NXP Moves to Amazon's Cloud in Bid to Speed Up Chip Design

Semiconductor companies are heading for the cloud.

NXP announced that it is moving the vast majority of its electronic design automation (EDA) workloads to Amazon's AWS cloud service. By moving more of its chip design work to the cloud, NXP said it is hoping to reduce costs and give its engineers more time to focus on building chips rather than managing compute resources.

Amazon said NXP is using a wide range of its compute, storage, analytics, and machine learning services to help boost productivity for engineers handling design and verification.

NXP believes that moving to the cloud will transform the way it designs and tests new generations of chips for automotive, industrial, and Internet of Things (IoT) sectors and give it a competitive edge over rivals.

Designing a modern chip is complex process that can take months or even years. It is impossible for even the most skilled engineers to test every single detail in the design by hand. NXP and other chip makers use EDA software to automate many of computations needed to configure and stitch together all the transistors in today's chips. That way, engineers can save time and focus on the more important aspects of a design.

Before a chip is manufactured, NXP said that it puts the design through rigorous testing and validation to make sure that it is functionally safe and secure and meets its goals for performance and power efficiency. NXP said it uses software tools for front-end design work, such as performance simulation and verification, along with back-end efforts such as timing and power integrity analysis and design rule checks to prepare a chip for production.

Historically, NXP and other semiconductor firms ran these workloads on-premises in in-house data centers with a fixed amount of resources. But because of the increasing complexity of chip designs, a vast amount of computing power is needed for each cycle. That put the onus on NXP to accurately forecast its computing needs and expand its in-house data centers.

To underscore the scale of the problem, NXP said it plans to store petabytes (a petabyte equals a million gigabytes) of design simulation data in Amazon's data centers.

Now NXP plans to rent out all the computing power and storage it needs through Amazon's AWS cloud arm instead of purchasing servers and software for internal use. That will give NXP the ability to buy computing power on demand, giving its engineers the flexibility to develop and test several chips at the same time—or run many performance simulations of the same chip design in parallel to reduce costs and time-to-market.

Renting out computing power from AWS will help stop NXP from over-spending on its internal data centers. It will also help reduce the risk of not having enough computing power on hand to test a chip design within its schedule, forcing it to delay new products, AWS said.

Shifting to the cloud also allows NXP to use Amazon's analytics and machine learning services to aid in its research and development endeavors. For instance, NXP said that it's using Amazon QuickSight, a machine learning-powered service, to help increase its productivity. By rapidly translating the results from one step of testing into modifications for another, it can reduce the time it requires to iterate on chip designs, NXP said.

NXP is taking advantage of a wide range of cloud computing services from AWS (also called instances). That allows it to use high-performance instances where it counts and less costly ones for other work.

NXP is not alone in moving more of its chip engineering to a public cloud. In January, Arm announced plans to move most of its EDA workloads to AWS to help it design chips faster and more cheaply than today. The company said at the time that running EDA software in a public cloud was a first for the industry and that it plans to reduce its internal data center footprint by 45% or more as it shifts to the cloud and 80% long-term.

Cadence, Synopsys, and other EDA vendors have started selling tools that can run in the cloud. But the chip industry has resisted a wholesale shift to the cloud in part because of the proprietary nature of chip designs.

Olli Hyyppa, senior vice president and chief information officer at NXP, said the rewards outweigh the risks. Shifting EDA workloads to the cloud will help it accelerate innovation and bring chips to market faster, he said.

"This will give precious time back to our design engineers to focus on innovation and lead the transformation of the semiconductor industry," he added.