Hark, one of the most secretive startups in AI, has raised $700 million in a Series A to build what it describes as a universal AI interface. The company says it expects to release its first multimodal models this summer, and will follow with hardware devices built specifically for that platform.
A $700 Million Statement
A $700M Series A is an extraordinary raise for any startup, let alone one that has shown few public details. It signals investor confidence that a new interface layer can sit between people and the apps and services they already use — without forcing them onto a proprietary ecosystem.
The Universal Interface Thesis
The pitch, as TechCrunch reports it, is simple: instead of another assistant locked to its own stack, Hark’s platform should work with the products and services people already rely on.
- Works everywhere — connects to tools users already have
- Multimodal first — voice, text, images, and more in one system
- Hardware later — dedicated devices built specifically for the platform
Models First, Hardware Second
The roadmap has two stages. First come the multimodal models, scheduled for this summer, which will anchor the personal AI platform. After that, Hark plans dedicated hardware — a bet that the interface deserves its own device rather than living only inside an app.
The Risks of Stealth
Stealth has costs. With no public model, benchmarks, or demos, Hark carries an unusually heavy load of expectations. The company is also entering a crowded field of device-focused AI startups, where differentiation will depend on whether the universal layer genuinely works across every app, and whether the hardware delivers something phones cannot already do.
What This Means
$700 million is a bet that the battle for AI is moving from models to interfaces. If Hark’s universal layer works, it could become the surface everyone else builds on. If it stumbles, the same capital will be remembered as the biggest Series A gamble in AI.