Hark raises $700M+ to build ‘personalized intelligence’ devices
Hark Inc. raised over $700 million in Series A funding led by Parkway Venture Capital, with backing from Nvidia, Intel, AMD, Qualcomm, and Salesforce, valuing the company at $6 billion. The startup plans to develop AI devices with personalized, multimodal models capable of tasks like restaurant reservations and voice commands, targeting a summer launch for its models and a later release for consumer hardware.
Hark Inc., a startup focused on AI-driven consumer devices, announced a $700 million Series A funding round led by Parkway Venture Capital. Investors including Nvidia Corp., Intel Capital, AMD Ventures, Qualcomm Ventures, and Salesforce Ventures joined, boosting the company’s valuation to $6 billion. Hark previously secured $100 million from CEO Brett Adcock, who also leads Figure AI and founded Archer Aviation. The company is developing custom AI models with persistent memory to personalize interactions, such as reserving restaurant tables or processing voice commands. These models will run on AI-optimized hardware, positioning Hark as an alternative to cloud-based AI services. A demo shows capabilities like e-commerce purchases and research, though business automation support remains unclear. Hark plans to launch its first AI models this summer, with consumer devices following later. The startup will use the funding to hire hardware engineers, AI researchers, and designers while upgrading its training infrastructure, which relies on on-premises Blackwell B200 GPUs. Competition looms from OpenAI, which is reportedly developing a screenless AI device and other wearables for a 2027 debut. The company’s approach contrasts with cloud-based multimodal models, potentially requiring optimized, less resource-intensive algorithms for local processing. Google’s Gemma 3n model, designed for limited hardware, may serve as a reference for efficiency. Hark’s vision aligns with Adcock’s statement about creating ‘personalized intelligence’ tailored to individual users and hardware.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.