Cyber-Physical Systems (CPS) integrate computation with physical processes, using embedded computers and networks to monitor and control physical systems with feedback loops. These systems have significant economic and societal potential but face challenges due to the unique safety and reliability requirements of physical components, which differ from general-purpose computing. Current computing and networking technologies may not be sufficient for CPS, requiring a rethinking of core abstractions to embrace physical dynamics and computation. CPS applications include medical devices, traffic control, automotive systems, energy conservation, and more. However, current software and networking technologies lack the necessary temporal semantics and concurrency models for reliable real-time performance. Software components, built on abstractions that match software better than physical systems, may not be sufficient for CPS without substantial changes in core abstractions. The physical world is unpredictable, requiring CPS to be robust to unexpected conditions and adaptable to subsystem failures. While digital circuits are highly predictable and reliable, software systems face challenges in achieving similar predictability and reliability. The abstraction layers in computing have failed to account for timing properties, leading to unpredictable behavior in CPS. Designing CPS requires predictable and reliable systems, which can be achieved through new technologies that address the lack of temporal semantics in current abstractions. Solutions include bottom-up approaches like modifying computer architectures for precision timing and top-down approaches like model-based design, which use models to represent system behaviors and synthesize software. Coordination languages also offer new semantics for component interaction. The core abstractions of computing need to be rethought to fully realize the potential of CPS. Incremental improvements will continue to help, but effective orchestration of software and physical processes requires semantic models that reflect properties of interest in both. The paper concludes that CPS will require fundamentally new technologies to address the challenges of predictability, reliability, and robustness in the face of physical dynamics and concurrency.Cyber-Physical Systems (CPS) integrate computation with physical processes, using embedded computers and networks to monitor and control physical systems with feedback loops. These systems have significant economic and societal potential but face challenges due to the unique safety and reliability requirements of physical components, which differ from general-purpose computing. Current computing and networking technologies may not be sufficient for CPS, requiring a rethinking of core abstractions to embrace physical dynamics and computation. CPS applications include medical devices, traffic control, automotive systems, energy conservation, and more. However, current software and networking technologies lack the necessary temporal semantics and concurrency models for reliable real-time performance. Software components, built on abstractions that match software better than physical systems, may not be sufficient for CPS without substantial changes in core abstractions. The physical world is unpredictable, requiring CPS to be robust to unexpected conditions and adaptable to subsystem failures. While digital circuits are highly predictable and reliable, software systems face challenges in achieving similar predictability and reliability. The abstraction layers in computing have failed to account for timing properties, leading to unpredictable behavior in CPS. Designing CPS requires predictable and reliable systems, which can be achieved through new technologies that address the lack of temporal semantics in current abstractions. Solutions include bottom-up approaches like modifying computer architectures for precision timing and top-down approaches like model-based design, which use models to represent system behaviors and synthesize software. Coordination languages also offer new semantics for component interaction. The core abstractions of computing need to be rethought to fully realize the potential of CPS. Incremental improvements will continue to help, but effective orchestration of software and physical processes requires semantic models that reflect properties of interest in both. The paper concludes that CPS will require fundamentally new technologies to address the challenges of predictability, reliability, and robustness in the face of physical dynamics and concurrency.