Option D is correct because Amazon DynamoDB is a serverless NoSQL database that is well suited for semi-structured data, and DynamoDB Accelerator (DAX) is designed to improve read performance for read-heavy applications. AWS documentation states that DAX can improve DynamoDB performance from milliseconds to microseconds, and the DynamoDB guide says DAX is most effective for applications with a high read-to-write ratio, which exactly matches this workload of infrequent writes and frequent reads.

Option A and B are incorrect because Amazon S3 and Iceberg are object storage and table-format solutions, not low-latency transactional databases for millisecond retrieval. Option C is not serverless and requires managing a database cluster, even if optimized reads improve performance. By contrast, DynamoDB is fully managed and serverless, and DAX adds the lowest-latency read path for hot data. The uploaded study guide also identifies DynamoDB as the right fit for dynamic or semi-structured data that requires low-latency access at scale.