Brett Adcock's Hark Raises $700M at $6B for AI Hardware and Foundation Models — Before Shipping a Product
Brett Adcock has raised $700 million for his third company before it has shipped a single product.
Hark, the AI lab Adcock founded in late 2025 with $100 million of his own capital, announced a Series A round led by Parkway Venture Capital at a $6 billion post-money valuation. Investors include NVIDIA, AMD Ventures, Intel Capital, Qualcomm Ventures, Salesforce Ventures, ARK Invest, Align Ventures, Brookfield, Greycroft, Prime Movers Lab, and Tamarack Global.
The company has roughly 70 employees. It has not disclosed a hardware form factor, a target price, a launch market, or a customer pipeline.
The Bet
Hark is positioning itself as the only major lab building foundation models, software, and native hardware together from a blank page — rather than adapting AI into existing form factors. Adcock’s argument is that integrated hardware-software systems, not API-first software layers, will produce the defensible consumer AI product.
The framing is familiar. Every AI hardware startup of the past three years has made a version of it. What’s different at Hark is the investor syndicate’s composition: having NVIDIA and AMD both on the cap table simultaneously addresses the supply allocation problem that has historically strangled AI hardware startups. Access to B200 compute for training is now answered.
On timing, every major lab is moving upstack. Anthropic is prioritising coding tools. OpenAI is heading into its IPO. Few are focused on building consumer-facing interfaces with native hardware from scratch. Adcock is betting the window is open.
Adcock’s Track Record
Adcock founded Archer Aviation, which has produced eVTOL aircraft that have reached commercial service. He founded Figure AI, the humanoid robotics company that has shipped the F.03 robot and reached production deployment. Both required integrating custom hardware with novel software systems at high capital intensity before any product revenue.
Hark follows the same pattern at higher stakes and a shorter hardware timeline.
What Ships Next
The company plans to release its first multimodal models this summer, made available through a personal AI platform. The models will be agentic and multimodal, with persistent memory across sessions and integrations with existing products and services.
Hardware comes after the models. The device layer is designed specifically for Hark’s foundation models — not a general-purpose compute platform retrofitted for AI, but silicon and form factor chosen to match the software from the start.
The company is training on NVIDIA B200s at a new company-operated data centre. The round funds infrastructure, hiring, and model development through to first product launch.
The Risk
$700 million at $6 billion for a company with no product is a bet on team and approach, not traction. The consumer AI hardware category has a poor success rate: Humane’s AI Pin, Rabbit R1, and several others burned significant capital and retreated or pivoted. Hark’s counterargument is that those products were software wrappers in hardware form, not systems where models and silicon were built together.
Whether that distinction translates to something users buy remains the open question. By the company’s own timeline, the answer starts arriving this summer.