Quadric Raises $30M to Drive On-Device AI Revolution

The drive to reduce cloud infrastructure costs and build sovereign AI capabilities is fueling a significant shift towards on-device AI inference. At the forefront of this movement is Quadric, a chip-IP startup co-founded by veterans of early Bitcoin mining firm 21E6. Quadric's innovative programmable AI processor IP, designed to run rapidly evolving AI models locally, is gaining considerable traction, evidenced by its recent $30 million Series C funding round and impressive revenue growth.

This expansion is already yielding significant returns. CEO Veerbhan Kheterpal told TechCrunch in an interview that Quadric is projected to reach $15 million to $20 million in licensing revenue for 2025, a substantial increase from approximately $4 million in 2024. The San Francisco-based company, with an additional office in Pune, India, is ambitiously targeting up to $35 million for the current year as it cultivates a royalty-driven on-device AI business. This impressive growth has boosted Quadric's post-money valuation to between $270 million and $300 million, a significant leap from its roughly $100 million valuation during its 2022 Series B round.

The company's strong performance has also attracted significant investor interest. Quadric recently announced a $30 million Series C funding round, spearheaded by ACCELERATE Fund, managed by BEENEXT Capital Management. This latest infusion brings Quadric's total funding to $72 million. Kheterpal noted to TechCrunch that the funding reflects a broader industry trend where investors and chipmakers are increasingly seeking solutions to shift AI workloads from centralized cloud infrastructure to local devices and servers.

From Automotive to Ubiquitous On-Device AI

Quadric initially focused on the automotive sector, where on-device AI is crucial for real-time functions such as driver assistance systems. However, Kheterpal explained that the proliferation of transformer-based models in 2023 catalyzed a significant shift, pushing AI inference into virtually "everything." This trend has created a sharp business inflection point over the last 18 months, as more enterprises prioritize running AI locally over cloud reliance.

Kheterpal drew a parallel, stating, "Nvidia is a strong platform for data-center AI. We were looking to build a similar CUDA-like or programmable infrastructure for on-device AI." Crucially, unlike Nvidia, Quadric does not manufacture chips. Instead, it licenses its programmable AI processor IP – which Kheterpal likens to a "blueprint" – allowing customers to embed it directly into their own silicon. This IP comes complete with a comprehensive software stack and toolchain, enabling on-device execution of various AI models, including vision and voice applications.

Quadric's customer base is diverse, encompassing sectors like printers, automotive, and AI-enabled laptops. Notable clients include Kyocera and Japan's automotive supplier Denso, known for developing chips for Toyota vehicles. Kheterpal anticipates that the first products integrating Quadric's technology, starting with laptops, will ship this year.

Beyond conventional commercial applications, Quadric is strategically expanding into markets pursuing "sovereign AI" initiatives. These strategies aim to lessen dependence on U.S.-based AI infrastructure, with Quadric actively exploring opportunities in India and Malaysia. Moglix CEO Rahul Garg serves as a strategic investor, advising on the company's sovereign AI approach in India. Globally, Quadric employs nearly 70 individuals, with approximately 40 in the U.S. and 10 in India.

The growing interest in distributed AI is largely driven by the escalating costs of centralized AI infrastructure and the significant challenges many nations face in establishing hyperscale data centers. This has spurred a demand for setups where AI inference can run on local devices like laptops or small on-premise servers, reducing reliance on cloud services for every query. This paradigm shift has been highlighted by prominent organizations; the World Economic Forum recently underscored the movement of AI inference closer to users, away from purely centralized architectures. Similarly, an EY report from November noted the increasing traction of the sovereign AI approach, as policymakers advocate for domestic capabilities across compute, models, and data, rather than full dependence on foreign infrastructure.

The Advantage of Programmable AI Processor IP

A critical challenge for chipmakers is the rapid evolution of AI models, which often outpaces traditional hardware design cycles. Kheterpal emphasizes that customers require programmable processor IP capable of adapting through software updates, thereby avoiding costly hardware redesigns with every architectural shift – from older vision-focused models to contemporary transformer-based systems.

Quadric positions itself as a compelling alternative to established players. Unlike chip vendors such as Qualcomm, which integrate AI technology within their proprietary processors, or IP suppliers like Synopsys and Cadence, who offer neural processing engine blocks, Quadric provides a more flexible solution. Kheterpal points out that Qualcomm's strategy can lock customers into its specific silicon, while many find traditional IP suppliers' engine blocks challenging to program.

Quadric's programmable approach empowers customers to support new AI models via software updates, eliminating the need for extensive hardware redesigns. This offers a significant competitive edge in an industry where chip development can take years, yet AI model architectures can transform in mere months.

Despite its rapid growth and strategic positioning, Quadric acknowledges it is still in the early stages of its buildout. The company's long-term success hinges on converting its current licensing agreements into high-volume shipments and a consistent stream of recurring royalties.