Skip to content
Chimera readability score 62 out of 100, Academic reading level.

Last Updated: 1 July 2026
OXMIQ Labs, founded by GPU architect Raja Koduri, develops licensable chip designs for companies that want to build their own custom AI silicon. The company does not manufacture finished processors. Instead, it creates the underlying GPU architecture and software and licenses both to semiconductor firms and AI system builders, giving them a way to develop custom AI chips without the cost of a full chip program.
Demand for custom AI silicon is growing as more companies try to reduce their reliance on a single GPU supplier. According to TrendForce, custom ASIC shipments from cloud providers are projected to grow 44.6% in 2026, nearly three times the 16.1% growth rate expected for GPU shipments. AI chip startups raised $8.3 billion globally in 2026 according to Dealroom. Investors are treating inference costs and supply chain concentration as problems that call for new hardware approaches.
OXMIQ has closed a $35 million Series A round, bringing total capital raised to $60 million. Fundomo and Samsung Catalyst Fund co-led the round. MediaTek, AM Intelligence Labs, Pegatron Venture Capital, CDIB-TEN, Darwin Ventures, Morgan Creek Digital, and Intel Capital also participated.
NVIDIA still holds approximately 70 percent of the AI chip market, but that share has been gradually declining as hyperscalers and semiconductor companies invest in custom silicon alternatives. Inference economics and the desire to control their own compute roadmaps are driving this shift. The barrier is steep, though. Building a competitive GPU from scratch takes billions of dollars, years of engineering, and a software ecosystem mature enough to attract developers. OXMIQ's approach is to license the core GPU IP so that other companies can skip that upfront cost.
Two board and advisory appointments came alongside the funding. Jim Keller, CEO of Tenstorrent and a widely recognized chip architect, has joined OXMIQ's board of directors. Dr. Valluri Rao, a retired Intel Fellow from the company's process technology group, joins as an advisor.
"I am excited to join the OXMIQ board. Raja and this team are creating an open GPU architecture, a much-needed step toward removing the artificial boundaries around AI innovation. 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."
Jim Keller, CEO of Tenstorrent and OXMIQ board member
The capital will go toward scaling OxCore, the company's GPU architecture, and moving it from validation to customer integration. OxCore is already running on FPGA, with live demonstrations available. OXMIQ follows an IP-first model, generating revenue from customer engagements while preserving capital for building the rest of the stack rather than pursuing full SoC development.
Funding will also support OXMIQ's broader product lineup. OxQuilt handles chiplet integration across different foundries, memory types, and packaging options. OxPython lets developers run existing CUDA and PyTorch code on OxCore without modifying anything. OxCapsule manages workload orchestration. Together, these tools are meant to offer design teams a complete path from licensed IP to working AI compute.
"Raja has built silicon at every layer of the stack, and he knows exactly where the constraints sit. 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. We backed this team because they will define how AI compute gets built this decade."
Rajeev Surati, Partner at Fundomo
OXMIQ Labs was founded in 2023. It is headquartered in Campbell, California, with a development site in Hyderabad, India. Founder and CEO Raja Koduri spent nearly three decades working in GPU architecture at senior leadership positions. At AMD, he led the Radeon Technologies Group and developed the Polaris, Vega, and Navi architectures, which shipped across PCs, Macs, and major game consoles. At Intel, he served as Executive Vice President and Chief Architect, leading the development of Intel Arc and Ponte Vecchio, the industry's first peta-scale GPU, assembled from 47 chiplets.
OXMIQ makes money by licensing its designs, not by selling chips. OxCore, its main product, is a scalable GPU core that integrates three compute engines into one: a CUDA-compatible GPU engine, a tensor processing engine, and an orchestration engine that coordinates workloads. In most AI chip designs, these three functions sit on separate chips. OxCore combines them in a single core built for near-memory compute, which means data travels shorter distances and the system uses less energy. The architecture works at any size, from single-core deployments to full datacenter configurations.
"A licensable core with an open architecture means design teams everywhere can build the custom AI silicon their work needs. Today, state-of-the-art AI reaches most people through a handful of channels, and the cost of the compute underneath is the reason. Bring that cost down, and you widen who gets to build with it. I believe AI is a force for good when it is a tool everyone can pick up and use, not just the few who can afford to build with it. Closing this round with investors who own the supply chain tells us we can get there."
Raja Koduri, Founder and CEO of OXMIQ Labs
The round was co-led by Fundomo, a New York venture firm focused on frontier compute infrastructure, and Samsung Catalyst Fund, Samsung Electronics' evergreen multi-stage venture capital fund investing in deep tech AI infrastructure. MediaTek, which invested in OXMIQ at the seed stage, reinvested in this round. Intel Capital participated as a strategic IP partner.
The remaining investors cover manufacturing, supply chain, and regional expertise. AM Intelligence Labs, part of the AM Green Group, is extending a collaboration behind a 5GW AI factory initiative in India. Pegatron Venture Capital brings systems integration and ODM manufacturing scale. CDIB-TEN, a joint fund between CDIB Capital and TEN Capital Corporation, is deeply connected to Taiwan's semiconductor ecosystem. Darwin Ventures invests across the Taiwan-Silicon Valley technology corridor, and Morgan Creek Digital backs the round on the thesis that compute capacity and architecture choice will define the AI economy this decade. Before this round, OXMIQ had raised $25 million.
"MediaTek is actively powering today's advanced AI capabilities from the edge to the cloud. Our investment in OXMIQ underscores this push and combines our AI ambitions with their highly flexible GPU architecture. We see this investment as a way to continue unlocking unprecedented on-device AI performance across all technology platforms."
Lawrence Loh, SVP of MediaTek

Sentinel — Human

Confidence

The text demonstrates high-quality, structured writing that effectively synthesizes complex financial and technical information, leaning toward professional human authorship rather than generic machine generation.

Signals Detected
low severity: High lexical diversity and sophisticated sentence structure; transitions are smooth but predictable.
low severity: Excellent flow; the narrative successfully integrates business context, technical specifics, and founder philosophy without mechanical balancing.
low severity: Specific financial figures (e.g., $8.3B raised, 44.6% growth) are explicitly attributed to named sources (TrendForce, Dealroom), indicating reliance on verifiable external data.
low severity: Quotes appear contextually authentic and align with the known roles of the speakers; no obvious LLM confabulation detected in historical or financial claims.
Human Indicators
The integration of highly specific, disparate entities (chip architects, venture capital funds, semiconductor history) into a cohesive narrative suggests deep domain expertise typically found in human reporting.
The nuance in the quotes reflects an understanding of the underlying conflict—cost vs. control—rather than simply summarizing facts.