## Factors So if you want to estimate the cost of **using** AI, you would need to factor in the following: - Size/type/number of underlying computer - RAM - Cores - GPU RAM - GPU Cores - Time taken (usually in hours) - Type of AI use - Training - Deployment - Prediction/inference - Availability - Dedicated (always available) - On-demand (pay-per-use, but susceptible to surge pricing, may also be slow to classify first image) There are other hidden costs: - Storage - Amount of images/metadata - Frequency of access - Cross-region data access (from data center to data center) - Number of times model is retrained - Idle time ## Options - [Vertex AI - Google](https://cloud.google.com/vertex-ai/pricing) - [AutoML - Google](https://cloud.google.com/vertex-ai/docs/beginner/beginners-guide) - [SageMaker - AWS Calculator](https://calculator.aws/#/createCalculator/SageMaker) - [Azure Calculator](https://azure.microsoft.com/en-us/pricing/calculator/?service=cognitive-services) - [HuggingFace Prices](https://huggingface.co/pricing) - [Lambda AI](https://lambda.ai/pricing) ## References - [Hidden Cost of AI in the Cloud - CloudOptimo](https://www.cloudoptimo.com/blog/the-hidden-cost-of-ai-in-the-cloud/)