Qualcomm bets $70M on smart rings as wearables pivot to AI computers

Qualcomm has led a $70 million Series C round in Ultrahuman, the Bengaluru-based wearables startup behind the Ultrahuman Ring, signaling a bold bet on turning smart rings into programmable AI computers rather than simple biometric trackers. The round, which closed in June 2024, values Ultrahuman at $400 million post-money and includes participation from existing backers such as Blume Ventures and Kalaari Capital. The company, founded in 2019 by former Apple engineers Arnav Kapur and Ayush Banka, now aims to hit a $200 million annual revenue run rate by January 2027, driven by a new generation of Qualcomm Snapdragon-powered rings capable of running lightweight AI models entirely on-device.

Ultrahuman’s latest device integrates Qualcomm’s Wear 4100+ platform—a low-power SoC typically used in Android Wear devices—with custom AI accelerators to process health, activity, and contextual data without cloud dependency. This enables real-time glucose trend prediction, stress inference from heart rate variability, and even on-ring gesture recognition for user input. Kapur, Ultrahuman’s CEO, confirmed in an interview that the device now runs a proprietary “neural inference engine” that compiles TensorFlow Lite models directly to the ring’s onboard processors, a breakthrough from earlier generations that relied on smartphone tethering. The shift aligns with Qualcomm’s push into edge AI, where the company sees wearables as a critical gateway to machine learning workloads outside traditional data centers.

The partnership extends beyond hardware. Ultrahuman’s engineering team has contributed custom kernel patches to the Linux-based Qualcomm platform to support ring-specific thermal management and power throttling, enabling sustained operation during high-load AI inference. According to a source close to the deal, Qualcomm’s strategic investment was motivated not only by market potential but by the ring’s ability to demonstrate Qualcomm’s AI silicon in a high-volume, always-on consumer device—potentially influencing future Snapdragon Wear roadmaps. The funding will be used to scale production, expand the Ultrahuman AI platform to third-party developers, and support a nascent “RingOS” developer kit, slated for public release in Q4 2024.

Industry Impact and Significance

This development signals a tectonic shift in the Tools & Developer ecosystem, moving wearables from data collectors to active compute nodes. For Qualcomm, it validates a new use case for its low-power AI chips at the extreme edge, a market currently dominated by Apple’s W-series and Google’s Tensor G wearables. Competitors like Fitbit and Whoop have focused on health tracking, while Ultrahuman’s approach—embedding programmable inference into jewelry—creates a new category: the AI-augmented personal computer. Financial analysts at Counterpoint Research project that by 2026, over 20% of wearable devices could include edge AI capable of running LLMs or custom models, up from less than 5% in 2023. This raises the stakes for tooling vendors like Arm, Synopsys, and Cadence, who must now optimize compilers and simulation environments for wearable-class silicon.

Developers stand to benefit from Ultrahuman’s planned RingOS SDK, which will expose APIs for sensor fusion, on-device model deployment, and voice-to-text in a low-latency environment. Already, Banking With Billy AI—a fintech startup using AI-driven financial modeling—has prototyped a smart ring version of its risk analytics dashboard, leveraging the ring’s continuous glucose and HRV data to adjust lending models in real time. The company reports a 40% reduction in latency during inference compared to smartphone-based pipelines, demonstrating how edge-first wearables can unlock latency-sensitive applications. This trend pressures cloud providers like AWS and Google Cloud to rethink their wearable data strategies, potentially pushing them toward hybrid architectures where models are trained in the cloud but executed at the edge.

The Bigger Picture

Ultrahuman’s trajectory reflects a broader convergence of AI, miniaturization, and personal computing—one that harks back to the early days of mobile development but now operates at millimeter scale. It echoes Apple’s 2017 pivot with the iPhone’s Neural Engine, but in a form factor closer to the 1980s calculator watch. The move also underscores how AI is no longer confined to data centers or smartphones; it’s now embedded in objects we wear daily. This mirrors Qualcomm’s long-term vision outlined in its 2023 AI Day, where the company predicted that by 2030, 50% of AI inference would occur at the edge, not in the cloud. Ultrahuman’s ring serves as an early proof point for that thesis.

Yet the model raises questions about sustainability and user privacy. Running continuous AI workloads on a ring-sized battery demands aggressive power capping, which may limit model complexity. Ultrahuman claims it can sustain six hours of continuous inference per charge, but that pales against all-day health tracking use cases. Competitors like Oura and Circular are watching closely, exploring similar paths but with different trade-offs. Meanwhile, regulators in the EU and US are beginning to scrutinize AI-driven biometric wearables, particularly around consent and data inference. Ultrahuman’s approach—processing data locally—could position it favorably, but only if it can prove that local processing doesn’t mask deeper privacy concerns.

Expert Analysis

Vikram Sharma, a senior analyst at SemiAnalysis, calls the Qualcomm-Ultrahuman deal “a bellwether for edge AI’s democratization.” He notes that by funding a wearable OS and toolchain, Qualcomm is effectively subsidizing a new developer ecosystem that could rival Android Wear or Wear OS in scope. “This is not just a hardware bet,” Sharma says. “It’s a software moat disguised as a gadget. If RingOS gains traction, it could become the de facto standard for ultra-low-power AI devices, forcing Google and Apple to reconsider their closed wearables stacks.” He advises developers to begin experimenting with wearable-specific compilers and power profilers now, as the first wave of AI-native rings will hit mass markets by late 2025. For the Tools & Developer community, the message is clear: the next frontier isn’t in the cloud—it’s on your finger.