Placement, VPC, and dedicated capacity

A capacity provider is the placement boundary for LMI. You define where compute runs (VPC, subnets, security groups) and the instance requirements Lambda can use (for example x86_64 or arm64). A function uses that provider through capacity_provider_config on aws_lambda_function.

Caution

capacity_provider_config cannot be added or removed after the function is first created. Decide whether the function runs on LMI before you create that function. See Publishing & activation.

Placement model

Element	What you decide	Why it matters
VPC placement	Subnets and security groups in your account	Defines network reachability and boundaries
Instance requirements	Architecture and performance constraints	Lets Lambda pick matching current-generation instances
Scaling config	scaling_mode, max_vcpu_count, and related limits	Caps fleet growth for this provider

You manage provider configuration; Lambda manages how that configuration is realized as running capacity.

Supported instance families

Family	x86_64 (Intel & AMD)	arm64 (Graviton)
C (compute optimized)	C5, C6, C7	C6G, C7G, C8G
M (general purpose)	M5, M6, M7	M6G, M7G, M8G
R (memory optimized)	R5, R6, R7	R6G, R7G, R8G

• Use .large or larger sizes within these families.
• Default: Lambda selects a suitable current-generation type from the allowed set for your architecture and function configuration.
• Override: narrow the pool with allowed or excluded instance types on the capacity provider (only one of those lists applies).

Note

Further reading: SKU-level availability is region-specific. On AWS Lambda pricing, choose your Region, open Lambda Managed Instances, and use the Management Fees section as the enumerated list. For API fields, see Capacity providers; for how overrides interact with scaling, see Scaling — Instance type selection.

Fully managed (what AWS operates)

With LMI, you do not run the fleet like normal EC2 operations. You configure placement and limits; Lambda runs lifecycle, patching, routing, and managed scaling behavior for the provider fleet.

For quota-sensitive and API-field details, use Lambda Managed Instances as the product reference.

See also

Tenancy & isolation Capacity provider as a security boundary; isolation and trust models.

IAM roles Execution identity for your code versus operator permissions Lambda uses to run the fleet.

Publishing & activation When managed instances and execution environments appear after you publish a version.