$700M for "Universal AI": The Hark Story Is Weirder Than It Looks

Brett Adcock doesn't do small. He built Figure AI to staff warehouses with humanoid robots. Before that, he built Archer Aviation to put you in electric aircraft. Now he's started Hark, which he has raised $700 million to build.

What does Hark make?

A "universal AI interface."

If you were hoping for more specificity, welcome to the club. But also: you're going to have to sit with some ambiguity, because the people who funded this also couldn't pin it down more precisely, and they handed over seven hundred million dollars anyway.

The round closed at a $6 billion post-money valuation. That makes Hark one of the most valuable AI companies you've never heard of. The investors include Nvidia, AMD Ventures, Qualcomm, and Salesforce Ventures. None of these companies invest in things they can't imagine building on top of, or selling hardware into.

Adcock founded Hark in late 2025 and seeded it with $100 million of his own money before asking anyone else. The company has 70 employees and runs a data center full of Nvidia B200 GPUs. Multimodal models are shipping in summer 2026. Hardware comes after.

The design director is Abidur Chowdhury, a former Apple executive. His explanation of why he joined is the most interesting thing anyone at Hark has said publicly: "I haven't seen anything that feels like something that will really help like the normal person."

Read that again. That's not a pitch. That's a critique of every existing AI product, buried in a recruiting announcement. It is, in its way, the best argument for why Hark might matter.

So What Is a "Universal AI Interface," Exactly?

The vision is something like this: instead of switching between apps, asking ChatGPT one question and Google another, manually entering every appointment and bookmark and reminder, your AI just knows. It runs in the background. It understands your life in aggregate. You express what you want; it figures out how to get there using whatever tools exist.

You've heard this pitch before. Google tried it again at I/O this month with its AI agent ecosystem, to decidedly mixed consumer reception. Apple tried it with Apple Intelligence. Siri was supposed to be this in 2011. Alexa, Cortana, and a small graveyard of now-defunct startups all took the same run at it. So what makes Hark different?

Two things, maybe.

First: hardware. Adcock isn't building another app. He's building a purpose-built device designed from the ground up for one job. Purpose-built hardware tends to outperform multitasking devices when the job is specific enough. Whether a personal AI assistant qualifies is an empirical question we don't have an answer to yet.

Second: timing. The AI industry just spent three years pretending chatbots were something smaller than they are, and the reckoning is landing. The category is genuinely in transition. Building the right product at the right inflection point is how a lot of enduring companies started.

The Part Nobody Wants to Talk About

Here's what every one of these announcements breezes past: privacy.

A universal AI interface is a universal data collection interface. If your AI knows your calendar, your email, your location, your conversations, and what you search for at 11 PM, then the company running that AI knows all of those things too. This is not a hypothetical concern. Hark's team has acknowledged that privacy is "largely unaddressed."

Think about that for a second. Seven hundred million dollars raised, pre-product, and privacy is described as largely unaddressed. Credit for honesty. But "largely unaddressed" is a strange thing to hear from a company that wants to know more about your life than your closest friends do.

Privacy is not a feature you bolt on after product-market fit. The architectural decisions you make at 70 employees calcify by 700 and become load-bearing walls at 7,000. Every AI company that has gotten this wrong discovered the problem only when it was expensive to fix. Hark has at least acknowledged it exists. That's a start. It's a long way from a solution.

Should You Care About Hark?

Probably, yes.

Adcock's track record is genuinely unusual. Figure AI has working humanoid robots doing warehouse tasks. Archer Aviation has working aircraft. This is not a first-time founder pitching a concept deck to anyone with a checkbook. He has a demonstrated ability to raise enormous capital for technically hard bets and then actually ship something.

The chipmaker investor roster matters too. When Nvidia and AMD both write checks into the same private company, the hardware thesis is usually real. They're not doing it for the founder's charisma.

Hark might not become what Adcock thinks it will. Most companies that raise $700M Series A rounds before shipping don't. The gap between "multimodal models in summer 2026" and "universal AI that normal people use every day" is large and full of unsolved problems, including the one above.

But the gap between what AI products feel like right now, and what they would need to feel like to actually be useful to a non-technical person on a Tuesday, is also real. Someone is going to close it. Adcock is betting it's him. He talked a lot of serious money into agreeing.

Seven hundred million to find out. That's Silicon Valley for you.

---

The About.chat newsletter covers what's actually happening in AI without the press release spin. Subscribe free at aboutchat.beehiiv.com and we'll send it twice a week.