200 GPU that the company has online, and they are also able to satisfy new orders without any delays. This clarification comes amidst reports that the company might be prioritizing sales of H20 GPUs, leading to supply constraints for other AI GPUs in the lineup.

The H20 chip, designed as a US regulations-compliant variant for export to China, has been in high demand. Despite its cut-down nature, it offers more AI horsepower and capabilities than Chinese-made chips like the Ascend 920 series. However, recent reports suggesting the presence of backdoors, spyware, or kill switches within the H20 chips were also rejected by Nvidia.

Nvidia's confidence in its supply chain and partnerships is evident in its recent earnings, where the AI segment reported a record $41.096 billion in revenue. The company is also moving to a yearly cadence for producing high-end AI GPUs, with Rubin launching in 2026, Rubin Ultra in 2027, and Feynman in 2028.

References:
[1] https://wccftech.com/nvidia-debunks-h100-h200-ai-sold-out-supply-constrained-h20-has-no-impact-on-other-products/
[2] https://seekingalpha.com/news/4491308-nvidia-says-supply-constraint-claims-are-categorically-false