The New Challenger in GPU Computing: OXMIQ's Bid to Redefine Accelerator Architecture

Raja Koduri, the architect behind revolutionary GPU designs at Apple, AMD, and Intel, has unveiled his most ambitious project yet: OXMIQ Labs. After two years operating in stealth, the startup emerges with $20 million in seed funding from investors including MediaTek and a mission to fundamentally reimagine GPU architecture for the multimodal AI era. Koduri has assembled an all-star team boasting over 500 collective years of GPU design experience and a track record generating $100B+ in prior revenue—a brain trust now targeting the computational limitations of modern AI workloads.

The Multimodal Imperative

Modern computing increasingly demands seamless integration of text, audio, video, images, and 3D environments—a shift rendering traditional fixed-function accelerators inadequate. "Unlike specialized AI chips," Koduri's team explains, "GPUs provide the general-purpose flexibility required for these diverse modalities while maintaining deep OS integration." This architectural advantage positions GPUs as essential infrastructure for next-generation AI, where heterogeneous workloads must coexist. Yet current designs struggle with efficiency at scale, creating an opening for OXMIQ's radical approach.

Breakthrough IP: From Nano-Agents to Light Transport

OXMIQ's licensable GPU IP rethinks silicon from first principles:

• Nano-agents: RISC-V cores embedded directly in silicon for fine-grained workload orchestration
• OXCORE™: Modular architecture integrating scalar, vector, and tensor engines with Python acceleration
• OXQUILT™: Chiplet-based scaling from edge devices to zettascale datacenters
• Light transport: Advanced photonics for accelerated data movement

This hardware foundation enables what Koduri terms "Physical AI"—systems capable of processing real-world sensory data at unprecedented speeds for robotics, autonomous systems, and immersive environments.

Software First: The CUDA Compatibility Play

OXMIQ's most immediate disruption comes through OXPython™—a flagship component of their OXCapsule™ software ecosystem. In a direct challenge to NVIDIA's dominance, OXPython allows Python-based CUDA applications to run unmodified on non-NVIDIA hardware. Tenstorrent CEO Jim Keller confirms the strategic partnership:

"OXPython’s ability to bring Python workloads for CUDA to AI platforms like Wormhole and Blackhole is great for developer portability. It aligns with our goal of letting developers open and own their entire AI stack."

OXCapsule abstracts hardware complexity through a unified API layer, enabling frictionless deployment across heterogeneous systems. This software-centric approach already generates revenue, with initial deployment on Tenstorrent platforms later this year and multiple vendor integrations underway.

Capital Efficiency and Industry Endorsements

Unlike capital-intensive chip startups, OXMIQ's IP licensing model avoids massive EDA and fabrication costs. MediaTek SVP Lawrence Loh highlights the strategic value:

"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."

Koduri has transitioned full-time from Mihira Visual Labs (the cinematic AI venture he co-founded) to lead OXMIQ, retaining an advisory role and minority stake. Mihira CEO Shobu Yarlagadda notes the symbiotic relationship: "As we scale into a global creative platform, we’re thrilled to integrate OXMIQ’s agentic GPU innovations into our storytelling stack."

The New Flexibility Frontier

OXMIQ arrives at an inflection point where AI's computational demands increasingly strain conventional architectures. By decoupling software ecosystems from hardware silos and rethinking GPU design at fundamental levels, Koduri's team isn't just building another accelerator—they're attempting to redefine how computational flexibility scales across the entire computing spectrum. For developers drowning in multi-platform porting headaches and enterprises seeking alternatives to monolithic vendor lock-in, OXMIQ's promise of truly heterogeneous acceleration could prove revolutionary. The success of their ambitious vision now hinges on execution—but with Koduri's pedigree and heavyweight backing, the GPU landscape may soon face its most credible disruption in decades.

Source: OXMIQ Labs Press Release