BMW Group Modernizes Connected Vehicle Platform with AWS Graviton Migration to Drive Efficiency and Sustainability

The BMW Group, in a significant technical milestone for its digital services division, has successfully modernized its massive Connected Vehicle backend infrastructure—known internally as Orbit—by migrating from traditional x86 architecture to Arm-based AWS Graviton processors. Executed by the BMW Connected Company in collaboration with Critical TechWorks (CTW), the large-scale transition encompasses a global fleet exceeding 24.5 million connected vehicles. The architectural overhaul, initiated in 2024, has already yielded roughly $600,000 in annualized infrastructure savings while accelerating price-performance by 40 percent and advancing the automotive giant’s overarching corporate environmental objectives.

Background Context and Strategic Imperatives

Modern automotive ecosystems rely heavily on continuous data processing, over-the-air updates, telematics, and real-time vehicle connectivity. For the BMW Group, maintaining a resilient, scalable, and economically efficient backend infrastructure is paramount. The Connected Vehicle platform handles immense daily operational loads, including more than 130,000 daily builds and deployments managed via self-hosted GitHub Actions runners.

As digital features expand across the multi-million vehicle fleet, the computational demands placed on cloud infrastructure grow exponentially. Traditionally reliant on x86 compute instances, BMW Group engineers identified an opportunity to optimize both cost and energy consumption by evaluating ARM-based architecture, specifically Amazon Web Services (AWS) Graviton processors. The shift aligns directly with BMW Group’s dual mandate: maintaining economic competitiveness while shrinking the carbon footprint associated with large-scale cloud computing operations.

Laying the Foundation: The 2024 Migration Chronology

The transition to AWS Graviton required careful, multi-phased architectural groundwork rather than a sudden, high-risk "rip-and-replace" maneuver. The engineering teams structured the journey into two primary phases: a build-foundation phase and a workload-migration phase, spanning six distinct stages.

In the initial stages of the project, the platform required a more flexible provisioning layer to handle diverse hardware architectures seamlessly. To achieve this, the team integrated Karpenter, an open-source Kubernetes cluster autoscaler, replacing rigid Managed Node Groups with dynamic, policy-driven node pools. This allowed the infrastructure to automatically provision the optimal instance types based on real-time workload demands.

Following the compute foundation upgrade, the continuous integration and continuous delivery (CD/CD) pipeline required modernization to support multi-architecture builds. Engineers introduced dedicated Graviton-based build runners alongside the legacy x86 fleet. Initial validations were performed on self-contained, isolated services to monitor stability and benchmark performance metrics. The early results confirmed immediate price-performance benefits, proving that Graviton runners delivered lower costs per build minute alongside noticeably faster execution times.

Overcoming Technical Hurdles in Multi-Architecture Workloads

A core challenge in migrating a live, high-availability platform is ensuring that software artifacts can run smoothly on disparate CPU architectures without disrupting ongoing developer workflows. To resolve this, BMW and CTW implemented a comprehensive multi-architecture build strategy.

Container images were engineered to support both amd64 (x86) and arm64 (Graviton) architectures natively. The engineering teams deployed two complementary approaches for container image construction: emulated builds utilizing Docker Buildx paired with QEMU, and native builds executed directly on Graviton-based runners. By publishing a unified manifest list alongside architecture-specific images, container runtimes automatically pull the correct variant depending on the underlying node architecture, rendering the transition entirely transparent to end-users and client services.

Furthermore, public base images were mirrored into a private Amazon Elastic Container Registry (Amazon ECR) using Google’s Crane tool. This ensured strict compliance with CVE (Common Vulnerabilities and Exposures) standards, maintained high availability, and filtered out unused architectures to optimize storage efficiency.

Coexisting Node Pools and Phased Rollout Strategies

To eliminate friction and empower individual application teams to migrate at their own pace, the platform engineering group designed a mixed-architecture Amazon Elastic Service (Amazon EKS) cluster environment. Shared, general-purpose Graviton node pools were established to maintain a consistent infrastructure model.

To prevent accidental misconfigurations or premature scheduling of non-compatible workloads, the new Graviton node pools were protected using Kubernetes taints. Application teams were required to explicitly opt-in by utilizing designated node selectors and matching tolerations. This policy ensured that only verified, Arm-ready workloads were scheduled onto Graviton instances, while legacy workloads continued operating securely on x86 nodes.

Before rolling the architecture out to broader application teams, the platform team executed a rigorous pilot rollout by migrating core platform services—such as Prometheus, Karpenter, and critical operational tooling—onto the Graviton infrastructure. This proactive validation surfaced edge cases early, refined node pool tuning, and established dependable rollback patterns.

By leveraging GitOps methodologies via ArgoCD, the platform guaranteed that if any unexpected irregularities surfaced, workloads could instantaneously fall back to x86 configurations simply by reverting Git pointers, eliminating complex manual interventions or image rebuilds.

Measurable Results and Economic Impact

The tangible outcomes of the AWS Graviton migration have exceeded initial internal projections. Today, more than 65 percent of the entire Connected Vehicle platform runs on Graviton instances across production environments.

The financial and operational implications are substantial:

• Annualized infrastructure cost reductions have surpassed $600,000.
• Price-performance metrics improved by approximately 40 percent, particularly for heavy networking components like the NGINX-based ingress controller.
• Energy consumption footprints associated with backend cloud workloads decreased correspondingly, contributing directly to BMW Group’s corporate sustainability targets.
• Build and deployment stability remained uncompromised throughout the multi-stage rollout, preserving the uninterrupted delivery of digital services to over 24.5 million connected vehicles worldwide.

Industry Implications and Lessons Learned

The successful transition executed by BMW Group, Critical TechWorks, and AWS offers a blueprint for enterprise-grade infrastructure modernization at scale. By prioritizing developer autonomy, minimizing adoption friction through clear opt-in mechanisms, and decoupling infrastructure changes from application logic, the engineering collective avoided the common pitfalls of legacy system migrations.

Industry analysts note that as enterprises face mounting economic pressures alongside increasing regulatory scrutiny regarding corporate carbon emissions, the adoption of specialized, energy-efficient silicon like AWS Graviton will likely accelerate across the automotive and manufacturing sectors. BMW’s approach demonstrates that complex cloud architectures can be overhauled incrementally without compromising system reliability, continuous delivery velocity, or service availability.

As the remaining platform workloads complete their transition to Arm-based infrastructure, the BMW Group continues to solidify its position not only as a traditional automotive manufacturer but as a sophisticated software-driven mobility provider. The tools, patterns, and methodologies established during this migration will serve as standard frameworks for future digital innovation across the BMW Connected Company ecosystem.