Businesses and government entities are increasingly seeking solutions to operate artificial intelligence directly on local devices, aiming to reduce cloud infrastructure expenses and develop independent technological capabilities. Quadric, a semiconductor intellectual property startup established by former members of the early bitcoin mining company 21E6, is positioning itself to enable this transition. The firm is expanding its reach from the automotive sector into laptops and industrial equipment with its on-device inference technology, and this strategic growth is already yielding positive results.

Headquartered in San Francisco with an additional office in Pune, India, Quadric is working toward generating up to $35 million in revenue this year as it cultivates a royalty-based business model for on-device AI. This progress has significantly increased the company’s valuation, which now stands between $270 million and $300 million post-money, a notable rise from approximately $100 million during its 2022 Series B funding round, according to Kheterpal. The company’s trajectory has also drawn increased investor interest. Last week, Quadric closed a $30 million Series C investment round led by the ACCELERATE Fund, managed by BEENEXT Capital Management, bringing its total raised capital to $72 million.

**Expanding from Automotive to Broad Applications**

Quadric initially focused on the automotive industry, where on-device AI can enable real-time features such as driver assistance systems. Kheterpal noted that the widespread adoption of transformer-based models in 2023 significantly broadened the application of AI inference, creating a pronounced business inflection over the last year and a half as more organizations opt to run AI locally instead of depending on cloud services. “Nvidia provides a robust platform for data-center AI,” Kheterpal stated. “Our goal was to develop a similar, programmable infrastructure akin to CUDA, but specifically for on-device AI.”

In contrast to Nvidia, Quadric does not manufacture its own chips. Instead, it licenses programmable AI processor intellectual property, which Kheterpal describes as a “blueprint” that clients can integrate into their own semiconductor designs. This is accompanied by a comprehensive software stack and toolchain to run various models, including those for vision and voice, directly on the device.

The startup’s client base includes companies in the printer, automotive, and AI laptop sectors, such as Kyocera and the Japanese automotive supplier Denso, which produces chips for Toyota vehicles. However, Quadric is now looking beyond conventional commercial applications toward markets pursuing “sovereign AI” strategies to lessen dependence on U.S.-based infrastructure, Kheterpal explained. The company is engaging with potential customers in India and Malaysia and has enlisted Moglix CEO Rahul Garg as a strategic investor to help guide its sovereign AI approach in India. Quadric currently employs nearly 70 people globally, with about 40 based in the United States and roughly 10 in India.

This strategic push is motivated by the escalating costs of centralized AI infrastructure and the challenges many nations face in constructing hyperscale data centers, Kheterpal said. These factors are spurring greater interest in “distributed AI” configurations, where inference tasks are handled on laptops or compact on-premise servers within offices, rather than relying on cloud services for every request. The World Economic Forum recently highlighted this trend toward moving AI inference closer to end-users and away from purely centralized systems. Likewise, EY noted in a November report that the sovereign AI approach has gained momentum as policymakers and industry advocates push for domestic capabilities across computing power, models, and data, reducing complete reliance on foreign infrastructure.

For semiconductor companies, a key challenge is that AI models are advancing more rapidly than hardware development cycles, Kheterpal pointed out. He emphasized that customers require programmable processor IP that can adapt via software updates, avoiding expensive redesigns each time architectures evolve—from earlier vision-centric models to contemporary transformer-based systems. Quadric presents itself as an alternative to chip manufacturers like Qualcomm, which typically incorporates its AI technology into its own processors, and to IP providers such as Synopsys and Cadence, which offer neural processing engine blocks. Kheterpal argued that Qualcomm’s model can lock customers into its proprietary silicon, while traditional IP suppliers offer engine blocks that many find difficult to program. Quadric’s programmable approach allows clients to support new AI models through software updates rather than hardware redesigns, providing a significant advantage in an industry where chip development can take years, while model architectures may shift in just months.

Nevertheless, Quadric is still in the early stages of its commercial rollout, with a limited number of signed customers to date. Its long-term success will largely depend on converting current licensing agreements into high-volume production and sustained royalty streams.