A research group has 2 million archived scientific PDFs that need a one-paragraph summary generated for each. The job will run overnight with no user waiting on results, and the team is choosing a deployment approach that minimizes cost since immediate response time does not matter. Which deployment choice fits this workload?

The correct answer is C. A batch endpoint is built exactly for this pattern: a large, known volume of work with no live user waiting on each response, so requests can be queued and processed at a pace that favors lower cost over speed, which matches an overnight job with 2 million documents and no latency requirement. Option A is incorrect because a real-time endpoint is optimized for immediate response to each request and is billed and provisioned for that responsiveness, which is unnecessary expense when nothing is waiting on the answer. Option B is incorrect because a dedicated throughput unit is meant to guarantee consistent low-latency capacity for steady interactive traffic; committing to that capacity model for an overnight, non-interactive job pays for responsiveness the task does not need. Option D is incorrect because autoscaling for second-by-second spikes solves a bursty, unpredictable interactive-traffic problem, not a single large known batch of offline work, so it adds complexity without addressing the actual requirement of processing a fixed dataset cheaply.