Can OXMIQ’s Open Architecture Break the AI Silicon Monopoly?

CAMPBELL, CA – July 01, 2026 – In the global rush to build the infrastructure of artificial intelligence, a single, formidable bottleneck has emerged: the prohibitive cost and proprietary nature of high-performance compute. Now, a new venture led by one of the semiconductor industry’s most respected architects is making a bold play to dismantle that bottleneck. OXMIQ Labs, founded by chip veteran Raja Koduri, has closed a $35 million Series A funding round, bringing its total capital to $60 million. Its mission is not merely to build a better chip, but to re-architect the entire system of how AI silicon is designed, built, and deployed.

With backing from a syndicate of strategic heavyweights including Samsung Catalyst Fund, MediaTek, and AM Intelligence Labs, OXMIQ is positioning itself as a foundational alternative in a market dominated by a few incumbents. The company’s licensable architecture model aims to empower a new wave of innovation by making custom AI silicon accessible to the many, not just the few with ten-figure R&D budgets. This isn't just a business plan; it's a structural challenge to the centralized nature of modern technological power.

A New Blueprint for Compute

At the heart of OXMIQ’s strategy is OxCore™, a licensable GPU core that fundamentally rethinks how processing is done. Instead of relying on disparate chips for different tasks, OxCore integrates three distinct compute engines into a single, cohesive unit: a CUDA®-compatible GPU engine for graphics and parallel processing, a dedicated tensor processing engine for AI workloads, and an orchestration engine (CPU) to manage tasks and agents across the system. This tight coupling is purpose-built for near-memory compute, a design that minimizes the costly and energy-intensive process of moving data around. The result is a blueprint for silicon that promises greater efficiency and scalability, from single-core edge devices to massive datacenter deployments.

This core IP is complemented by OxQuilt™, a chiplet integration architecture that introduces a radical level of flexibility. Today, most AI silicon designs are locked into a specific foundry, memory type, and packaging technology. OXMIQ flips that model. As Rajeev Surati, Partner at co-lead investor Fundomo, noted, “Most compute IP makes the customer bend their memory, packaging, and foundry around the chip. OXMIQ does the opposite, and that flips a cost center into leverage.” OxQuilt allows customers to design across different process nodes, memory types like HBM or GDDR, and interconnect standards. This modular, adaptable approach could fundamentally alter the economics of chip design, allowing companies to build custom AI silicon without the astronomical cost of a full, ground-up chip program.

Breaking the Software Lock-In

A brilliant hardware architecture is of little use if no one can write software for it. This is the moat that has protected NVIDIA’s market dominance for years: its CUDA® software platform. OXMIQ is tackling this challenge head-on with OxPython™, a software layer designed to run existing CUDA® and PyTorch® code on its hardware without any changes. This is a critical move for democratization. It promises developers full portability, allowing them to leverage a vast ecosystem of existing code and talent while freeing them from hardware lock-in.

This commitment to an open ecosystem is further strengthened by the addition of legendary chip architect Jim Keller, CEO of Tenstorrent, to OXMIQ's board of directors. Keller’s endorsement is a powerful signal to the industry. “Raja and this team are creating an open GPU architecture, a much-needed step toward removing the artificial boundaries around AI innovation,” Keller stated. “As the industry concentrates around a few incumbents, this is more important than ever. OXMIQ’s open, configurable foundation, which developers can build on and own, is exactly where compute should be heading.” By building its architecture on the open RISC-V instruction set and providing a seamless software bridge, OXMIQ is laying the tracks for a parallel ecosystem that thrives on choice, not constraint.

The Architect's Grand Vision

To understand OXMIQ is to understand Raja Koduri. With a storied career that includes leading GPU and architecture divisions at Apple, AMD, and Intel, Koduri has seen the inner workings of the world’s most advanced technology firms. His vision for OXMIQ is a direct response to the structural problems he has witnessed firsthand. He argues that the current cost of AI infrastructure—where a single hyperscale data center can approach $50 billion—is unsustainable and inherently limits who can participate in the AI revolution.

“A licensable core with an open architecture means design teams everywhere can build the custom AI silicon their work needs,” Koduri explained. “Today, state-of-the-art AI reaches most people through a handful of channels, and the cost of the compute underneath is the reason.” His mission is to drive that cost down, widening access and enabling a more distributed, equitable landscape for innovation. He draws a parallel to ARM’s licensing model, which enabled companies like Apple to create custom chips that ultimately made the mobile internet affordable for billions. Koduri is betting he can do the same for AI.

The Smart Money Signals a Shift

The composition of OXMIQ’s investors reveals a broader industry preparing for a seismic shift. This is not just speculative venture capital; it is strategic capital from across the supply chain. Co-lead investor Samsung Catalyst Fund sees the potential for “efficient, custom inference solutions serving large-scale agentic workloads,” a clear nod to integrating this technology into its vast ecosystem of consumer and enterprise products. MediaTek, a powerhouse in mobile and edge silicon, has reinvested, with SVP Lawrence Loh stating that the partnership will help unlock “unprecedented on-device AI performance.”

Perhaps most telling is the involvement of AM Intelligence Labs, which is partnering with OXMIQ on a massive 5GW AI factory initiative in India. This isn't just an investment; it's a customer engagement that provides a direct path for OXMIQ’s architecture to be deployed at a monumental scale. With other investors like Pegatron (manufacturing), CDIB-TEN (semiconductor ecosystem), and Intel Capital (strategic IP partner) also on board, the message is clear: the industry is actively seeking viable, flexible, and cost-effective alternatives to the current AI hardware oligopoly. They are betting that OXMIQ holds the key.