Synopsys and Amazon signed a multi-year agreement valued at more than $1 billion to expand Amazon's use of Synopsys silicon IP and EDA tools across its custom AI chip portfolio, the companies said September 30, 2026, from Sunnyvale, California.

The deal marks the first time Synopsys will deploy a license-plus-royalty business model, tying a portion of its IP revenue to production volumes of Amazon-designed silicon. Amazon becomes the lead customer for Synopsys' newly launched application-optimized IP initiative.

What chips fall under the agreement?

The agreement extends a 15-year-plus engineering relationship. Amazon will gain broader access to Synopsys' silicon IP blocks, EDA tools, simulation and analysis (S&A) software, and agentic AI technologies. Joint engineering teams will focus on multiphysics simulation and optimization for Trainium (AI training and inference) and Graviton (general-purpose Arm compute).

Amazon's custom-silicon portfolio also includes Nitro, the system-on-chip that handles cloud security, networking and storage offload in AWS data centers. The hyperscaler has historically paired each of these chip families with its own internal design teams, supplemented by third-party IP and EDA flows.

What changes in Synopsys' revenue model?

Until now, Synopsys has sold IP primarily through upfront licenses bundled with maintenance. Under the Amazon agreement, the company layers a royalty component on top, scaling with chip volumes. The shift converts part of Synopsys' IP revenue from one-time sales into a recurring stream tied to customer shipments.

"Special purpose silicon is at the heart of AWS innovation," said Synopsys president and CEO Sassine Ghazi. "As we expand into application-optimized IP solutions, we're proud to have Amazon as our lead customer for this next phase of growth."

"More broadly, by combining Amazon's leadership in AI infrastructure and custom silicon with Synopsys' leadership in silicon IP, EDA and multiphysics solutions, we can dramatically increase the productivity of our teams and shape the future of AI-native engineering," Ghazi added.

How does AI enter the design loop?

Both companies plan to apply AI across silicon-to-system workflows. Synopsys will deploy its AI-powered EDA, physics-based simulation and agentic AI tools inside Amazon's design flow. The two firms will also develop custom agentic AI capabilities for Amazon's chip teams, targeting the design, analysis, optimization and validation stages of complex processor development.

In the other direction, Synopsys will migrate parts of its own IP and EDA tool development onto Amazon EC2, Amazon cloud storage, and Amazon Bedrock. The company will use Bedrock to build and deploy internal AI agents intended to accelerate R&D.

Why does this matter for the broader IP market?

Synopsys competes with Cadence and Siemens EDA in the merchant IP and EDA markets. A royalty-tied deal with one of the world's largest chip buyers signals a possible template for future agreements, particularly as hyperscalers continue to pull chip design in-house.

"Amazon has invested in custom silicon for more than a decade, and it has fundamentally changed what's possible for AWS and our customers," said Peter DeSantis, SVP of foundational AI, custom silicon and quantum computing at Amazon. "From Graviton to Trainium, purpose-built chips deliver better performance at lower cost because they're designed for exactly what customers need."

"As our chip designs grow more ambitious and AI reshapes the engineering process itself, Synopsys helps us move faster across the design cycle, helping us deliver more capable, efficient computing for customers worldwide," DeSantis added.

What remains undisclosed?

Synopsys (Nasdaq: SNPS) did not disclose the royalty rate, the share of revenue tied to license versus royalty, or specific delivery milestones. Neither company gave a timeline for when the first application-optimized IP blocks would tape out. The financial impact of the new model will hinge on Trainium and Graviton shipment volumes, which AWS has historically reported in aggregate rather than per-chip, leaving the royalty stream's scale to surface in future Synopsys earnings reports.