A company streams IoT data from automobile sensors to Amazon S3 via Amazon Kinesis Data Firehose, producing trillions of S3 objects per year. Each morning the company retrains ML models using the previous 30 days of data; quarterly it uses the previous 12 months for other analysis. Data must be available with minimal delay for up to 1 year, and retained for archival purposes after 1 year. Which storage solution is the most cost-effective?

The correct solution is the most cost-effective because it leverages S3 Standard for immediate access, S3 Standard-IA for data accessed less frequently (after 30 days, covering the 30-day ML model retraining window), and S3 Glacier Deep Archive for long-term archival after 1 year. This aligns with the access patterns: frequent access for 30 days, less frequent for up to 1 year, and archival beyond that. S3 Intelligent-Tiering is less cost-effective here because the access patterns are well-defined, making manual lifecycle policies more precise and potentially cheaper than Intelligent-Tiering's monitoring and auto-tiering overhead. S3 Intelligent-Tiering does not automatically move objects to S3 Glacier Deep Archive; it moves to S3 Glacier Flexible Retrieval (formerly S3 Glacier) or S3 Glacier Instant Retrieval. Using S3 Standard-IA from the start would incur retrieval fees for the daily ML model retraining within the first 30 days, making it more expensive.