Bedrock cross-region inference, explained (2026).
Cross-region inference lets Bedrock serve your request from another AWS Region when one is busy. This page covers geographic vs global profiles, how to turn it on, which models are covered, where your data is processed, the throughput gains, the cost, and when to stay in one Region.
- Cross-region inference routes a Bedrock request to another AWS Region with spare capacity. You call an inference-profile ID instead of a model ID. Geographic profiles (US, EU, APAC) keep processing inside that geography. Global profiles can use any supported commercial Region and, per AWS, cost about 10% less.
- It raises effective throughput and smooths spikes that would throttle one Region. There is no extra routing fee, and the price is based on the Region you call from. Data stays on the AWS network, encrypted in transit, and CloudTrail records which Region processed each request.
- Use a geographic profile when you have data residency rules, a global one when you do not, and a single Region when a rule pins data to one Region. Inference profiles do not support Provisioned Throughput.
IWhat cross-region inference profiles are
One Region has finite on-demand capacity for a model. A spike can hit that ceiling and throttle you. Inference profiles let one request draw on several Regions.
By default you call a model ID in one Region, and every request uses that Region's shared on-demand capacity. Requests beyond its ceiling get a ThrottlingException.
A cross-region inference profile represents a model across a set of Regions. You call the inference-profile ID, and Bedrock routes each request to a Region in the profile with capacity. Your application sees the same API, model and response.
AWS offers two kinds (see the cross-Region inference guide). Geographic profiles (such as US, EU or APAC) route only within that geography. Global profiles route to any supported commercial Region worldwide, and AWS lists them at about 10% lower cost.
Do not confuse this with Provisioned Throughput. Provisioned Throughput reserves capacity in one Region for an hourly fee. Cross-region inference spreads on-demand capacity across Regions at no extra fee. AWS notes that inference profiles do not support Provisioned Throughput, so they are separate tools.
IIHow it works, and how to enable it
Turning it on is mostly a one-identifier change. Two Regions are involved in each request.
- Source Region: The Region your application calls. Pricing is based on this Region, and CloudTrail logs the request here.
- Destination Region: The Region that runs the model, picked by Bedrock from the profile's Regions based on capacity. It may be the source Region or another one. CloudTrail's additionalEventData.inferenceRegion field shows which.
To enable it, pass the inference-profile ID instead of the model ID in InvokeModel or Converse, and grant IAM permission to invoke the model in the profile's Regions. Requests can route to Regions you have not manually enabled in your account. If you use SCPs, they must allow the destination Regions; for global profiles AWS says to allow "aws:RequestedRegion": "unspecified".
Besides the profiles AWS publishes, you can create application inference profiles that wrap a system-defined profile and add your own cost-allocation tags. See the inference profiles guide.
IIIWhich models and Regions support it
Coverage is per model and per profile, and it changes as AWS adds models.
AWS publishes profiles for many high-demand models, including the Anthropic Claude family, Amazon Nova and Meta Llama. Each model may have geographic profiles, a global profile, or both. Check the "Supported Regions and models for inference profiles" page in the Bedrock docs before you design around a specific model.
| Profile type | Where requests can be processed | Cost | Best for |
|---|---|---|---|
| Geographic (US, EU, APAC and others) | Regions inside that geography | Standard pricing | Data residency requirements |
| Global | Any supported commercial Region worldwide | About 10% lower, per AWS | Cost and throughput without location limits |
| Single Region (no profile) | Only the Region you call | Standard pricing | Rules that pin data to one Region |
IVData residency and compliance
Routing a request to another Region means your prompt is processed there. The profile type decides how far it can go.
With a geographic profile, a request is processed only in a Region inside that geography: a US profile stays in US Regions, an EU profile in EU Regions. With a global profile, it can be processed in any supported commercial Region. AWS says data in transit between Regions stays on the AWS network and is encrypted.
Bedrock does not use your prompts or outputs to train the base models, wherever the request runs.
Decision steps: (1) state your obligation precisely: none, geography-level, or Region-level; (2) none → a global profile is fine; geography-level → a geographic profile; Region-level ("data must stay in Frankfurt") → call that one Region and skip cross-region routing for that path; (3) record the choice and check CloudTrail's inference Region field in audits.
VThroughput and availability benefits
The reason to adopt it, where residency allows, is more burst capacity and better resilience at no extra fee.
Higher effective throughput. Spreading requests across Regions raises your ceiling above one Region's capacity. Bursty traffic is absorbed instead of throttled.
Resilience under load. A busy Region stops being a single point of pressure; requests flow elsewhere in the profile without you writing failover logic. It is not a full disaster-recovery plan, but it helps the inference call.
No commitment. Unlike Provisioned Throughput, you reserve nothing.
Limits: it is still best-effort capacity, not a guaranteed contract, and a distant destination Region can add a little latency.
VICost, and cross-region vs single-region
AWS charges no extra routing fee. The price is based on the Region you call from, and global profiles are about 10% cheaper.
Use a geographic profile when traffic is high or bursty and geography-level residency is acceptable. Use a global profile when you have no location rules and want the lower price. Stay single-Region when a rule or contract pins data to one Region, or when traffic is low and steady. Many systems mix them by workload.
| Dimension | Single Region | Geographic profile | Global profile |
|---|---|---|---|
| Effective throughput | One Region's capacity and quota | Pooled across the geography | Pooled worldwide |
| Throttling at spikes | More likely | Smoothed | Smoothed |
| Price | Standard | Standard, based on source Region | About 10% lower (AWS) |
| Where processed | Your chosen Region | Any Region in the geography | Any supported commercial Region |
| Setup | Call the model ID | Call the profile ID; IAM and SCPs | Call the global profile ID; SCP allows "unspecified" |
| Best for | Region-pinned rules | Geography-level residency | No location rules; cost focus |
VIIWho can fund a resilient Bedrock architecture
Cross-region requests are ordinary Bedrock spend, so they draw down AWS credits automatically. The hard part is the architecture: mapping each workload's residency rule to the right profile, IAM and SCPs, and monitoring.
Partner-filed funding: a vetted AWS partner can file POC (proof of concept) funding for a new GenAI workload, MAP funding if you are moving from another cloud, or incremental workload (IW) funding if you already run on AWS and are launching new work. For startups this is typically $50K to $150K. Depends on the workload; AWS approves.
AWS Activate: you apply yourself. Activate Founders gives self-funded startups $1,000 to start, up to $5,000. Activate Portfolio gives up to $200,000 with the Org ID from your VC, accelerator or angel group, if it is an Activate Provider. A consulting partner cannot file Activate for you.
CloudRoute matches you with one vetted AWS partner who can design the profile and residency setup and check which funding fits. Startups pay CloudRoute nothing. Partners pay CloudRoute a monthly membership plus 10% of AWS-approved funding.
| Funding | Who files it | Amount | When it fits |
|---|---|---|---|
| Migration Acceleration Program (MAP) | Your matched AWS partner | $50K to $150K for startups. Depends on the workload; AWS approves. | You are not on AWS yet and want to migrate. |
| Proof of concept (POC) funding | Your matched AWS partner | $50K to $150K for startups. Depends on the workload; AWS approves. | You want to test something, whether or not you are on AWS today. |
| Incremental workload (IW) funding | Your matched AWS partner | $50K to $150K for startups. Depends on the workload; AWS approves. | You are already on AWS and are launching new work: a new AI feature, a new customer or a new contract. |
| AWS Activate Founders | Founder applies | $1,000 to start, up to $5,000 | Self-funded startups with no investor or accelerator behind them. |
| AWS Activate Portfolio | Founder applies | up to $200,000 | Startups backed by a VC firm, accelerator or angel investor that is an AWS Activate Provider. |
| AWS Credits for AI Startups | Invite only (AWS account manager) | $200,000 or more, by invitation | AI startups ready to scale after Activate Portfolio, invited by AWS. |
Cross-region inference vs Provisioned Throughput
Both address capacity, in opposite ways.
| Variable | Cross-region inference | Provisioned Throughput |
|---|---|---|
| What it does | Spreads on-demand requests across Regions | Reserves dedicated capacity in one Region |
| Commitment | None | Hourly, optional term |
| Guarantee | Best-effort | Guaranteed throughput |
| Extra cost | None (global profiles about 10% cheaper) | Paid even when idle |
| Custom models | n/a | One way to serve fine-tuned models |
| Works with inference profiles | Yes (it is one) | No, per AWS |
How a match works (example)
- You tell us about the work
You fill in a short form: what you run today, what you want to move, test or build, your stage and your region.
- We pick one partner
We read your answers and pick one vetted AWS consulting partner that fits the work, the stack and the region. We reply within 24 hours; your matched partner reaches out within 5 business days.
- The partner checks POC funding
Your matched partner scopes the test with you. Then it checks whether the test qualifies for POC (proof of concept) funding. A POC can fit whether or not you run on AWS today.
- The partner files, AWS decides
If the work qualifies, the partner files the funding request with AWS. AWS reviews it and decides. The amount depends on the workload.
- You pay nothing
CloudRoute is free for startups. Partners pay CloudRoute a monthly membership plus 10% of AWS-approved funding.
AWS Activate credits are separate, and you apply for them yourself. Use Activate Founders if no investor backs you. Use Activate Portfolio with the Org ID or invite from your VC, accelerator, incubator or angel group if it is an Activate Provider. A consulting partner cannot apply for Activate on your behalf.
Common questions
What is cross-region inference in Amazon Bedrock?
What is the difference between geographic and global cross-region inference?
How do I enable cross-region inference?
Does cross-region inference cost more?
Where is my data processed with cross-region inference?
Can I use Provisioned Throughput with cross-region inference?
When should I stay in a single Region?
Can AWS funding cover a cross-region Bedrock build?
Build resilient, residency-correct inference with a partner who can file the funding
A vetted AWS partner can map each workload to the right profile or Region, set up IAM and SCPs, and add monitoring. The same partner checks whether the work qualifies for POC, MAP or IW funding: typically $50K to $150K for startups. Depends on the workload; AWS approves. CloudRoute is free for startups.