Amazon SageMaker AI

Machine Learning

Fully managed platform for building, training, and deploying your own ML models.

Verified against AWS documentation · 2026-08-26

Key facts

  1. 01Covers the whole ML lifecycle in one service: label, build in notebooks, train, tune, deploy, monitor
  2. 02You bring the data and the model — AWS brings managed compute. It is not a pre-trained API
  3. 03Four inference options: real-time (persistent endpoint), serverless (intermittent traffic), asynchronous (large payloads, scales to zero), batch transform (offline bulk)
  4. 04Limits: real-time 25 MB / 60s, serverless 4 MB / 60s, async 1 GB / 1 hour, batch transform runs GB-scale datasets for days
  5. 05Naming: the original service is now SageMaker AI. "Amazon SageMaker" now refers to the wider Unified Studio platform that wraps it

Exam triggers

Phrases and scenarios that point here.

  • "Custom model", "train on our own data", "our own algorithm" → SageMaker, not Rekognition/Comprehend/Transcribe
  • "No ML expertise", "business analyst", "no code" → Canvas; "automatically find the best model" → Autopilot
  • Endpoint sits idle most of the day → Serverless Inference; large payload or minutes-long inference → Asynchronous Inference; no endpoint needed at all → Batch Transform
  • "Reduce training cost" → Managed Spot Training (up to 90% off)
  • "Accuracy degrades over time" → Model Monitor; "explain the prediction" or "detect bias" → Clarify

Not this service

  • Off-the-shelf vision, speech, or text prediction → Rekognition, Transcribe, Comprehend
  • Call a foundation model by API and pay per token → Bedrock
  • Train and predict without leaving SQL → Redshift ML or Athena ML

Use cases

  1. 01Forecasting demand from five years of your own POS data, where no off-the-shelf model has ever seen your SKUs
  2. 02Hosting a fraud-scoring model behind a real-time endpoint that checkout calls inline on every transaction
  3. 03Async inference for a claims processor chewing through 200 MB scanned PDF bundles, scaling to zero overnight
  4. 04Batch transform to re-score the entire customer base for churn once a month — nothing left running between runs
  5. 05Managed Spot Training for overnight hyperparameter sweeps, where an interruption costs patience but not money

How it's charged

  1. 01No platform fee — you pay ML instance-hours per feature, billed per second; SageMaker Savings Plans discount that rate for a 1- or 3-year $/hour commitment
  2. 02Real-time endpoints bill every second they are InService, traffic or not — the single most common source of surprise SageMaker bills
  3. 03Training, processing, and batch transform jobs bill only for job duration; Managed Spot Training cuts that by up to 90% in exchange for interruptions
  4. 04Serverless and asynchronous inference scale to zero: serverless bills compute-time plus data processed, async bills nothing while its queue is empty
  5. 05Storage bills separately and quietly — notebook and Studio EBS volumes keep charging while the instance is stopped, plus S3 for datasets and model artifacts

Commonly confused with

Amazon BedrockMachine LearningAmazon RekognitionMachine LearningAmazon ComprehendMachine Learning

References