Quadric’s On-Device AI Shift Pays Off Big
▼ Summary
– Companies and governments are seeking to run AI locally to reduce cloud costs and build sovereign capability, creating a market for on-device AI technology.
– Quadric, a chip-IP startup, is expanding from automotive into laptops and industrial devices, with its licensing revenue growing significantly from $4 million in 2024 to a projected $35 million in 2025.
– The company recently raised $30 million in a Series C round, reflecting investor interest in shifting AI workloads from centralized cloud infrastructure to devices and local servers.
– Quadric licenses programmable AI processor IP and a software stack, allowing customers to embed the technology into their own silicon to run models on-device, rather than manufacturing chips itself.
– The startup is targeting markets like India and Malaysia for sovereign AI strategies and pitches its programmable IP as a flexible alternative to vendors like Qualcomm or traditional IP suppliers.
The growing demand for on-device AI processing is creating significant opportunities for companies that enable local inference, moving workloads away from expensive cloud infrastructure. Quadric, a processor intellectual property startup, is capitalizing on this trend with a business model that has recently accelerated its financial trajectory. The company reported a substantial jump in licensing revenue, reaching between $15 million and $20 million in 2025, a notable increase from approximately $4 million the previous year. This performance supports a current post-money valuation estimated between $270 million and $300 million, a rise from around $100 million during its 2022 Series B funding round.
This momentum recently attracted a $30 million Series C investment led by the ACCELERATE Fund, managed by BEENEXT Capital Management. The financing brings Quadric’s total funding to $72 million. According to CEO Veerbhan Kheterpal, this investor interest aligns with a broader industry push to transition AI workloads from centralized cloud systems to devices and local servers.
Originally focused on the automotive sector for applications like real-time driver assistance, Quadric has expanded its reach. Kheterpal notes that the proliferation of transformer-based models created a sharp business inflection point, driving demand for local AI execution across diverse industries. The company’s strategy involves licensing programmable AI processor IP, essentially a blueprint that clients integrate into their own chips, alongside a complete software stack for on-device model deployment.
Unlike chip manufacturers such as Nvidia or Qualcomm, Quadric does not fabricate its own silicon. Instead, it provides a chip-agnostic, programmable architecture. This approach is designed to offer flexibility, allowing customers to adapt to new AI models through software updates rather than undergoing lengthy and costly hardware redesigns. The company positions this as an advantage over both integrated chip vendors and traditional IP suppliers offering fixed-function neural processing blocks.
Current customers include firms in the printer, automotive, and AI laptop sectors, such as Kyocera and automotive supplier Denso. The first consumer products incorporating Quadric’s technology, starting with laptops, are anticipated to ship this year.
Looking ahead, Quadric is targeting markets pursuing sovereign AI strategies, aiming to reduce dependence on U.S.-based cloud infrastructure. The company is exploring opportunities in countries like India and Malaysia, citing rising cloud costs and the challenges of building hyperscale data centers as key drivers. This shift toward distributed AI, where inference occurs on local devices or small on-premise servers, is gaining recognition among policymakers and industry analysts focused on building domestic AI capabilities.
While the company’s early growth is promising, its long-term success hinges on converting initial licensing agreements into high-volume product shipments and sustained royalty streams. The broader industry challenge remains the rapid evolution of AI models, which outpaces traditional hardware development cycles, underscoring the potential value of a software-updatable, programmable processor foundation.
(Source: TechCrunch)
