The Higher-Level Services: AI, Big Data, IoT

The building blocks covered so far — compute, storage, networking, databases, security — are the foundation that most cloud applications share. But each major cloud provider also offers a second tier of services that sit on top of that foundation: ready-made AI capabilities, tools for processing enormous datasets, and platforms for connecting thousands of physical devices to the internet.

These higher-level services are where much of the cloud's current growth and hype live. You don't need deep expertise in any of them right now — but you do need a map of the categories, what they're for, and which providers are known for what.

The analogy: a hardware megastore that also rents specialists by the hour. Instead of learning plumbing, electrical work, and carpentry from scratch, you hire the specialist you need, use their expertise for your project, and pay only for the time. The provider has already built the specialty; you consume it. That's the pattern behind all higher-level services.

AI and Machine Learning

Machine learning (ML) is the practice of training a computer model on large amounts of data so it can recognize patterns and make predictions. The trained model can then classify images, transcribe speech, detect spam, suggest products, or do many other tasks that used to require human judgment.

All three providers offer two broad flavors. The first is ready-made AI APIs — services that accept input (a photo, a sentence, an audio file) and return an intelligent output (what's in the photo, the sentiment of the sentence, the transcription of the audio). No data science required; you call the service like any other API. Cloud Vision (Google), Amazon Rekognition (AWS), and Azure AI Vision are examples.

The second flavor is an ML platform — tools for teams that want to train their own models on their own data. You bring the data and the problem; the platform provides the computing power, storage, and workflow tooling. AWS SageMaker and Google Gemini Enterprise Agent Platform (formerly Vertex AI) are the leading examples.

Generative AI — models that produce text, images, or code from a prompt — is the most prominent recent addition. These are the same large language models behind tools like ChatGPT. Cloud providers now offer them as hosted APIs. Azure OpenAI Service hosts OpenAI's models directly. Google's Gemini Enterprise Agent Platform includes Gemini. AWS Bedrock provides access to a range of foundation models. A developer can call these APIs to build a product that uses generative AI without training anything themselves.

Big Data and Analytics

A regular database — even a managed one — is designed for transactional work: look up one user, update one order, insert one record. When the question is instead "analyze all 500 million transactions from last year to find patterns," a different toolset is needed. That toolset is broadly called big data.

Big data services are built to process and analyze enormous volumes of data in parallel — spreading the work across many machines simultaneously to handle scales that would be impractical on a single server. The output is typically a report, a dashboard, or a trained model, not a real-time transaction. Google BigQuery — a managed data warehouse where you can query petabytes of data using standard SQL — is one of the most prominent examples. AWS has Redshift (data warehouse) and Athena (query engine). Azure has Synapse Analytics.

IoT — The Internet of Things

IoT (Internet of Things) refers to physical devices — sensors in a factory, thermostats in a building, GPS trackers on delivery vehicles — that connect to the internet, send data, and sometimes receive instructions. A large organization might manage tens of thousands of such devices simultaneously.

IoT cloud services handle the two hardest parts of this at scale: connecting and authenticating each device securely, and ingesting the constant stream of data those devices generate. Without a managed platform, building this plumbing yourself for thousands of devices would be enormous work. AWS IoT Core, Azure IoT Hub, and Azure IoT Central (Microsoft's managed SaaS IoT offering) are designed to solve exactly this problem. Google Cloud IoT Core was retired in August 2023; Google directs users seeking IoT connectivity to third-party platforms.

The Common Pattern

Across all three categories — AI, big data, IoT — the same logic applies as for managed services generally: the provider has already built the complex, hard-to-maintain infrastructure, and you consume a capability rather than constructing it. A startup can add image recognition to a product by calling an API, without hiring a team of machine learning researchers. A retailer can analyze a year of sales data in minutes without buying a supercomputer. A logistics company can track a fleet of vehicles without building its own data ingestion pipeline.