Vertex AI

GCP Vertex AI — commitment brief

Data sourced: March 2026. Verify current figures at the Google Cloud Pricing Calculator.

Coverage summary

Vertex AI workloads that consume Compute Engine VM SKUs — including online prediction, batch prediction, and training jobs — are covered by compute flexible CUDs (the same spend-based commitment instrument used for Compute Engine and GKE). When Vertex AI predictions run on Compute Engine-backed resources, those VM costs are eligible for the 28% (1-yr) or 46% (3-yr) compute flexible CUD discount. Certain Vertex AI SKUs such as Vertex AI Search, Vision AI, and managed dataset storage are outside the compute flexible CUD scope and are billed at on-demand rates.

What is covered: Vertex AI online prediction and batch prediction workloads running on Compute Engine VM SKUs (billed as GCE SKUs with a vertex-ai-online-prediction label), Vertex AI training jobs using Compute Engine instances, and Vertex AI Workbench notebook instances. The compute flexible CUD discount applies automatically to the Compute Engine compute portion of these resources.

What is not covered: Vertex AI Search and Conversation, Vertex AI Vision, managed dataset storage, the Vertex AI Management Fee SKU (billed separately at on-demand rates), Model Garden API calls (e.g., Gemini API), and any Vertex AI feature that does not consume Compute Engine VM SKUs.

CUD types

Vertex AI does not have a dedicated Vertex AI–specific CUD product. Coverage is provided through the compute flexible CUD, which applies to Compute Engine VM usage across Compute Engine, GKE, Cloud Run, and Vertex AI within the same billing account. The compute flexible CUD offers 28% off on 1-yr terms and 46% off on 3-yr terms. Vertex AI Management Fee SKUs are billed separately and are not discounted by CUDs.

Machine type / SKU coverage

SKU	Compute Flex CUD	1-yr discount	3-yr discount	Notes
Online prediction (Compute Engine–backed, e.g., n1, n2, c2 node types)	✅ Yes	~28%	~46%	GCE SKU with vertex-ai label; CUD applies to compute portion
Batch prediction (Compute Engine–backed)	✅ Yes	~28%	~46%	Compute portion eligible
Training jobs (Compute Engine–backed)	✅ Yes	~28%	~46%	Compute portion eligible
Vertex AI Workbench (Compute Engine VM)	✅ Yes	~28%	~46%	Underlying VM eligible for flex CUD
Vertex AI Management Fee SKU	❌ No	—	—	Separate billing SKU; not eligible for CUDs
Vertex AI Search / Conversation	❌ No	—	—	Serverless product; no CUD coverage
Vertex AI Vision	❌ No	—	—	Serverless product; no CUD coverage
Model Garden API (Gemini, etc.)	❌ No	—	—	Per-token / per-request billing; no CUD
Managed dataset storage	❌ No	—	—	Storage charges not covered

Regional availability

Vertex AI is available in 40+ regions globally for prediction and training workloads. The compute flexible CUD that covers Compute Engine-backed Vertex AI workloads is a billing-account-level commitment with no Vertex AI-specific regional restrictions — it applies automatically wherever those VM SKUs are consumed. Not all Vertex AI features are available in all regions; feature availability varies by region and should be verified for your specific use case (AutoML, custom training, batch prediction, etc.).

Notable restrictions: GPU-backed prediction and training workloads on A2 or A3 machines are limited to the zones where those machine types are available. CUD coverage for these workloads follows the same zone-level restrictions as the underlying machine type.

⚠️ CUD availability varies by machine type and region. Always verify at GCP regions and zones before purchasing.

Archera

Google Cloud Vertex AI is within Archera's commitment management scope for workloads backed by Compute Engine resources. Archera manages Vertex AI coverage through compute flexible CUDs — automatically sizing commitments across your combined Vertex AI, GKE, and Compute Engine footprint, monitoring utilization, and wrapping commitments in a GRI/GSP to eliminate downside risk on over-commitment.

---

Sources

• Vertex AI pricing

• Compute flexible committed use discounts (blog)

• Expanded coverage for Compute Flex CUDs

• Spend-based committed use discounts overview

⚠️ Discount percentages are approximate. Coverage of specific Vertex AI SKUs under compute flexible CUDs depends on how the underlying compute is billed. Always verify with the Google Cloud Pricing Calculator.