Hark raises $700M to build personal AI hardware

Per Business Insider, Yahoo Finance, and Dealroom, San Jose-based AI lab Hark raised $700 million in a Series A at a $6 billion post-money valuation. Parkway Venture Capital led the round, with participation from Nvidia, AMD Ventures, Salesforce Ventures, Intel Capital, Qualcomm Ventures, ARK Invest, Brookfield, Greycroft, and others, according to Yahoo Finance and Dealroom. Dealroom and Yahoo Finance report Hark was founded by Brett Adcock, who seeded the company in late 2025 with $100 million of his own capital.

Reporting by Yahoo Finance states Hark is developing personalized AI systems and hardware intended to interact naturally with people and the physical world, not only via chat. Yahoo Finance and Dealroom report the company expects to release its first multi-modal models this summer and to provide early access to a personal AI platform, with hardware devices to follow. Per Yahoo Finance, Hark operates a data center using Nvidia B200 GPUs and has roughly 70 employees. Dealroom and Yahoo Finance note hires from Apple, Google, Meta, and Amazon, and identify Abidur Chowdhury, a former Apple product executive, as Hark's director of design.

Industry context: Large, early-stage rounds above half a billion dollars for startups focused on integrated AI software plus custom hardware are uncommon and reflect strong investor appetite for vertically integrated consumer-AI plays. Participation from chipmakers and their corporate VC arms (for example Nvidia, AMD Ventures, Intel Capital, Qualcomm Ventures) typically signals investor interest in securing downstream demand for silicon and hardware partnerships, as reported by the outlets covering the round. For practitioners, the combination of multi-modal model development and a dedicated hardware roadmap raises questions about compute procurement, model compression, on-device inference, and end-to-end systems engineering that teams building similar products routinely confront.

Companies attempting to ship tightly integrated AI devices and personal assistants commonly need to solve several engineering challenges concurrently: large-scale fine-tuning for personalization, low-latency multi-modal inference, systems integration between cloud and edge, and hardware-design constraints for power and thermals. Industry observers note that successful productization often depends on early alignment between chip vendors, component suppliers, and software stacks; the investor mix reported by Yahoo Finance and Dealroom is consistent with that pattern.

Observers should track the timing and technical shape of the model releases reported by Yahoo Finance and Dealroom, any published model specs or benchmarks, announced hardware partners or suppliers, and hires in systems-on-chip, firmware, and embedded software. Also watch for developer APIs, privacy and personalization controls in any early-access platform, and how on-device versus cloud inference responsibilities are described by the company or participating partners. Per Business Insider, Adcock's quoted press release emphasizes personalization; watching how that goal maps to data collection, privacy posture, and on-device model design will be important for engineers and product teams.

Business Insider reproduces a press-release quote from Brett Adcock: "We're building the AI that everyone deserves, but no one has built yet - one that actually knows you, speaks your language, is highly personalized, and lives on hardware made for you."