IN Brief:
- The multi-year agreement exceeds $1 billion and expands Amazon’s use of application-optimised Synopsys semiconductor IP.
- Amazon will increase its use of Synopsys EDA, simulation, and AI engineering technology.
- Synopsys will deepen its own use of AWS compute, Bedrock, Trainium, and Graviton infrastructure.
Synopsys and Amazon have signed a multi-year agreement worth more than $1 billion that extends their semiconductor relationship across application-optimised IP, electronic design automation, simulation, AI-assisted engineering, and cloud infrastructure. Amazon will become the lead customer for the next phase of Synopsys’ application-optimised IP business, building on more than 15 years of work between the two companies while increasing the amount of engineering technology tied directly to Amazon’s custom silicon programmes.
The commercial model moves part of Synopsys’ IP business towards a combination of upfront licensing and royalties linked to production volume, giving the supplier greater exposure to the eventual scale of the devices in which its technology is used. For Amazon, the attraction is a development model in which reusable IP can be shaped more closely around the architecture, interfaces, power targets, and workloads of a particular processor rather than adopted unchanged from a catalogue.
AWS has already pushed heavily into purpose-built silicon through Nitro infrastructure devices, Graviton processors, and Trainium AI accelerators, giving Amazon more control over the relationship between hardware, software, networking, data-centre infrastructure, and the workloads customers run. Designing a processor for one tightly defined environment creates opportunities to remove unnecessary functions and optimise resources more aggressively, although the gain comes with responsibility for architecture, verification, manufacturing, software support, and the long lifecycle of the resulting chip.
Synopsys is not replacing Amazon’s internal silicon engineering teams under the agreement; its role is expanding around the reusable IP, EDA software, simulation systems, and application-specific engineering assets those teams use. As that IP becomes more closely aligned with Amazon’s requirements, the boundary between a standard licensable block and a customer-specific development programme becomes less distinct, particularly where interface behaviour, memory architecture, power management, or implementation choices are shaped around one very large deployment.
Peter DeSantis, senior vice president for foundational AI, custom silicon and quantum computing at Amazon, has argued that purpose-built processors improve performance and cost because they can be designed around specific customer requirements. The latest agreement gives that strategy a deeper connection to Synopsys’ engineering stack, rather than announcing another finished Amazon processor in its own right.
Amazon is pursuing similar customisation through several suppliers at different layers of the system, including its separate collaboration with Qualcomm on AI inference silicon and high-speed connectivity. The Synopsys arrangement sits further upstream, concentrating on the design tools, reusable semiconductor building blocks, simulation, and engineering automation used before a finished accelerator or processor reaches production.
AI-assisted engineering forms another part of the agreement, with Amazon increasing its use of Synopsys systems intended to automate repeated optimisation and analysis tasks during semiconductor development. The attraction is straightforward in large custom designs, where each architectural change can trigger fresh rounds of timing, power, physical, and verification work, although useful automation still depends on deterministic tools proving that the proposed changes meet the constraints of the design.
The relationship also runs in the opposite direction, since Synopsys will expand its own use of Amazon EC2, cloud storage, and Amazon Bedrock while optimising software for Trainium and Graviton processors. Amazon therefore uses Synopsys technology to help build its infrastructure silicon, while Synopsys increasingly uses Amazon processors and cloud services to develop and run parts of the software portfolio used in semiconductor engineering.
That mutual dependence gives both companies an incentive to improve the efficiency of the other’s engineering environment, but it also illustrates how tightly custom silicon is becoming connected to the infrastructure on which it is designed. Cloud compute, AI-assisted engineering, EDA software, semiconductor IP, and the finished processor no longer sit in entirely separate procurement categories when one organisation controls the deployment environment at AWS scale.
No process nodes, package architectures, manufacturing schedules, or production volumes have been disclosed, and Synopsys has not identified which future Amazon processors will first contain the new application-optimised IP. The agreement consequently establishes a long-term engineering and commercial framework rather than providing specifications for a new device, with its practical significance becoming measurable only as future AWS silicon enters production and Synopsys begins to show how much of its IP business is moving towards royalty-bearing customer-specific development.
