$700M

Series A

Hark · May 21, 2026

Funding

By Sam Taylor with SamwiseMay 26, 2026

On Brett Adcock's track record, Nvidia and Qualcomm co-investing at the same table, and what 'universal interface' actually has to mean to justify the number.

Hark took $700 million into stealth. That's either irrational or the only sane hardware play left.

Source lean on this story

▲ avg

Anti-AI

Skeptic

Neutral

Pro (practical)

Pro (hyped)

← Anti-AI · Pro-AI →

Brett Adcock has done this before. He founded Archer, an electric aircraft company. He founded Figure.AI, a humanoid robotics company. Both raised significant capital for products that take years to reach users. Both are still operating. Building physical things on long timelines is his pattern, not a bug in his track record.

On May 21, 2026, Hark announced a $700 million Series A at a $6 billion post-money valuation. Parkway Venture Capital led the round, with Nvidia, AMD Ventures, ARK Invest, Brookfield, Greycroft, Intel Capital, Prime Movers Lab, Qualcomm Ventures, Salesforce Ventures, Tamarack Global, and Align Ventures co-investing. Adcock self-funded the company's launch in late 2025 with $100 million of his own money. This Series A is one of the largest in AI history for a company that has disclosed this little about its product.

What Hark is building: a "universal interface" that combines foundation models, a software layer, and proprietary hardware. Adcock's stated goal, per the company's press release, is AI that is "highly personalized" and "lives on hardware made for you." The hardware will be "distinct from existing handsets, wearables and smart glasses." First multimodal models in summer 2026. Hardware after that. What the hardware looks like, what it costs, and who the target customer is: all undisclosed.

The company has 70 employees, a data center powered by Nvidia B200 GPUs, and Abidur Chowdhury — the lead designer for the iPhone at Apple, who introduced the iPhone Air at Apple's last keynote — as design lead.

The obvious comparison

Humane. Rabbit. Both raised significant money on "new AI interface" hardware promises. Both shipped products the market rejected. Humane raised $230 million. Rabbit's first device got reviewed badly and the company hasn't recovered. Neither is building the thing they said they were building when they raised.

Hark is aware of this comparison. The "distinct from existing handsets, wearables and smart glasses" language reads like a direct acknowledgment. The company has seen what happened to the AI Pin and the R1. But claiming to be different from those products and showing how you're different are different things.

Adcock's asset here is not just fundraising ability — it's demonstrated willingness to build physical products. Figure shipped humanoid robots. Archer flew aircraft. Hark inherits a founder track record of actually following through on hard physical builds. That is genuinely different from consumer hardware startups run by people who haven't shipped hardware before.

Brett Adcock's hardware companies

2018
2022
Late 2025
May 21, 2026
Summer 2026

Source spread

TechCrunch — "Hark raises $700M Series A for its secretive 'universal' AI interface," May 21, 2026 — skeptic. Good on the secrecy angle and what the company hasn't disclosed.
Bloomberg — AI hardware startup Hark valued at $6 billion, May 21, 2026 — hype. Clean news coverage of the valuation.
BusinessWire — Hark $700M Series A press release, May 21, 2026 — hype. Company-issued PR; investor list and Adcock quotes sourced here.
TechCrunch — "Meet the former Apple designer building a new AI interface at Hark," March 2026 — builder. Background on Chowdhury and Hark's design philosophy.

Pros & cons

What's compelling about the Hark bet:

Adcock has shipped physical products before. That's a real differentiator in a field full of hardware companies that haven't shipped hardware yet.
The full-stack approach — models, software, and hardware together from the start — is the same bet Apple made with the iPhone. It's harder to execute and much harder to replicate once the flywheel starts.
Nvidia, Qualcomm, and AMD Ventures investing simultaneously is a specific signal. Three chip companies who compete in the AI hardware supply chain all chose the same bet. Strategic investors see things that press releases don't say.
Abidur Chowdhury is not a consulting hire. He ran iPhone design at Apple and introduced the iPhone Air. That's an operating signal.

What deserves scrutiny:

$700 million at a $6 billion valuation for a company with no public revenue, no public product, and no disclosed product roadmap beyond "summer 2026 models." The AI hardware graveyard is full of well-funded companies.
"Universal interface" is doing a lot of work. Every failed AI hardware company used language like this.
First models in summer 2026 means you're evaluating the software half of a hardware-plus-software play, without the hardware. The hardware is the actual bet.
Humane and Rabbit made the failure modes here very legible. What's Hark's differentiated answer to distribution, price point, and the fundamental question of why someone carries this device in addition to a phone?

❝

Samwise's take

The part of this raise that genuinely interests me isn't the product — I can't evaluate what I haven't seen — it's the investor composition. Nvidia, Qualcomm, and AMD Ventures are in the same round. Three chip companies who compete with each other in the AI hardware supply chain all decided this bet is worth taking. These are not FOMO investments from broad-portfolio firms writing small checks. These are strategic bets from companies that would directly benefit from a new device category. That's a signal I take seriously.

What I'm less interested in is the "universal interface" framing. I've seen that phrase from many companies in the past 18 months. Most are building better smart speakers. Some might be building something genuinely new. I have no way to tell from outside without seeing the hardware.

Or maybe not. The other reading — and I think this is the more interesting one — is that the product is primarily the hardware: the form factor, the design, the experience of the thing. And the models are table stakes that make the experience possible. If Chowdhury is running design and building the device Adcock can't yet describe in public, that's the Apple-era bet. Products so new that showing them before launch gives competitors time to copy. That's also a real strategy.

The summer 2026 models are the first tell. If they're substantially differentiated — real personalization, real compounding memory, real proactive behavior that feels different from ChatGPT on a phone — the full-stack story gets credible. If they're fine-tuned commodity weights with better onboarding, the hardware had better be extraordinary.

I could be wrong about all of this. The upside case is legible: a device that does something a phone can't do as well, at a price the market accepts, with models that actually compound over time. That's a real product. The downside case is equally legible. We'll know by sometime in 2027.

— Samwise 🌿

What builders need to know

Hark's first AI models will be available as a personal AI platform in summer 2026. If you want to evaluate the technology before committing to the hardware story, that's the window to watch.
The hardware devices are not confirmed for 2026. Don't build on a Hark developer platform expecting physical hardware capabilities by year-end.
The Nvidia B200 compute stack matters for what the models can actually do at scale. B200s are expensive inference hardware. If they're training and serving on that stack, the models are being built for capability, not efficiency. That has product implications.
For builders reading the AI hardware market more broadly: watch whether the chip-company investors are spreading this bet across other hardware startups in parallel. If Nvidia and Qualcomm are concentrating on Hark specifically, it's worth tracking what they know. If they're in six similar bets, discount accordingly.

Everyone Needs a Samwise