The "proactive" AI trend has blown into Silicon Valley, and Hark has secured $700 million in funding

None

Text | Alpha Society

AAI startup founded by the end of 2025 that has not yet publicly released products has secured $700 million in Series A funding, with its valuation soaring to $6 billion. The round was led by Parkway Venture Capital, with participating investors including NVIDIA, AMD Ventures, Intel Capital, Qualcomm Ventures, and Salesforce Ventures.

As we can see, this company has received massive funding in a short period, and industry-leading hardware and software tech giants have endorsed it.

This company, called Hark, has a clear entry point: they aim to use a combination of “self-developed foundational models + customized hardware” to create the next generation of general human-machine interfaces.

Essentially, this is a new type of AI interface, manifested as AI-native hardware, which consists of a series of customized native hardware devices and computing devices with intelligent agent capabilities, equipped with end-to-end speech models and highly personalized memory abilities. All these AI systems are multimodal, capable of understanding and interacting naturally.

When we see Hark completing a $6 billion valuation financing round with NVIDIA and Qualcomm entering simultaneously, we are not surprised. Since 2024, Alpha Society has been actively布局 in the “proactive AI” direction — our early investment, Looki, has already sold multimodal AI wearable devices to users worldwide, becoming the world's largest multimodal general-purpose wearable AI device; Guangfan Technology has developed an AI operating system native to hardware and pioneered AI earphones with visual perception capabilities.

Hark’s massive funding further confirms an increasingly obvious trend: the next decade of AI will be not only on screens but also in the real world.

AI has become smarter, but it still uses old shells and interaction methods

Hark was founded by Brett Adcock at the end of 2025, initially with a personal investment of $100 million. Brett Adcock has previously founded companies such as Archer, Figure, and Vettery.

Among them, Archer entered the electric vertical takeoff and landing (eVTOL) aircraft market and successfully went public. Figure is a humanoid robot company; in 2024, Figure raised $675 million, and in September 2025, it completed a Series C round exceeding $1 billion, with a valuation reaching $39 billion. Its investors include Jeff Bezos, NVIDIA, Microsoft, OpenAI, and others.

Why would Brett Adcock want to start a business in the “proactive” AI native hardware direction? Because Figure’s core system is fundamentally “AI + hardware + real-world interaction,” which is very similar to the underlying tech stack of AI-native hardware, so he knows where the pitfalls are. Recently, Figure demonstrated a robot performing long-term parcel sorting tasks, indicating they have solved some problems.

Image source: Brett Adcock’s personal website

Besides Brett Adcock himself, Abidur Chowdhury has joined Hark as the head of design. He was a senior Apple product designer, involved in designing products like iPhone Air. Hark has also attracted engineers from Apple, Meta, Google, Tesla, and top AI labs, covering AI research, hardware engineering, and design.

Looking at the history of personal hardware terminal development, it’s essentially a history of alternating advances in hardware form, interaction methods, and applications: as hardware forms and interaction methods evolve, new applications are generated, unlocking new capabilities, and spreading to larger user groups.

For example, once the PC form factor was established and its size became small enough, combined with mature interaction interfaces like mouse and GUI, it became easier for ordinary people to use. After the internet became widespread, it moved from business and creative professionals to the general public.

The next breakthrough came with the iPhone, which not only integrated the capabilities of computers and phones into a small form factor but also introduced multi-touch interaction, further lowering the interaction threshold, making the user base of smartphones (including tablets) surpass that of PCs.

Moreover, its App Store ecosystem directly set the software standard for the mobile internet era, generating approximately $1.3 trillion in developer revenue and sales globally in 2024.

Currently, the problem with AI is that it is intelligent and software-capable, but mainly runs and interacts through chat interfaces and non-AI-native devices like computers and smartphones, lacking continuous memory of user identities and hardware designed specifically for intelligent interaction.

An industry consensus is that the next stage requires intelligent agent systems capable of natural interaction with humans and the real world. These systems should be able to anticipate needs, reduce cognitive load, and operate as collaborative partners rather than waiting for commands like traditional software.

At present, AI at the software level has already spawned super startups like OpenAI and Anthropic, with valuations approaching one trillion dollars. As AI-native hardware further develops, its impact on the tech industry could be comparable to the iPhone.

However, for “proactive” AI native hardware to mature, it is a complex engineering challenge. For example, Hark needs to build this entire system from the model, AI hardware, interaction, and memory layers.

First, their models will have agent capabilities, multimodal abilities, and memory functions, able to remember who the user is, what was said, and work across products and services already used by the user.

They will design AI-native hardware and integrate it with Hark’s foundational models. Judging by their recruitment for real-time speech infrastructure roles, their interaction interface will likely start with voice.

Developing “proactive” AI, Chinese startups have more advantages

Current AI, whether chatbots or agents, are still just tools because they are confined to screens; people only give commands when they need them and then receive results.

Compared to these “passive” AIs, why is “proactive” AI important? Because it transforms AI from a tool into a collaborator. AI can independently think, do, and complete tasks to some extent, helping humans.

Building a “proactive” AI system requires a combination of software and hardware AI-native hardware, capable of perception, memory, intelligence, and new, lower-threshold interaction, and must be always on and close to users.

In the previous AI hardware exploration phase (e.g., smart speakers), perception existed but was limited to storage; intelligence was insufficient, and interactions were rigid.

In this new AI explosion phase, perception capabilities have further improved, and AI’s memory and intelligence have made huge progress. Interaction is still being explored, but the path is now clearer.

True “proactive” AI has already taken a big step forward.

To make “proactive” AI more mature, it’s not a matter of single breakthroughs but a comprehensive competition involving foundational models, agent operating systems, personalized memory, and hardware terminals. The competition for AI-native hardware is a holistic one.

In this field, innovation and exploration, Chinese startups have more opportunities for success, thanks to three unique advantages: first, manufacturing ecosystem strength, with places like Shenzhen having the most complete global supply chain; second, market scale advantage, as China is both the largest manufacturing country and the largest application market; third, policy support, as the government has prioritized AI as a strategic focus, providing certainty for long-term investments.