Raja Koduri Returns to GPU Industry with Oxmiq Labs Startup, Leveraging 500 Years of Expertise from Top GPU & AI Architects

In a significant shift within the GPU industry, Raja Koduri, former chief architect at Intel, has embarked on an innovative venture with Oxmiq Labs. His mission: to create advanced AI GPUs from the ground up by assembling a team of top-tier architects.

Raja Koduri Launches Oxmiq Labs: Revolutionizing the GPU Ecosystem with a “Software First”Approach

After nearly two years in retirement from Intel, Raja Koduri has re-entered the tech arena with Oxmiq Labs, a startup dedicated to revolutionizing GPU design and implementation. Koduri previously served as Senior Vice President and Chief GPU Architect at AMD, where he honed expertise that he now channels into this new project.

As highlighted in their official announcement, Oxmiq Labs boasts a staggering collective 500 years of experience among its team members and plans to leverage hundreds of patents in the development of new AI hardware. Rather than merely enhancing existing designs, the startup is committed to an ambitious goal: entirely rebuilding the GPU ecosystem with a focus on ‘Atoms to Agents’.This strategy emphasizes low-level hardware innovation while advancing high-level AI capabilities.

The rise of multimodal AI capabilities necessitates GPUs that can process diverse data forms—text, images, and video—intelligently and simultaneously. Unlike AI chips designed for specific tasks, Oxmiq’s architecture is engineered to harness the inherent flexibility of GPUs, driving improvements in performance and efficiency.

A cornerstone of Oxmiq’s hardware innovation is the OXCORE, a modular GPU core designed to combine Scalar, Vector, and Tensor units for varied operations. This architecture will facilitate support for nano agents and enable accelerated processing for Python and CUDA/SIMD compatibility, scaling seamlessly from compact devices to extensive data center environments.

Moreover, the OXQUILT chiplet architecture will allow for the customization of GPUs, optimizing the combination of compute, memory, and interconnect components. This adaptability is expected to drastically reduce both research and development time, as well as manufacturing costs, compared to traditional GPU designs. Central to Oxmiq’s vision is a “Software First”strategy aimed at developing an intuitive software stack that integrates with both its proprietary GPU technology and third-party AI hardware, like Tenstorrent.

We’re excited to partner with OXMIQ on their OXPython software stack,

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

– Jim Keller, CEO of Tenstorrent

Oxmiq is also introducing a unified software ecosystem known as OXCapsule, designed to alleviate complexities in hardware deployment, allowing developers to seamlessly implement AI applications across multiple platforms.

OXMIQ has an impressive bold vision and world-class team,

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.

– Lawrence Loh, SVP of MediaTek

Additionally, Oxmiq’s strategy centers around a licensing-first model, which allows the company to offer its intellectual property licenses without the need to develop complete GPUs. This approach helps sidestep the costly and time-consuming tape-out process. With $20 million in seed funding from key industry players including Mediatek, the path forward for Oxmiq appears promising. This venture marks a significant milestone as Raja Koduri claims it to be the first GPU startup in Silicon Valley in over 25 years, setting the stage for potentially transformative developments in the tech industry.

Source & Images