Question

A media company is launching a new AI-powered article summarization feature. During the first two weeks, marketing expects unpredictable traffic spikes as the launch is promoted, and the team wants guaranteed throughput without on-demand throttling. However, leadership is unwilling to sign a 1-month or 6-month provisioned throughput commitment because the long-term traffic pattern is completely unknown and may drop sharply after launch. Which approach best satisfies both the guaranteed-throughput and no-long-term-commitment requirements during the launch window?

Accepted Answer

Purchase Provisioned Throughput with no commitment (hourly pricing) for the launch period, then release the model units once traffic patterns stabilize.

Answer

Purchase Provisioned Throughput with a 1-month commitment to get the lowest hourly rate, then cancel it after two weeks.

Answer

Rely solely on on-demand inference and request a service quota increase to handle the spikes.

Answer

Use Batch inference for all summarization requests to avoid throttling during the launch.