MIPS, by GF (MIPS), a leading technology provider for efficient workload-focused platforms based on open technology, today announced a strategic non-exclusive partnership with Xcelsa Labs to make custom-chip design more accessible and accelerate the development of application-specific silicon through Xcelsa’s verified compute intelligence platform. The partnership adds Xcelsa’s capabilities to MIPS’ broader ecosystem of technology providers and partners, supporting customers across the diverse design, verification, implementation, and optimization requirements of workload-focused silicon.

MIPS’ software-first approach leverages Xcelsa’s Apex Design Optimization platform to accelerate the development of workload-focused custom silicon. Apex combines Xcelsa’s fast verified and physical intelligence stacks to iterate over thousands of design alternatives at the speed of AI to deliver improved designs faster. The verified intelligence stack guarantees functional equivalence and the physical intelligence stack reasons through physical consequences without relying on EDA tools. Together, the technologies can help customers reduce design iteration time, improve power, performance, and area, and deploy differentiated silicon for demanding AI, automotive, industrial, robotics, and edge applications.

“Customers need custom silicon that reflects the real requirements of their software and workloads, delivered within product schedules they can manage,” said Sameer Wasson, CEO of MIPS. “By combining MIPS’ software-first, workload-focused platforms with Xcelsa’s verified compute intelligence, we can shorten critical design iterations, preserve functional behavior through formal equivalence, and improve power, performance, and area. This partnership advances our workload-native platforms for Physical AI by giving customers a more efficient path from workload definition to high-quality, differentiated silicon.”

In work on MIPS production-scale design, Xcelsa’s Apex platform autonomously closed a major timing-violating path in a load-store unit in under three hours, reducing critical-path delay by 33 percent. The optimization rewrote a substantial design block and was formally proven equivalent to the reference design. MIPS estimates that the comparable manual workflow would have required approximately one engineer-month.

As intelligent systems move into automotive, industrial, robotics, and other physical environments, semiconductor designers must meet workload-specific requirements for deterministic performance, energy efficiency, functional safety, and real-time response. These constraints create difficult tradeoffs across architecture, software, timing, power, and silicon area. MIPS and Xcelsa are addressing this challenge by connecting workload requirements to verified design optimization, helping designers explore more implementation alternatives and reach differentiated silicon faster.

“Custom silicon is at an inflection point. As AI workloads become more specialized, the industry needs a faster way to move from workload requirements to high-performance, production-ready hardware,” said Srikanth Jagabathula, CEO and Co-Founder of Xcelsa Labs. “Our work with MIPS shows that verified and physical intelligence stacks can together autonomously improve design in hours while formally preserving its intended functionality. By combining Xcelsa’s Verified Compute Intelligence for Hardware Design with MIPS processor IP and software-first approach, we are helping customers explore more design alternatives at AI speed, improve power, performance, and area, and reduce the time required for critical optimization work.”

“Efficiency gains now come from silicon built for the specific workload. The problem is that software iterates in weeks and hardware design cycles run in years,” said Dylan Patel, Founder and CEO of SemiAnalysis. “Closing that gap requires continuous hardware-software co-design, which only works if verification and physical design can keep pace: fast formal verification, and PPA optimization that doesn’t break correctness. The MIPS and Xcelsa partnership will drive the direction the custom silicon market needs to go.”

The partnership extends the MIPS ecosystem for workload-native platforms for Physical AI. It combines MIPS workload-focused design methods and industry-leading open technology portfolio with Xcelsa’s verified compute intelligence capabilities to help customers shorten custom silicon design cycles, improve design quality, and optimize hardware. The resulting custom platforms will be optimized for the performance, power, and real-time requirements of automotive, aerospace, industrial, robotics, communications infrastructure, and other embedded applications.

For more information, visit https://mips.com/.

About MIPS, by GF

MIPS, by GF, develops inference-native and safety-capable platforms for energy-efficient Physical AI systems and embedded devices. MIPS, now combined with ARC processor solutions, delivers software-first platforms for automotive, aerospace, robotics, IoT, and other embedded markets. Based on the open RISC-V instruction set architecture, the MIPS portfolio spans processor IP, virtual platforms, custom application-specific silicon, and Physical AI platforms, backed by GlobalFoundries process technology, packaging, and manufacturing. For more information, visit MIPS.com.

About Xcelsa Labs

Xcelsa Labs is pioneering verified compute intelligence to enable high-performance, energy-efficient silicon for the AI age. Its flagship Apex platform combines fast verified and physical intelligence stacks to autonomously optimize custom chip designs, improving power, performance, and area while formally guaranteeing functional equivalence to reference designs. Founded by leading researchers in AI, mathematics, and chip design, Xcelsa helps semiconductor teams compress critical optimization work from weeks to hours and accelerate silicon development for AI, automotive, industrial, robotics, and edge applications. For more information, visit xcelsa.ai.

Media gallery

About The Author
Business Wire