The most cost efficient choice is Amazon EC2 Spot Instances. They let you consume spare EC2 capacity at steep discounts and the pipeline can tolerate interruptions and resume from checkpoints so occasional preemptions do not jeopardize completion.

Amazon EC2 Spot Instances provide deep discounts compared with standard pricing and they are ideal when jobs are restartable and time flexible. Because your job finishes in about 75 minutes and can resume from checkpoints a Spot based approach reduces total compute cost while still allowing the job to complete reliably after interruptions.

AWS Lambda is unsuitable because each invocation has a 15 minute maximum runtime and a 75 minute batch cannot finish in a single invocation. Breaking the work into many short functions increases orchestration complexity and often raises cost versus a long running instance for CPU bound batch jobs.

Amazon EC2 On-Demand Instances will run the job without interruption but you pay the full on demand rate for the entire runtime. That makes them more expensive than using spare capacity when the workload can tolerate preemption.

Amazon EC2 Reserved Instances lower cost for steady predictable usage but they require long term commitments and they do not align well with intermittent batch jobs that benefit most from spot pricing.

Full AWS Practitioner Certification Question