Companies and governmental organizations are increasingly seeking solutions for localizing AI processing to reduce cloud infrastructure costs and bolster domestic capabilities. Quadric, a chip-design startup established by former executives from the pioneering bitcoin mining firm 21E6, is poised to facilitate this transformation with its innovative on-device inference technology.
Exploring Beyond Automotive Applications
Quadric initially focused on the automotive sector, where on-device AI enhances real-time functionalities like driver assistance. However, according to CEO Kheterpal, the surge of transformer-based models in 2023 has catalyzed a shift toward integrating AI into a broader range of applications, leading to a significant business growth trajectory over the past 18 months. More companies are opting for decentralized AI solutions instead of relying on cloud services.
“As Nvidia dominates data-center AI, we aim to develop a comparable infrastructure for on-device AI,” Kheterpal stated. Unlike Nvidia, Quadric does not manufacture chips but licenses programmable AI processor intellectual property (IP). This technology serves as a “blueprint” for customers, who can incorporate it into their own silicon, complete with a software suite for running models such as visual and auditory processing on their devices.
Strategic Expansion into Sovereign AI Markets
Quadric is now venturing beyond conventional commercial applications, targeting markets implementing “sovereign AI” strategies to reduce dependency on U.S.-based technology. Kheterpal mentioned that the startup is exploring client relationships in India and Malaysia, supported by strategic investor Rahul Garg, CEO of Moglix, who aids in shaping Quadric’s approach in the Indian market. The company employs nearly 70 staff globally, with significant operations in both the U.S. and India.
The motivation behind this shift stems from escalating costs associated with centralized AI infrastructures and the challenges many nations face in establishing large-scale data centers. Kheterpal pointed out the increasing interest in “distributed AI” frameworks where inference computations can be executed on laptops or smaller on-site servers instead of relying solely on cloud platforms.
Industry Perspective and Market Trends
The World Economic Forum recently highlighted this trend, noting that AI inference is moving closer to end users, transitioning from purely centralized systems. Similarly, a November report by EY indicated that the concept of sovereign AI is gathering momentum as governmental bodies and industry organizations strive for local AI capabilities encompassing compute resources, models, and data, moving away from complete dependency on foreign infrastructure.
Yet, challenges persist for chip manufacturers as AI models evolve at a pace that often outstrips current hardware design cycles. According to Kheterpal, the industry requires programmable processor IP that can adapt through software changes, avoiding expensive hardware redesigns with each architectural shift.
Positioning Itself as an Alternative
Quadric positions itself as a viable alternative to established chip manufacturers like Qualcomm, which typically embed their AI technology within their own processors. Additionally, it distinguishes itself from IP suppliers like Synopsys and Cadence, which provide neural processing engine blocks that present programming challenges for many customers. Quadric’s programmable approach enables clients to adapt to new AI models through software updates, providing a critical advantage in a rapidly evolving industry where chip development can take years.
Although Quadric is still in the early stages of its expansion, it has secured a handful of customers so far. The future growth potential hinges on converting current licensing agreements into high-volume shipments and sustainable royalty streams.
