Rambus partners with ChipAgents to use agentic AI for chip design

A new partnership for silicon

The semiconductor company Rambus has entered into a strategic collaboration with ChipAgents to integrate agentic AI into its core operations. This partnership focuses on using autonomous AI agents to streamline the design and verification of silicon intellectual property (IP). By adopting this technology, Rambus intends to increase the speed at which it develops complex chips and components.

The agreement involves Rambus implementing the ChipAgents platform across its various engineering workflows. This move is part of a larger effort to enhance engineering productivity and strengthen the company's position in the data-center infrastructure market. The collaboration combines Rambus's deep industry expertise with ChipAgents's AI-native capabilities to create more efficient design paths.

As techmonitor.ai reported, Rambus has also made a strategic investment in ChipAgents to support the development of these AI-driven capabilities. This investment underscores the importance of moving toward a system where AI agents can handle end-to-end design tasks rather than just isolated functions. The goal is to provide customers with improved speed and quality within their preferred design environments.

How agentic AI handles design

Agentic AI refers to systems that use autonomous agents to perform complex, multi-step tasks rather than just following simple commands. In the context of semiconductor production, these agents are designed to convert high-level specifications and code into production-ready register-transfer Level (RTL). RTL is a common hardware description language used to model the behavior of digital electronic systems.

Beyond just generating code, the ChipAgents platform provides verification assets and automated root-cause analysis. Verification is the process of ensuring that a chip design functions correctly before it is manufactured. By automating these steps, the platform aims to significantly reduce the cycle times required for both design and verification phases.

The technology is built to be tool-agnostic, meaning it can work with the various different software environments that engineering teams already use. This allows design engineers, verification engineers, and embedded systems teams to adopt AI assistance without having to abandon their established tools and technologies. It focuses on moving more efficiently from a design intent to a final silicon implementation.

Impact on semiconductor engineering

The primary beneficiaries of this collaboration are the engineering teams at semiconductor firms who manage the production of state-of-the-art SoC designs. These teams often face increasing complexity as demand for advanced chips grows. By using agentic AI, these engineers can move through the development pipeline with greater speed and fewer manual bottlenecks.

Rambus CEO Luc Seraphin noted that AI is creating a virtuous cycle of innovation by driving demand for more complex systems while simultaneously transforming how those systems are designed. The partnership aims to help these teams manage the high volume of work required for modern AI and data-center infrastructure. It provides a way to handle the increasing complexity of silicon IP more effectively.

For customers of Rambus, the collaboration means they can expect faster turnaround times and higher quality for their custom chip designs. The initiative seeks to improve the speed of integration for silicon IP. This is particularly important for companies looking to stay competitive in the rapidly evolving hardware landscape where time-to-market is a critical factor.

Benchmarks and future goals

To measure the progress of these technologies, ChipAgents recently released a new benchmark called Build-A-Chip-Bench. This benchmark was developed in collaboration with Nvidia to evaluate how well AI agents can complete end-to-end digital chip design tasks. It tests the ability of these systems to move from initial specifications all the way through physical implementation to a verified GDS.

A GDS is a file format used to describe the physical layout of an integrated circuit. The benchmark serves as a standardized way to measure the progress of AI capabilities in the semiconductor sector. It provides a clear metric for how far the industry has come in automating the transition from a conceptual design to a manufacturable product.

Moving forward, Rambus and ChipAgents plan to work on specific workflows that combine Rambus's silicon IP with the AI platform's capabilities. They intend to introduce the ChipAgents platform to their internal engineering teams involved in both chip design and silicon IP projects. This will serve as a testing ground for how agentic AI can be scaled across large-scale semiconductor production.