Domain 1 of 4 · Chapter 3 of 4

Gen AI Landscape

Unlock the complete study guide + 1,040 practice questions across 16 full exams.

Bundled into the existing Generative AI Leader premium course — no separate purchase.

14-day money-back guarantee — no questions asked.

Included in this chapter:

  • The gen AI landscape as a five-layer stack
  • A tour of the five layers
  • The decision at each layer: how far down?
  • Exam-pattern recognition

The five layers on the decisions that separate them

Decision axisInfrastructureModelsPlatformsAgentsApplications
What it providesCompute to train and serveFoundation modelsManaged build-and-deploy environmentModel plus tools and orchestrationFinished end-user experience
Google Cloud exampleAI Hypercomputer (TPUs, GPUs)Gemini, Gemma, Imagen, VeoVertex AI, Model GardenVertex AI Agent BuilderGemini app, Gemini for Google Workspace
Primary ownerPlatform or infra teamML or developer teamML or developer teamDeveloper or solution teamBusiness or line-of-business user
Control and customizationHighestHighMediumMediumLowest
Time to valueSlowestSlowMediumFasterFastest
Operating burdenHighestHighMediumLowLowest (fully managed)

Decision tree

Ready-made app meets it?ApplicationsGemini for Google WorkspaceYesNoMust act with tools?AgentsVertex AI Agent BuilderYesNoCustomize/self-host model?PlatformsVertex AI, Model GardenNoYesCost driven by scale?ModelsFine-tune or self-host GemmaNoInfrastructureAI HypercomputerYes
Cheat sheet

Cheat sheet

  • The gen AI landscape is a five-layer stack
  • Operate at the highest layer that meets the requirement
  • Higher layers trade control for speed
  • Descend the stack only when a concrete requirement forces it
  • AI Hypercomputer is Google Cloud's Infrastructure layer
  • Gemini, Gemma, Imagen, and Veo make up the Models layer
  • Vertex AI is the Platform layer, not a model
  • An agent is a model wired to tools and orchestration
  • Applications are finished experiences you buy, not build
  • Google Cloud sells at every layer of the stack
  • Self-host an open model when data cannot go to a shared model
  • Adopt top-down: start at Applications and descend as needs grow
  • Sort vendor offerings by layer before comparing them

Unlock with Premium — includes all practice exams and the complete study guide.

Also tested in

References

  1. AI Hypercomputer overview
  2. Introduction to Vertex AI
  3. AI Hypercomputer consumption models
  4. Google models on Vertex AI
  5. Gemma open models
  6. Explore models in Model Garden
  7. Vertex AI Agent Builder
  8. Conversational Agents (Dialogflow CX): agents
  9. Gemini app
  10. Gemini for Google Workspace