Can You Run IBM i on AWS? What Is Actually Possible in 2026
Can you run IBM i on AWS?
Not natively. Amazon EC2 runs on Intel and AMD x86 processors and AWS's own Arm-based Graviton chips, and IBM i runs only on IBM Power processors, so an IBM i partition cannot boot on EC2. That has not changed as of August 2026, and no announcement suggests it will. What does work: managed IBM Power hosting in data centers physically adjacent to AWS regions, IBM Power Virtual Server connected to AWS over a private interconnect, or replatforming applications off IBM i onto Linux on EC2. Which pattern fits depends mostly on how tightly your applications are coupled to Db2 for i and how much change you can absorb.
Practical decision support
Pick the integration pattern before the provider
Use this guide to match your IBM i estate's coupling to Db2, tolerance for application change, and existing AWS footprint to one of the workable patterns before shortlisting vendors.
- Profile application-to-database chattiness first: it decides whether a split architecture is even viable.
- Attribute and verify every latency figure; contract for a measured SLA, not a marketing number.
- Score replatforming as a modernization project with full regression testing, not as a hosting change.
At a glance
Key takeaways
IBM i cannot boot on EC2
EC2 runs x86 and Graviton (Arm) processors; IBM i runs only on IBM Power. No native Power instance had been announced as of August 2026.
AWS's answer is partners next door
AWS partner content endorses managed Power hosting adjacent to its regions over Direct Connect, and AWS signed a Strategic Collaboration Agreement with Connectria in November 2023.
Latency decides split architectures
Chatty application-to-Db2 traffic fails over WAN distances, per IT Jungle's reporting. Adjacent providers engineered for roughly 2ms round trips; verify with a measured SLA.
Replatforming is a modernization project
Recompiling RPG and COBOL for Linux on EC2 is a documented path (AWS names Infinite i), but plan for testing and operational change, not a lift-and-shift.
Can You Run IBM i or AIX on AWS? The Short Answer
You cannot run IBM i natively on AWS. There is no EC2 instance type built on IBM Power processors, and IBM i does not run on anything else. The same is true of AIX. This is not a controversial reading: CloudFirst's parent company, Data Storage Corporation, states plainly in its press releases and SEC filings that AWS, Microsoft Azure, and Google Cloud do not natively support IBM i or AIX workloads, and its hosting business is built on that fact.
What you can do is put IBM i close to AWS, connect the two privately, and operate them as one environment. AWS's own partner content endorses exactly that pattern. The rest of this guide walks the realistic options: managed Power hosting adjacent to AWS regions, IBM Power Virtual Server with an interconnect, and replatforming off IBM i entirely, along with the latency constraints that decide whether a split architecture works at all.
Why EC2 Cannot Boot IBM i or AIX
Amazon EC2 offers instances on Intel and AMD x86 processors and on AWS's own Graviton processors, which are built on Arm Neoverse cores. IBM i and AIX are built for the POWER instruction set and run only on IBM Power hardware. An operating system cannot boot on a processor family it was never built for, and AWS sells no Power-based instances, so no EC2 instance type, size, or region changes the answer.
Nothing public suggests this is about to change. According to AWS's own blog coverage, the current AWS and IBM collaboration showcased at IBM Think 2026 covers agentic AI (IBM's Bob on Amazon Bedrock), mainframe modernization for z/OS, autonomous databases, and enterprise automation. IBM Power, IBM i, and AIX are absent from that list. No announcement of native Power instances on AWS existed as of this writing, and the silence should not be read as intent in either direction.
What AWS Itself Offers for IBM i and AS/400 Workloads
AWS addresses IBM i through partner programs and data services rather than compute. Three published pieces matter. First, AWS's Partner Network blog promotes hosting IBM i and AIX systems with low-latency connectivity to AWS through Connectria, which amounts to AWS endorsing adjacent hosting plus Direct Connect as its answer for Power workloads. Second, AWS's migration-options blog ("Demystifying Legacy Migration Options to the AWS Cloud") categorizes the replatforming paths and names specific tools, including Infinite i for recompiling IBM i applications to run on x86. Third, in November 2023 AWS announced that AWS Mainframe Modernization Data Replication with Precisely supports IBM i, replicating Db2 for i data in near real time into AWS databases including RDS and Aurora; the underlying tool is Precisely Connect CDC (MIMIX Share). Note the scope: the Mainframe Modernization service is aimed at z/OS, and for IBM i the documented capability is the Precisely data replication feature, not application migration.
One clarification that trips people up: Amazon RDS for Db2, launched in November 2023, runs Db2 for Linux, Unix, and Windows version 11.5.9. That is a different product from Db2 for i, which IBM's documentation describes as fully integrated into the IBM i operating system. Moving an IBM i database to RDS for Db2 is therefore a conversion project, not a like-for-like migration.
Pattern 1: Managed IBM Power Hosting Adjacent to AWS
The closest thing to IBM i "on" AWS is IBM Power hardware racked in a data center next door. Connectria launched this model in November 2020, hosting IBM i and AIX on Power systems in facilities physically adjacent to AWS data centers, connected by private low-latency links. Connectria markets it as IBM on-net with AWS and states round-trip latency of under 2 milliseconds to AWS availability zones in US-East-1 and US-West-1. Treat that figure as a provider claim: actual latency depends on the specific facility, region, and availability zone, so ask for a measured latency SLA before designing around it. Connectria has also announced two Singapore data centers to extend the model to Asia Pacific, so coverage is growing; verify current locations directly with the provider.
AWS has leaned into this model rather than competing with it. It published APN blog content promoting the architecture, and in November 2023 it signed a Strategic Collaboration Agreement with Connectria to accelerate modernization of IBM Power workloads with AWS. IT Jungle's reporting on the deal was blunt about the motivation: AWS lacked a dedicated Power strategy of its own. In 2024, Lightedge acquired Connectria, and the combined company now markets IBM i and AIX hosting connected to AWS, Azure, and Google Cloud, again with a stated sub-2ms figure (data center counts and cloud partner tiers are the company's own claims and change; verify current facilities with the provider). CloudFirst, a Data Storage Corporation subsidiary, runs the same play with direct connections into AWS, Azure, Google Cloud, and IBM Cloud, and has announced UK and European expansion. Our independent directory of IBM Power hosting providers profiles these firms individually.
Pattern 2: IBM Power Virtual Server with a Private Interconnect to AWS
IBM Power Virtual Server is IBM's infrastructure-as-a-service offering for IBM i, AIX, and Linux on Power, running in IBM data centers. It connects privately to IBM Cloud and to external environments through IBM Cloud Direct Link, which IBM documents for multicloud and data-center-to-data-center connectivity. For reaching other clouds, IBM has published guidance on connecting Power Virtual Server through the network exchange provider Megaport; the documented reference pattern targets Microsoft Azure. The same exchange-provider approach can reach AWS in principle, but we did not find an IBM-published AWS-specific pattern, so validate the design with IBM and the network provider before committing.
Latency is the honest caveat. We found no verified figure for PowerVS-to-AWS round trips, and the result depends entirely on which PowerVS site you pair with which AWS region. It is reasonable to assume the number will be higher than physically adjacent hosting, because the facilities were not placed next to each other. If your architecture keeps chatty traffic inside PowerVS and uses AWS for batch, API, or replicated-data integration, that may not matter. If you are counting on per-record database calls across the link, test with production-shaped traffic first.
Pattern 3: Replatforming Off IBM i onto x86 and Linux on AWS
The third pattern removes the Power dependency instead of working around it. AWS's migration blog categorizes the approaches (hardware emulation, middleware emulation, automated refactoring, and middleware replatforming) and, for IBM i, names Infinite i, which recompiles RPG, COBOL, CL, and DDS source so the applications run on Linux on x86; it is listed on AWS Marketplace. Be realistic about what that involves. Recompiling is the start of a project that also includes testing every job stream, retraining operations, and revalidating every integration, which is why we treat replatforming as a modernization decision rather than a hosting decision. Our migrate-versus-modernize comparison walks that trade-off in detail.
There is also a data-first variant that leaves the application on IBM i and moves the data. The Precisely replication capability announced by AWS replicates Db2 for i into AWS databases for analytics or a phased migration. Partner case studies document the fuller journey: an AWS APN blog describes how DXC migrated a Db2 database on AS/400 to a cloud-native architecture on Amazon S3, AWS Glue, and Amazon Athena, and Lightedge published a case study of a luxury brand that migrated part of its IBM i LPAR workloads while keeping the remainder on managed Power alongside an AWS estate. Partial moves like these are common in practice; a full exit is a much larger commitment.
The Split-Stack Architecture: Latency Rules When Apps Live on AWS and Db2 Stays on IBM i
Whichever hosting pattern you choose, most real architectures end up split: transactions stay on IBM i and Db2 for i, while AWS hosts what surrounds them. IT Jungle's April 2026 coverage describes exactly that division of labor, with IBM i keeping core transaction processing while hyperscalers host analytics, customer-facing web and mobile applications, DevOps pipelines, and resilience architectures.
Latency is what makes or breaks the split. IT Jungle reported that customers historically could not move x86 application tiers to hyperscale clouds while keeping IBM i on premises, because chatty application-to-Db2 traffic made the applications non-functional over WAN latency. That is the engineering rationale for adjacent hosting: Connectria's stated design goal was roughly 2 millisecond round trips into each availability zone, close enough that record-level database access still behaves. As a working rule, batch transfers, API-based integration, and replicated data tolerate distance well; per-record, chatty database access tolerates it worst. Connectria's Transflo case study describes the pattern in production, with real-time applications talking to a fully managed IBM i Db2 stack on-net with AWS, per the provider's own account. Measure your application's chattiness before you pick a pattern, not after.
How AWS Compares with Azure and IBM Cloud for Power Workloads
AWS is the only one of the major hyperscalers covered here without a Power-inside-the-facility story. Microsoft Azure hosts IBM Power workloads (AIX, IBM i, and Linux on Power) natively inside Azure data centers through Kyndryl Cloud Uplift, the former Skytap service Kyndryl acquired in 2024, and Microsoft publishes a reference architecture for it. Kyndryl has also offered the capability on IBM Cloud, though Azure is the flagship destination and current IBM Cloud availability should be verified with Kyndryl. IBM Cloud itself offers Power Virtual Server. AWS, by contrast, relies on adjacent partner hosting.
That difference is not automatically decisive. If the rest of your estate, your data platform, and your DevOps tooling already live on AWS, adjacent managed Power hosting keeps everything close and under one integration model. If you specifically want Power hardware inside a hyperscaler's own facility, that points to Azure or IBM Cloud today. The right answer falls out of your application coupling, your team's cloud skills, and your exit constraints, which is exactly what an independent assessment of your IBM i estate is designed to surface.
Four Paths for IBM i Workloads That Need AWS
Use this table to match each path to your tolerance for application change and latency.
| Option | How it works | Trade-offs to weigh |
|---|---|---|
| Managed Power hosting adjacent to AWS | IBM i runs on provider-managed Power hardware next to an AWS region, linked by Direct Connect; providers state sub-2ms round trips | Least application change; you take on a hosting contract and must verify locations, measured latency, and exit terms |
| IBM Power Virtual Server plus interconnect | IBM i runs in IBM data centers on PowerVS; Direct Link plus an exchange provider reaches AWS | IBM's published reference pattern targets Azure; AWS designs need validation, and latency reflects the real distance between sites |
| Replatform to Linux on EC2 | Recompile RPG, COBOL, CL, and DDS (for example with Infinite i) so applications run on x86 EC2 instances | Removes the Power dependency but is a full modernization project with heavy testing and operational change |
| Data-first replication, IBM i stays | Replicate Db2 for i in near real time into AWS databases with Precisely Connect CDC for analytics or phased migration | Transactions stay on IBM i; RDS for Db2 is Db2 LUW, not Db2 for i, so any database move is a conversion |
Risks and Common Mistakes
Use caution
Treating sub-2ms latency claims as guarantees
The under-2-millisecond figures for adjacent hosting come from provider marketing, corroborated mainly by trade-press interviews with those providers. Actual latency depends on the facility, region, and availability zone, so contract for a measured latency SLA before you design around the number.
Use caution
Assuming RDS for Db2 accepts an IBM i database
Amazon RDS for Db2 runs Db2 for LUW 11.5.9. Db2 for i is integrated into the IBM i operating system and cannot be detached from it, so moving the database to RDS is a conversion project with real schema and application work.
Use caution
Planning around native Power on AWS appearing
No announcement of native IBM Power instances on AWS existed as of this writing, and the current AWS and IBM collaboration covers AI and z/OS modernization, not Power hardware. Plan with the patterns that exist today.
Frequently Asked Questions
Can you run IBM i (AS/400) on AWS EC2?
No. EC2 instances use Intel or AMD x86 processors or AWS's own Graviton (Arm) processors, and IBM i runs only on IBM Power processors, so it cannot boot on EC2. That remains true as of August 2026, and AWS has announced nothing that changes it. The workable options are managed Power hosting adjacent to AWS, IBM Power Virtual Server with a private interconnect, or replatforming off IBM i.
Can you run AIX on AWS?
No, for the same architecture reason as IBM i: AIX requires Power hardware. The common answers are managed AIX hosting connected to AWS over Direct Connect (offered by Connectria, now part of Lightedge, and by CloudFirst), AIX on IBM Power Virtual Server, or migrating the application from AIX to Linux on EC2, which AWS partner content treats as a modernization project, not a lift-and-shift.
What is the closest thing to native IBM i on AWS?
Managed Power hosting in data centers physically adjacent to AWS regions, linked by AWS Direct Connect. Providers state round-trip latency under 2 milliseconds to AWS availability zones; treat that as a provider claim and ask for a measured latency SLA for your specific facility. AWS's own Partner Network blog endorses the pattern, and AWS signed a Strategic Collaboration Agreement with Connectria in November 2023.
Does Amazon RDS for Db2 let me move my IBM i database to AWS?
Not directly. RDS for Db2 runs Db2 for Linux, Unix, and Windows (version 11.5.9), a different product from Db2 for i, which is integrated into the IBM i operating system and cannot be detached from it. Moving the data means conversion and replication, for example AWS Mainframe Modernization Data Replication with Precisely, available for IBM i since November 2023, typically for analytics or a phased migration while transactions stay on IBM i.
How much latency can a split IBM i and AWS architecture tolerate?
It depends on how chatty the application-to-database traffic is. Trade press (IT Jungle) documented that ordinary WAN latency made split architectures non-functional for record-level Db2 access, which is why adjacent-hosting providers engineered for roughly 2 millisecond round trips. Batch, API-based, and replicated-data integrations tolerate far more; per-record, chatty access tolerates the least. Test with real workloads before committing.
What are my options for an AS/400 to AWS migration?
Four realistic paths: rehost onto managed Power adjacent to AWS and integrate over Direct Connect; move to IBM Power Virtual Server and interconnect to AWS; replatform by recompiling RPG, COBOL, and CL for Linux on EC2 (Infinite i is named in AWS's own migration-options blog); or go data-first, replicating Db2 for i data to AWS services while the application stays put. Each trades cost, risk, and timeline differently, and an independent assessment should precede tool selection.
Is Azure or IBM Cloud a better home for IBM i than AWS?
They are different, not automatically better. Azure hosts IBM Power natively inside its data centers via Kyndryl Cloud Uplift, the former Skytap service Kyndryl acquired in 2024, and IBM Cloud offers Power Virtual Server. AWS has no in-facility Power service and relies on adjacent partner hosting. If the rest of your estate is on AWS, adjacent hosting keeps everything close; if you specifically want Power inside a hyperscaler's own facility, that points to Azure or IBM Cloud today.
Has AWS announced native IBM Power support?
No such announcement was found as of August 2026. The current AWS and IBM collaboration showcased at IBM Think 2026 covers agentic AI, z/OS mainframe modernization, and IBM software running on AWS, not Power hardware in AWS data centers. Treat the answer as current as of this writing and revisit if AWS or IBM announce otherwise.
Sources
- AWS APN Blog: Hosting IBM i and AIX Systems with Low-Latency Connectivity to AWS with Connectria
- AWS APN Blog: Demystifying Legacy Migration Options to the AWS Cloud
- AWS What's New: Mainframe Modernization Data Replication with Precisely for IBM i (Nov 2023)
- Amazon RDS for Db2 (Db2 LUW)
- IBM Documentation: Db2 for i overview (integrated into IBM i)
- IBM Cloud docs: Direct Link Connect for IBM Power Virtual Server
- Microsoft Learn: Deploy IBM Power workloads on Azure (Skytap / Kyndryl Cloud Uplift)
- IT Jungle: A Year Later, Diving Deeper into How Connectria Brings IBM i Workloads Closer to AWS
Not Sure Which Pattern Fits Your IBM i Estate?
Get an independent, directional read on whether adjacent hosting, Power Virtual Server, or replatforming fits your applications and your AWS footprint.