The years of sanctions and subsequent circumventions remind me of Wiley E. Coyotee trying to get the roadrunner antics. This may create some alternative to cuda, which would be a pretty great outcome.
For context, these are cards for ML on rack servers. A100 is the current line, H100 will be the next line.
Does anyone have a feel for the supply chain of the A100, especially as it relates to export to China? Just wondering out loud how much they really need to ‘import’ it so much as … let it stay there.
EDIT: Partially answered by the second paragraph of the filing if I had actually read it first: “The new license requirement may impact the Company’s ability to complete its development of H100 in a timely manner or support existing customers of A100 and may require the Company to transition certain operations out of China”.
I read this as saying they were doing material development or manufacturing in China and do believe they need to shift this out of China to comply with the requirement to prevent export, not selling directly to Chinese customers would not be enough.
I read it the same way. Note that NVIDIA made another filing to state that US government authorized exemptions for the purpose of H100 development. In fact, the filing states that it happened after the previous filing on 2022-08-31, but the filing itself was made on 2022-09-01, so it must happened within a single day.
(Note that in SEC Form 8-K, “Date of Report” at the top is not the date of filing. To see the date of filing, scroll to the bottom.)