Multicloud AI & ERP / Cloud Migration, Integration & AI Operations
Why it matters
Google reorganized its AI platform around agents this year. The capabilities teams relied on continue, but under new names, alongside new governance features and next-generation TPUs.
How it works
The big themes
Agent Platform replaces the Vertex AI name (announced April 22, 2026): Vertex AI Studio is now Agent Studio, Agent Engine is now Agent Runtime and Model Garden keeps its name. Agent governance arrives with Agent Identity, Agent Gateway and Agent Registry. TPU 8 splits into 8t for training and 8i for inference.
At a glance
| Build | Govern | Run | |
|---|---|---|---|
| Headline release | Gemini Enterprise Agent Platform | Agent Identity and Agent Gateway | TPU 8t and 8i (announced) |
| Renamed | Agent Studio (was Vertex AI Studio) | Google SecOps (was Chronicle) | Cloud Run functions (was Cloud Functions) |
| Building blocks | ADK, Agent Runtime | Agent Registry, Model Armor | BigQuery with Gemini |
| For employees | Gemini Enterprise | Agent Observability | AI Hypercomputer |
What stands out
Build
Gemini Enterprise Agent Platform unifies models, tools and agents: build with the open-source ADK, deploy to Agent Runtime and use Agent Studio for prompts, with A2A and MCP for interoperability.
Govern
Agent Identity gives each agent its own identity, Agent Gateway controls agent traffic and Agent Registry lists approved agents, with Model Armor screening prompts and responses.
Run
TPU 8t and 8i target training and inference separately, BigQuery adds AI functions inside the warehouse and Gemini Enterprise brings agents and search to every employee.
Design decisions that matter
- 01Update names in docs and code
Avoid confusion across teams.
- 02Adopt Agent Identity early
Before agents multiply.
- 03Plan TPU 8 benchmarks
For training and serving cost.
- 04Use ADK for new agents
Portable and open source.
Common mistakes we see
- Assuming features disappeared. Most moved under new names.
- Agents on shared service accounts. Agent Identity solves this.
- Waiting for TPU 8 to plan. Benchmark current TPUs now.
A practical 30-day plan
- Week 1
Map renamed services in your estate.
- Week 2
Pilot an ADK agent on Agent Runtime.
- Week 3
Set up Agent Identity and Gateway policies.
- Week 4
Benchmark TPUs for your main model.
How NTS helps
NTS helps Google Cloud clients adopt Gemini Enterprise Agent Platform, set up agent governance with Agent Identity and Gateway, and plan TPU-based training and inference.
What you get
- A briefing on the latest Google Cloud AI services
- Use-case fit for Gemini, Vertex AI and agents
- Pilot projects on new capabilities
- An adoption roadmap
Typical outcomes
- Early, safe adoption of Google AI innovations
- Clear priorities for AI investment
- Competitive advantage from new capabilities
How we work with you
- 1Assess
A focused review of your Google Cloud AI adoption across AWS, Azure and Google Cloud. We look for the gaps we find most often: assuming features disappeared, agents on shared service accounts and waiting for TPU 8 to plan.
- 2Design
A target architecture and roadmap built on the decisions that matter most: update names in docs and code, adopt Agent Identity early, plan TPU 8 benchmarks and use ADK for new agents.
- 3Build
Hands-on delivery alongside your team. We map renamed services in your estate, pilot an ADK agent on Agent Runtime, set up Agent Identity and Gateway policies, and finally benchmark TPUs for your main model.
- 4Run and improve
We brief your team every quarter on new Google Cloud AI services and update your roadmap.
Why teams choose NTS for Google Cloud AI adoption
- Google Cloud AI expertise
We know Gemini, Vertex AI and BigQuery AI in depth.
- Hype filtered out
We separate production-ready services from previews.
- Hands-on proof
Pilots validate value quickly.
Book a free 30-minute review
Talk to an NTS architect about your Google Cloud AI adoption on AWS, Azure and Google Cloud.