Quadric's Shift to On-Device AI Inference Yields Results

Businesses and governments are increasingly seeking tools to run AI on local devices, aiming to reduce cloud infrastructure expenses and develop sovereign capabilities. Quadric, a chip-IP startup founded by veterans from early Bitcoin mining company 21E6, is positioning itself to enable this transition. The company is expanding beyond automotive applications into laptops and industrial equipment with its on-device inference technology.

This strategic expansion is already delivering results.

Quadric reported licensing revenue between $15 million and $20 million in 2025, a significant increase from approximately $4 million in 2024, according to CEO Veerbhan Kheterpal. The San Francisco-based company, which also maintains an office in Pune, India, is targeting up to $35 million in revenue this year as it builds a royalty-based on-device AI business. This growth has elevated the company's post-money valuation to between $270 million and $300 million, up from around $100 million during its 2022 Series B round, Kheterpal stated.

The company's progress has also drawn investor interest. Quadric recently announced a $30 million Series C funding round led by the ACCELERATE Fund, managed by BEENEXT Capital Management, bringing its total funding to $72 million. Kheterpal noted that this investment comes as both investors and chipmakers seek ways to shift more AI workloads from centralized cloud infrastructure to devices and local servers.

From automotive to everything

Quadric initially focused on automotive applications, where on-device AI enables real-time functions like driver assistance systems. Kheterpal explained that the widespread adoption of transformer-based models in 2023 expanded inference applications across numerous sectors, creating significant business momentum over the past 18 months as more companies seek to run AI locally instead of relying on cloud services.

"Nvidia provides a robust platform for data-center AI," Kheterpal said. "We aimed to create similar programmable infrastructure, comparable to CUDA, specifically for on-device AI applications."

Unlike Nvidia, Quadric doesn't manufacture chips directly. Instead, the company licenses programmable AI processor IP, which Kheterpal describes as a "blueprint" that clients can integrate into their own silicon designs. This is complemented by a software stack and toolchain for running various models, including vision and voice applications, directly on devices.

Disrupt 2026 Tickets: One-time offer

Tickets are now available! Secure savings of up to $680 during this limited-time offer, and be among the first 500 registrants to receive 50% off your additional guest pass. TechCrunch Disrupt gathers industry leaders from Google Cloud, Netflix, Microsoft, Box, a16z, Hugging Face, and many others for 250+ sessions focused on driving growth and enhancing competitive advantage. Engage with hundreds of innovative startups and participate in curated networking opportunities that facilitate business deals, valuable insights, and creative inspiration.

Disrupt 2026 Tickets: One-time offer

Tickets are now available! Secure savings of up to $680 during this limited-time offer, and be among the first 500 registrants to receive 50% off your additional guest pass. TechCrunch Disrupt gathers industry leaders from Google Cloud, Netflix, Microsoft, Box, a16z, Hugging Face, and many others for 250+ sessions focused on driving growth and enhancing competitive advantage. Engage with hundreds of innovative startups and participate in curated networking opportunities that facilitate business deals, valuable insights, and creative inspiration.

The startup's client base includes manufacturers of printers, vehicles, and AI laptops, such as Kyocera and Japanese automotive supplier Denso, which produces chips for Toyota vehicles. According to Kheterpal, the first products incorporating Quadric's technology are scheduled to launch this year, starting with laptop devices.

Beyond traditional commercial applications, Quadric is now exploring markets pursuing "sovereign AI" strategies to decrease dependence on U.S.-based infrastructure, Kheterpal explained. The company is engaging with potential clients in India and Malaysia, with Moglix CEO Rahul Garg serving as a strategic investor helping shape Quadric's sovereign approach in India. Quadric employs nearly 70 people globally, including approximately 40 in the United States and about 10 in India.

This strategic direction is fueled by rising costs associated with centralized AI infrastructure and the challenges many nations face in constructing hyperscale data centers, Kheterpal noted. These factors are generating increased interest in "distributed AI" configurations where inference occurs on laptops or small on-premise servers within office environments, rather than depending on cloud-based services for every computational request.

The World Economic Forum recently highlighted this transition in an article discussing how AI inference is moving closer to end users and away from purely centralized architectures. Similarly, EY's November report indicated that sovereign AI approaches are gaining momentum as policymakers and industry organizations advocate for domestic AI capabilities encompassing computing resources, models, and data, rather than complete reliance on foreign infrastructure.

For chip manufacturers, the primary challenge lies in the rapid evolution of AI models compared to hardware design cycles, Kheterpal observed. He emphasized that clients require programmable processor IP that can adapt through software updates, avoiding expensive redesigns each time architectures transition from earlier vision-focused models to contemporary transformer-based systems.

Quadric positions itself as an alternative to chip suppliers like Qualcomm, which typically integrates its AI technology within its own processors, and IP providers such as Synopsys and Cadence, which offer neural processing engine blocks. Kheterpal noted that Qualcomm's approach can create vendor lock-in with its proprietary silicon, while traditional IP suppliers provide engine blocks that many clients find challenging to program effectively.

Quadric's programmable methodology enables clients to support new AI models through software updates rather than hardware redesigns, providing a crucial advantage in an industry where chip development often requires years, while model architectures can evolve within months.

Despite these advantages, Quadric remains in the early stages of its development, with a limited number of signed clients to date. The company's long-term success will largely depend on converting current licensing agreements into high-volume product shipments and sustainable royalty streams.