The $700M Bet on a Product That Doesn't Exist
One AI lab. Zero shipping product. Twelve of the world's most informed investors. Here's the math.
Brett Adcock raised $700 million in Series A last Thursday. The kicker: Hark, the AI lab he founded, hasn’t shipped anything. No public product. No demo. No models in market. Just a deck, 70 engineers, a fresh Nvidia B200 cluster, and a $6 billion post-money valuation.
Most Series A rounds clock in between $20M and $40M. Hark’s is roughly 20x that. Parkway Venture Capital led. Nvidia, AMD Ventures, Intel Capital, Qualcomm Ventures, Salesforce Ventures, ARK Invest, Brookfield, Greycroft, and Prime Movers Lab all wrote checks. Basically the entire silicon supply chain plus a few growth funds, all backing the same idea. Adcock himself seeded the company with $100 million of his own money in late 2025.
That’s wild. It also might be one of the more rational deals of the year.
The mental model: Convex Bets
In my book I write about convexity as the shape of a payoff. A bet is convex when your downside is bounded but your upside isn’t. You can lose 1x. You can win 100x. If you can find one of those with even a small probability of payoff, the math says you take it. Then you take the next one.
VCs already think this way for seed rounds. What’s new here is doing it at $700M check size.
The bet isn’t “Hark will ship a great product.” The bet is: if Hark becomes the universal interface between humans and AI, our piece of a $6B cap table compounds into something that makes the check look small.
Run the rough math. If the probability of Hark hitting a generational outcome is even 3%, the expected value covers the round. And the strategics aren’t paying just for that 3%. Nvidia gets a marquee chip customer either way. AMD and Intel get the same option. Salesforce gets distribution intel. ARK gets a public-markets-ready position long before the IPO conversation starts. Everyone gets a front-row seat to what Adcock’s actually building.
That’s not one bet. That’s six bets stacked, three of which pay even if Hark flops.
The contrarian take
The reflex is to call this peak froth. It might be.
But the deeper read: when an asset has skewed payoffs, the right size of a position scales with conviction in the operator, not the maturity of the product. The investors didn’t underwrite Hark’s roadmap. They underwrote Adcock’s pattern. He turned Figure AI into one of the most valuable humanoid robotics companies on earth and took Archer Aviation public. That’s a different kind of diligence than what most pre-revenue rounds run.
For founders, the lesson isn’t to chase $700M rounds. It’s this: capital prices convexity. Build something where the downside is small, the upside is large, and the path between them isn’t a function of luck. Build it visibly. Then do it again.
At /mkt, we think about this every day. We’re building infrastructure for athlete tokenization through Reg A+ qualified offerings, with tZERO as the trading infrastructure partner. Regulated markets don’t reward stealth or hype. They reward operators who can build that same shape inside the rules. Different game. Same model. toolkit of 50 mental models, my book is coming soon.
Spencer Gareiss is CPO at /mkt. Former Senior TPM at Robinhood (derivatives & prediction markets). Army Reserve Captain. Author of "Mental Models: How to Think, Act, and Win." Nothing here is investment advice or a solicitation of any security.
