CONTEXT: Digital Twins (DTs) undergo a rapid shift from monolithic cyber-physical systems to distributed, cloud-native stacks. While the elasticity of public cloud providers benefits this transition—opaque, provider-specific pricing makes economically viable deployment, especially across federated clouds, difficult. OBJECTIVES: This work aims to enable cost-aware deployment of digital twins across federated cloud providers by modeling DTs as multi-layered architectures and optimizing service selection per layer to minimize operational expenses. METHODS: We introduce Twin2Clouds, a cost-driven DT engineering framework that (1) organizes DT functionality into five cloud-oriented layers to help engineers map twin components to suitable cloud services, (2) offers cloud-agnostic cost and pricing primitives that capture heterogeneous pricing schemes across providers and services, and (3) integrates these elements in a cost-aware deployment model that predicts cloud costs and prescribes a reproducible deployment plan selecting the lowest-cost services for each layer. RESULTS: A quantitative evaluation of three real-world scenarios, ranging from a large smart building (30,000 devices) to a smart home (100 devices), demonstrates that Twin2Clouds consistently outperforms single-provider baselines. Depending on scale and workload, monthly costs decrease by an average of 24.6% and by up to 79% compared with single-cloud alternatives. CONCLUSION: Twin2Clouds equips DT engineers with a practical, vendor-neutral method for navigating today’s complex cloud landscape and for realizing scalable, sustainable, and economically viable digital twins.