IN Brief:
- Cadence OnCloud is available from sovereign infrastructure in Munich for European engineering customers.
- The service supports semiconductor EDA, verification and simulation workloads with European data residency.
- NVIDIA Vera and Grace computing, Millennium M2000 capacity and further digital twin instances are planned additions.
Cadence Design Systems has made its sovereign OnCloud infrastructure available to European semiconductor and electronic system developers from a Munich data centre, offering design and simulation services with data residency within European jurisdiction.
Cadence confirmed the service on 21 October during NVIDIA GTC Berlin. The available OnCloud environment combines Cadence electronic design automation software with accelerated computing resources, while additional NVIDIA Vera and Grace systems, a Millennium M2000 AI Supercomputer and further Reality Digital Twin Platform instances are planned as the facility expands. Cadence has not disclosed commissioning dates for that equipment.
Semiconductor projects generate commercially sensitive information long before a circuit reaches fabrication. Logic descriptions, physical layouts, timing constraints, simulation results and process-related design rules may involve intellectual property belonging to several organisations. Locating computing workloads within an agreed operating environment can help customers control where these materials are processed and retained.
European data residency does not, on its own, resolve every security or contractual obligation. Engineering teams must establish who administers the systems, how access is authorised, whether supporting services transfer information elsewhere and which controls protect files during processing. Cadence identifies ISO 27001, ISO 27017 and SOC 2 Type II assurance credentials for OnCloud, but individual customers still need to assess their own compliance requirements.
Cloud computing is attractive to chip developers because demand for processing capacity changes across a programme. Early functional simulation can involve many independent test runs, whereas later implementation stages may require substantial memory, rapid data access and repeated optimisation of timing or power. Maintaining enough local hardware for peak demand can leave expensive resources underused at other times.
Cadence’s Munich arrangement gives design teams access to remote capacity for such workloads. Simulation can test how a proposed design responds to selected inputs, while timing analysis examines whether signals can move through the intended circuit within defined intervals. Physical verification, power estimation and checks against manufacturing design rules add further computational demands before fabrication.
Some jobs can be distributed across numerous processors, although others are constrained by shared memory, storage access or the time required to exchange information between machines. Accelerated hardware is therefore only one part of performance: scheduling, available software licences and the organisation of large design datasets also determine how efficiently work progresses.
The Munich deployment expands Cadence’s European computing footprint alongside existing cloud resources in Frankfurt. Cadence also intends the sovereign service to support eligible start-ups participating in the EuroCDP programme, which could give smaller chip design teams access to engineering capacity without equivalent investment in their own computing installations.
Cadence’s proposed NVIDIA Vera and Grace processing systems and Blackwell-powered Millennium M2000 infrastructure would extend the available compute configurations. These additions remain planned, with installation dates, customer allocation and eventual operational capacity unconfirmed.
Cadence is also extending the use of its Reality Digital Twin Platform to model the physical infrastructure needed to operate high-density computing equipment. These installations must distribute substantial electrical loads and remove heat while maintaining acceptable operating conditions across processors, storage systems and supporting power equipment.
Digital models can represent rack positions, electrical distribution, cooling arrangements and the thermal loads of computing hardware. Engineers can alter a proposed configuration and examine predicted changes before installing equipment. Cadence has developed a validated digital model of its Millennium M2000 system to support this work, although model accuracy still depends on the conditions and assumptions used in each simulation.
The platform supports NVIDIA Omniverse DSX Blueprint workflows and is being used in related projects with infrastructure suppliers and operators. Cadence intends to apply the same modelling approach to its Munich environment, examining energy performance and operational risks as the additional computing systems are installed.
Thermal simulations cannot replace commissioning measurements. In an operating data centre, coolant flow, ambient conditions, equipment power draw and control-system settings may differ from the assumptions used during design. Engineers must compare predictions with observed performance and adjust models where necessary.
Cadence has reported improved capacity utilisation and availability in another digital twin deployment, but those results are not measurements from the newly available Munich OnCloud service. Similarly, the wider Reality platform’s customer projects do not establish that every associated application is already hosted in Munich.
The available sovereign service now gives European engineering teams a defined environment for semiconductor design and simulation. Its broader AI infrastructure capabilities will depend on the commissioning of the planned processors, Millennium computing capacity and additional digital twin resources, together with customer verification that the operating arrangements satisfy their data protection requirements.
