Raja Koduri's Oxmiq Labs Unveils Revolutionary GPU Technology for AI and Graphics Computing

Legendary GPU architect Raja Koduri's new startup leverages RISC-V and targets CUDA workloads -- Oxmiq Labs supports running Python-based CUDA applications unmodified on non-Nvidia hardware

Another startup developing GPUs that are not meant for graphics emerges from stealth mode. Raja Koduri, a legendary GPU architect from ATI Technologies, AMD, Apple, and Intel, on Tuesday said he had founded a new GPU startup that emerged from stealth mode today. Oxmiq Labs is focused on developing GPU hardware and software IP and licensing them to interested parties. In fact, software may be the core part of Oxmiq's business as it is designed to be compatible with third-party hardware. Oxmiq develops a vertically integrated platform that combines GPU hardware IP with a full-featured software stack aimed at AI, graphics, and multimodal workloads where explicitly parallel processing is beneficial. On the hardware side, Oxmiq offers a GPU IP core based on the RISC-V instruction set architecture (ISA) called OxCore, which integrates scalar, vector, and tensor compute engines in a single modular architecture and can support near-memory and in-memory compute capabilities. Oxmiq also offers OxQuilt, a chiplet-based system-on-chip (SoC) builder that enables customers to create their own SoCs that integrate compute cluster bridge (CCB, which probably integrates OxCores), memory cluster bridge (MCB), and interconnect cluster bridge (ICB) modules based on specific workload requirements in a rapid and cost-efficient manner. For example, an inference AI accelerator for edge applications can pack a CCB and an ICB or two, an inference SoC requires more CCBs, MCBs, and ICBs, whereas a large-scale SoC for AI training can pack dozens of chiplets. Oxmiq does not disclose whether its OxQuilt enables building only multi-chiplet system-in-packages (SiP), or is designed to assemble monolithic processors too. Oxmiq's software stack is perhaps an even more important product that the company has to offer. The software package is designed to abstract the complexity of heterogeneous hardware and enable deployment of AI and graphics workloads across a range of hardware platforms, not just those using the company's IP. The core of the software stack is OXCapsule, a unified runtime and scheduling layer that manages workload distribution, resource balancing, and hardware abstraction. The layer encapsulates applications into self-contained environments, which the company calls 'heterogeneous containers.' These containers are designed to operate independently of the underlying hardware, enabling developers to target CPUs, GPUs, and AI accelerators without modifying their codebase or dealing with low-level configuration. A standout component of this stack is OXPython, a compatibility layer that translates CUDA-centric workloads into Oxmiq's runtime and allows Python-based CUDA applications to run unmodified on non-Nvidia hardware without recompilation. OXPython will first launch not on Oxmiq's IP, but on Tenstorrent's Wormhole and Blackhole AI accelerators. In fact, Oxmiq's software stack is fundamentally designed to be independent from Oxmiq hardware, and that is a core part of its strategy. "We are 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 great for developer portability and ecosystem expansion. It aligns with our goal of letting developers open and own their entire AI stack." Having developed graphics processors at S3 Graphics, ATI Technologies, AMD, Apple, and Intel, Raja Koduri is primarily known as a GPU developer. In fact, he even positions Oxmiq as the first GPU startup in Silicon Valley in decades. "We may be the first new GPU startup in Silicon Valley in 25+ years," wrote Koduri in an X post. "GPUs are not easy." However, it should be noted that Oxmiq is not building a consumer GPU like AMD Radeon or Nvidia GeForce. In fact, it does not develop all the IP blocks necessary to build a GPU, unlike Arm or Imagination Technology: it does not support full consumer graphics features out-of-the-box (such as texture units, render back ends, display pipeline, ray tracing hardware, DisplayPort or HDMI outputs), so Oxmiq licensees must implement them in silicon themselves, if they plan to build a GPU. Oxmiq has secured $20 million in seed funding from major tech investors, including mobile and custom AI silicon developer MediaTek, and has already recorded its first software revenue. By focusing on IP licensing instead of costly chip production or even actual silicon implementation, the company maintains high capital efficiency without relying on expensive EDA tools or tape-outs. "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."

Chip startup Oxmiq launches GPU tech for license

SAN FRANCISCO, Aug 5 (Reuters) - Oxmiq Labs said on Tuesday that it planned to launch licensable graphics processor tech geared for artificial intelligence data crunching. Founded by Intel's (INTC.O), opens new tab former chief architect, Raja Koduri, Oxmiq said that it has raised $20 million in seed capital to help launch the new GPU intellectual property. The funding round includes investments from angel investors, and corporate strategic investors, including MediaTek (2454.TW), opens new tab, Oxmiq said. The company did not disclose its valuation. Oxmiq's GPU technology is capable of scaling from a single core for physical AI applications such as robotics, to thousands of cores that would be useful in a cloud computing company's data center. The company said it can customize the GPU architecture for specific types of computing. "We want to be Arm for the next generation," Koduri told Reuters. The Campbell, California-based company said it was taking a software-first approach to constructing its chip designs and has built a tool to allow software programs written for Nvidia's (NVDA.O), opens new tab CUDA to work on non-Nvidia hardware "without code modification of recompilation." The company said it opted to pursue building intellectual property instead of a complete chip design because it would avoid the high costs. A cutting-edge chip can cost more than $500 million to design. At Intel, Koduri oversaw the development of the company's graphics chips. Koduri has held senior positions at Advanced Micro Devices (AMD.O), opens new tab and Apple (AAPL.O), opens new tab. Reporting by Max A. Cherney in San Francisco; Editing by Mrigank Dhaniwala Our Standards: The Thomson Reuters Trust Principles., opens new tab * Suggested Topics: * Technology Max A. Cherney Thomson Reuters Max A. Cherney is a correspondent for Reuters based in San Francisco, where he reports on the semiconductor industry and artificial intelligence. He joined Reuters in 2023 and has previously worked for Barron's magazine and its sister publication, MarketWatch. Cherney graduated from Trent University with a degree in history.

Ex-Intel and AMD graphics chief Raja Koduri promises a RISC-V based IP that 'rearchitects the GPU from first principles' but with few details in the startup word-salad

Tech startups are nothing new, but this one might stand a chance of making it because of who's behind it. These days, it seems like there's a new tech startup company appearing every day, promising to revolutionise whatever market it's targeting. Oxmiq Labs Incorporated is a name that few will have heard of right now, but its founder is practically a household name in the world of processors, and Raja Koduri's latest venture is making some big claims. Admittedly, those claims are so packed full of tech buzzwords that I had to reread them all a dozen times to get any sense of what makes Oxmiq Labs different from any other chip startup. "Oxmiq delivers solutions that balance multimodal computing flexibility with the radical performance improvements required for next-generation graphics and AI workloads for its customers," says the press announcement at Business Wire. It gets better: "Oxmiq's licensable GPU IP rearchitects the GPU from first principles incorporating breakthrough technologies including nano agents in silicon leveraging RISC-V cores, near-memory and in-memory computing, and light transport." Got all that? Good. Unfortunately, there's not a whole lot of additional, concrete information with which to explore whether any of those claims have any weight to them. Oxmiq has a single diagram to explain its new GPU IP (intellectual property), but even that's pretty vague. At the heart is the OxCore, of which almost nothing is said, other than it's based on the RISC-V ("risk-five") architecture and that it "integrates scalar, vector, and tensor compute engines in a modular architecture customizable for specific workloads, enabling nano-agents, native Python acceleration, and compatibility with SIMD/CUDA paradigms." These cores are designed to be stacked on top of a slice of relatively slow but spacious DRAM or relatively fast but small SRAM, with the resulting chiplets labelled as MCBs and CCBs in the diagram. What do the letters stand for, you might ask? That's not explained, but I'd guess they're something along the lines of memory/cache core something. Bonds, badgers, burgers, boffins -- not a clue. Anyway, each MCB/CCB is a complete SoC (system-on-chip), with interconnect bridges to allow them to be hooked up to other chiplets. Oxmiq's diagram shows two such examples: a MCB-heavy variant for doing AI inference and MCB/CCB-balanced version for AI training. Despite the press statement mentioning "next-generation graphics", this product absolutely is not for gaming PCs, sadly -- it's for massive data centers, to process AI or scientific data. To compete against AMD, Intel, and Nvidia, any hardware startup needs to have a comprehensive and powerful software stack, and to that end, Oxmiq has developed OXPython. This "enables Python-based Nvidia CUDA AI applications to execute seamlessly on non-NVIDIA hardware without code modification or recompilation." I'm not sure Nvidia would be best pleased about this, but another famous tech person startup certainly is: Jim Keller. His Tenstorrent company -- which also makes RISC-V processors for AI -- has teamed up with Oxmiq: "We're excited to partner with Oxmiq. 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." Keller and Koduri both have very long careers in the semiconductor industry, with Keller spending many years at AMD, leading the development of the Athlon K8 and Ryzen Zen architectures. Koduri also spent a long time at AMD, working in its graphics division, leading the development of Polaris, Vega (below), and Navi. He also spent time with Intel, with the Arc series of GPUs coming under his leadership. In short, both know what they're doing, and their names will certainly help in no small way with attracting investors. Tenstorrent does have one advantage over Oxmiq, though. It already has hardware available to buy, whereas Koduri's startup appears to be focused on IP only, i.e. it'll be down to another company to license the tech and make the chips themselves. That's more cost-effective than having to purchase fabrication slots at the likes of TSMC, but as Arm will testify, processor IP licensing isn't necessarily an easy path to big bucks. MediaTek has already invested in Oxmiq, though how much of the $20 million in seed funding it contributed to is unknown. Koduri's company claims to have already generated revenue from its software (Tenstorrent will certainly be responsible for some of this) and that area might ultimately be its biggest earner. The G in GPU might stand for graphics, but the money just doesn't lie in gaming anymore (and hasn't for a good few years). Calling them AIPUs might seem like a silly idea, but at least it would stop anyone from unrealistically hoping that Nvidia's near-monopoly of the gaming GPU industry was under threat from a potent startup. The possibility of Oxmiq's 'rearchitected GPUs' prising AI revenue away from Team Green might be a little more believable, though.

Chip startup Oxmiq launches GPU tech for license - The Economic Times

Founded by Intel's former chief architect, Raja Koduri, Oxmiq said that it has raised $20 million in seed capital to help launch the new GPU intellectual property.Oxmiq Labs said on Tuesday that it planned to launch licensable graphics processor tech geared for artificial intelligence data crunching. Founded by Intel's former chief architect, Raja Koduri, Oxmiq said that it has raised $20 million in seed capital to help launch the new GPU intellectual property. The funding round includes investments from angel investors, and corporate strategic investors, including MediaTek, Oxmiq said. The company did not disclose its valuation. Oxmiq's GPU technology is capable of scaling from a single core for physical AI applications such as robotics, to thousands of cores that would be useful in a cloud computing company's data center. The company said it can customize the GPU architecture for specific types of computing. "We want to be Arm for the next generation," Koduri told Reuters. The Campbell, California-based company said it was taking a software-first approach to constructing its chip designs and has built a tool to allow software programs written for Nvidia's CUDA to work on non-Nvidia hardware "without code modification of recompilation." The company said it opted to pursue building intellectual property instead of a complete chip design because it would avoid the high costs. A cutting-edge chip can cost more than $500 million to design. At Intel, Koduri oversaw the development of the company's graphics chips. Koduri has held senior positions at Advanced Micro Devices and Apple.

GPU Visionary Raja Koduri Launches OXMIQ Labs to Redefine AI and Graphics Computing

OXMIQ takes a Software First approach, focusing on developer experience through a unified stack that works across both OXMIQ silicon and third-party GPU and AI platforms. Raja Koduri, announced the launch of OXMIQ Labs Inc., a new GPU software and IP powerhouse set to transform AI and graphics computing globally. OXMIQ emerges from two years of stealth R&D with a world-class team holding 500+ years of combined GPU & AI experience, hundreds of patents, and over $100B in past product revenue impact. The company's mission: to re-architect GPUs for the multimodal AI era, where text, images, audio, video, and 3D worlds work together seamlessly.