In terms of measuring the effectiveness of the project, I mostly work only in terms of the sizing of the infra piece for AI workloads. What exactly, what type of AI workloads the customer is having? And whether the primary workload is training-heavy or inferencing, what AI models they have? And in terms of performance, we just mainly ask in terms of what is the target for that token latencies. When you talk about AI, it is all about tokens. What are the expected average and peak tokens? That is the kind of sizing I understand. Regarding whether my clients have NVIDIA AI Enterprise on cloud or on-premise, I can say it is a mix. It is mixed because it depends on the usage of your AI workload. If it is frequent, where people are trying to access, upload, and download, then definitely on-prem will be ideal, where they will go with NVIDIA AI Enterprise. And if it is not that much, then they will go with NVIDIA AI Enterprise from AWS or any cloud where you are able to spin the GPUs of NVIDIA in the cloud. I am not much into AWS on the cloud part. My overall rating for NVIDIA AI Enterprise is eight out of ten.
In terms of measuring the effectiveness of the project, I mostly work only in terms of the sizing of the infra piece for AI workloads. What exactly, what type of AI workloads the customer is having? And whether the primary workload is training-heavy or inferencing, what AI models they have? And in terms of performance, we just mainly ask in terms of what is the target for that token latencies. When you talk about AI, it is all about tokens. What are the expected average and peak tokens? That is the kind of sizing I understand. Regarding whether my clients have NVIDIA AI Enterprise on cloud or on-premise, I can say it is a mix. It is mixed because it depends on the usage of your AI workload. If it is frequent, where people are trying to access, upload, and download, then definitely on-prem will be ideal, where they will go with NVIDIA AI Enterprise. And if it is not that much, then they will go with NVIDIA AI Enterprise from AWS or any cloud where you are able to spin the GPUs of NVIDIA in the cloud. I am not much into AWS on the cloud part. My overall rating for NVIDIA AI Enterprise is eight out of ten.