
Oxmiq Labs Inc.™: Revolutionizing GPU Architecture for the Multimodal AI Era
After two years of operating in stealth mode, Oxmiq Labs Inc.™, a groundbreaking GPU software and IP startup founded by renowned GPU architect and visionary Raja Koduri, has officially emerged into the spotlight. Backed by a world-class team of GPU and AI architects with over 500 years of combined experience, hundreds of patents, and a collective track record of generating more than $100 billion in revenue at prior companies, Oxmiq is poised to redefine the future of computing. The company’s mission? To re-architect the GPU stack from the ground up—leveraging cutting-edge innovations in hardware and software to meet the demands of modern multimodal AI workloads.
The Shift to Multimodal Computing
Modern computing has undergone a fundamental transformation, moving toward multimodal experiences where text, audio, video, images, and 3D environments interact seamlessly. This shift has positioned the GPU as the cornerstone of the computational landscape. Unlike fixed-function AI accelerators designed for specific tasks, GPUs offer unparalleled general-purpose computational flexibility. They maintain deep integration with mainstream operating systems through standardized APIs and unified memory models, making them indispensable for both current applications and the emerging era of multimodal AI, where diverse workloads must be processed harmoniously.
Oxmiq Labs addresses this opportunity head-on with its licensable GPU IP, reimagining GPU architecture from first principles. By incorporating breakthrough technologies such as nano-agents embedded in silicon using RISC-V cores, near-memory and in-memory computing, and light transport innovations, Oxmiq delivers solutions that balance multimodal computing flexibility with the radical performance improvements required for next-generation graphics and AI workloads.
A Software-First Approach
Drawing on decades of industry evolution, Oxmiq embraces a Software First strategy that prioritizes developer experience. At the heart of this approach is OXCapsule™, the company’s unified software ecosystem, which abstracts away hardware complexity to provide frictionless deployment across diverse computing platforms. This eliminates the configuration challenges that traditionally plague heterogeneous environments, empowering developers to focus on innovation rather than compatibility issues.
A standout component of OXCapsule is OXPython™, a revolutionary tool that enables Python-based NVIDIA® CUDA™ AI applications to run seamlessly on non-NVIDIA hardware without requiring code modification or recompilation. Set to launch initially on Tenstorrent’s AI platform later this year, with multiple vendor integrations underway, OXPython exemplifies Oxmiq’s commitment to breaking down hardware silos and democratizing high-performance computing across the industry.
“We’re excited to partner with Oxmiq on their OXPython software stack,” said Jim Keller, CEO of Tenstorrent™. “OXPython’s ability to bring Python workloads for CUDA to AI platforms like Wormhole™ and Blackhole™ is a game-changer for developer portability and ecosystem expansion. It aligns perfectly with our vision of letting developers own their entire AI stack.”
The Oxmiq Breakthrough
Beyond its software innovations, Oxmiq delivers a complete GPU hardware IP stack designed to power silicon solutions ranging from edge devices and autonomous robots to enterprise infrastructure and zettascale data centers. Central to this offering is OXCORE™, the company’s scalable GPU core, which integrates scalar, vector, and tensor compute engines in a modular architecture customizable for specific workloads. OXCORE supports nano-agents, native Python acceleration, and compatibility with SIMD/CUDA paradigms, ensuring versatility across applications.
Through OXQUILT™, Oxmiq’s chiplet architecture, customers can configure optimal ratios of compute, memory, and interconnect to suit their needs. This modular approach significantly reduces time-to-market, R&D costs, and production expenses compared to traditional methodologies, enabling rapid innovation and scalability.
Capital Efficiency and Market Traction
Oxmiq has raised $20 million in seed funding from prominent technology investors, including strategic players like MediaTek®, a leader in mobile and AI silicon. The company has already recorded its first round of software revenue, demonstrating strong market traction. By adopting a licensing-first model, Oxmiq avoids the heavy capital requirements typical of chip startups dependent on expensive EDA tools and physical tape-outs, achieving remarkable capital efficiency.
“Oxmiq has an impressive bold vision and world-class team,” said Lawrence Loh, SVP of MediaTek. “The company’s GPU IP and software innovations will drive a new era of compute flexibility across devices—from mobile to automotive to AI on the edge.”
Collaboration with Mihira Visual Labs™
Raja Koduri, who co-founded Mihira™ two years ago alongside Shobu Yarlagadda and SS Rajamouli, has transitioned out of day-to-day operations at Mihira to focus full-time on leading Oxmiq. He now serves as a strategic advisor to Mihira Visual Labs™, which is led by CEO Shobu Yarlagadda. Oxmiq holds a minority stake in Mihira and will continue to support its growth by providing foundational agentic and GPU IP technologies that power Mihira’s cinematic AI platform.
“Raja’s early contributions to Mihira helped shape our foundational vision of cinematic AI,” said Shobu Yarlagadda, Co-Founder and CEO of Mihira Visual Labs™, and acclaimed producer of the Baahubali film series. “Now, with Oxmiq, he and his team are building the deep-tech infrastructure that powers our next chapter. As we scale Mihira into a global creative platform, we’re thrilled to continue our close collaboration with Oxmiq and integrate their agentic GPU innovations into our storytelling stack.”



