Newsletter · 2026-07-06 · Ashutosh Agarwal

Anthropic Alumni Raise a $200M Seed and Groq's Salary-for-Equity Save - The Raise - Week of June 29 - July 5, 2026

The Raise for the week of June 29 - July 5, 2026. Anthropic alumni closed a $200M seed at a $1B valuation with a16z, Kleiner and NVIDIA in, Groq's founder detailed the salary-for-equity gambit that saved the company, and 'seed round' stopped meaning anything.

The Raise

Week of June 29 - July 5, 2026: Anthropic Alumni Raise a $200M Seed and Groq's Salary-for-Equity Save

This was the week "seed round" officially lost all meaning. On June 29th, a startup you'd never heard of a week ago, Mirendil, founded by a group of Anthropic defectors, closed a $200 million seed round at a $1 billion valuation, with Andreessen Horowitz, Kleiner Perkins, and NVIDIA all writing checks. Read that again: two hundred million dollars. Seed.

I've been doing this long enough to remember when a seed round was two engineers, a deck, and $1.5M if you were lucky. Now it's a unicorn valuation before there's a product. And the punchline is almost too perfect: Mirendil's whole pitch is that frontier models are about to commoditize. They're building self-improving AI (recursive self-improvement, the exact thing Anthropic has been on the public record lobbying to regulate) so that anyone can spin up specialized in-house models in medicine or materials. In other words, the smartest money in the Valley just paid a billion-dollar entry price to bet that billion-dollar frontier labs are worth less than everyone thinks.

I don't hate the trade. If Mirendil is right, the moat around Anthropic and OpenAI is a puddle, and OpenAI quietly punting its IPO from 2026 to 2027 this same week, to avoid a down round per the reporting, tells you the public market already smells it. But let's be honest about what a $200M seed actually is: it's a Series B with a seed-round story attached so the founders keep more control and the headline sounds spicier. When the label stops meaning anything, that's usually late in the cycle, not early.

This Week's Rounds

Mirendil, $200M seed at a $1B valuation. Led by Andreessen Horowitz, Kleiner Perkins, and NVIDIA. Anthropic alumni building self-improving ("recursive") AI so customers can train specialized in-house models: one of the larger seed valuations for a new AI company in recent memory. Venture In The South, "E250 The Week In Venture" (June 29, 2026)
Cowboy Space Corporation, $365M Series A at a $2B valuation. Just 60 employees, and lead investors were not named on the pod. Orbital-energy startup from Robinhood co-founder Baiju Bhatt, building satellites that harvest solar power in low-Earth orbit and, next, AI data centers in space that beam computed data back down by laser. Second in Command, "Ep. 593, Cowboy Space Corporation COO & CLO Joe Yaffe" (July 2, 2026)
DeepSeek, closing on a ~$7.4B round. Led by Tencent, battery maker CATL, and China's state-backed National AI Investment Fund. The Chinese lab is doubling headcount and opening to outside capital for the first time, though as Scott Galloway's guest flagged, even a $7.4B "mega-round" is maybe half what a comparable US lab would command. The Prof G Pod, "China Decode: Apple's China Chip Play, DeepSeek Seeking Billions..." (June 30, 2026)
Venice AI, $65M at a $1B valuation. Its first outside round, led by Dragonfly and Coinbase Ventures. Private, "uncensorable" AI that routes users across 200 models and stores nothing on its own servers: already ~3M users and ~$70M in annualized revenue. FOMO HOUR, "'Robinhood Chain' Debuts... Venice AI $1B Valuation" (July 2, 2026)
Integral, $18M Series A. Lead undisclosed on-air. Co-founder/CEO Shubh Sinha is building privacy "sanitization" for proprietary datasets (think medical records and financial transactions) so AI labs get the signal without the PII. TBPN, "MetaCloud, Bending Spoons IPO... Integral" (July 1, 2026)
SpeedLabs, $6.5M seed. Led by Parlay Capital (PrizePicks' seed lead), with Bullpen Capital (FanDuel's first backer), a former PrizePicks board member, and Betting Startups Capital. Turns live sports narratives into "momentum markets" for prediction-market and sportsbook venues. Founder Nick Meader set out to raise $3–4M and got talked up to $6.5M by demand. The Betting Startups Podcast, "Ep. 217: Turning sports narratives into 'momentum markets' w/ Nick Meader" (June 30, 2026)
Frozen One, $5.75M seed. Led by Brand Foundry Ventures (Graza, Olipop), with Supernatural Ventures, Abe Burns, and Grammy-winner Ryan Tedder. High-protein ice cream (40g protein, under 400 calories a pint) scaling national retail after a strong Target launch in April: the raise comes just months after an oversubscribed $2M round. CPG Week by BevNET & Nosh, "A Sweet Acquisition, a Bitter Trademark Battle and Fancy Food Finds" (July 2, 2026)

Founder Story of the Week: Groq

The best founder tape this week wasn't about a raise that closed, it was about the raise that almost didn't happen. On July 5th, Jonathan Ross, founder of Groq, sat down with David Senra and walked through the near-death experience that built the company now doing a rumored $20 billion partnership with NVIDIA, a deal, he says, that went from first phone call to money-in-the-bank in about three weeks.

The part that stuck with me is why Groq nearly died first. For years, West Coast VCs wouldn't touch it. Ross's theory of the case is brutal and, I think, mostly right:

"West Coast VCs are more like Lemmings. And typical East Coast VCs all think that they're smarter than each other."

One coastal pass, he argues, cascades into a dozen; Groq ended up funded largely by East Coast crossover funds while the Sand Hill Road crowd chased "safer" bets, and then watched NVIDIA do the biggest deal in its history around the company they skipped.

But the masterstroke came when the money actually ran out. Instead of the layoff list his leadership team had drawn up (a list that would have cut the exact engineers building Groq's never-been-done compiler) Ross put "World War II looking pictures of war bonds" on the wall at an all-hands and pitched Groq bonds:

"It wasn't technically a bond. It was an exchange of salary for equity."

Roughly 80% of employees opted in; many cut down to the statutory minimum wage. It bought two months of runway against three weeks of cash, the margin between existing and not. His framing for why attrition actually fell: "put everyone's hands on the steering wheel." Passengers panic on a scary road; drivers lean in. It's the most honest fundraising lesson I heard all week, and no term sheet was involved.

Also Heard

The best "how I raised" masterclass came from GoldCast's Palash Soni on The Neon Show: an F1-visa founder racing an O1-visa clock, landing an Acfor term sheet in two days (Tuesday pitch, Thursday paper), closing ~$1.5M at a $5M cap, then $10M at a $38M post led by Unusual's $7.5M check with Scott Belsky and Lenny Rachitsky as angels. His advice to first-timers: survive before you optimize valuation, "Zero into zero is zero." The Neon Show, "How to Beat a Competitor With 100x Your Funding | Palash Soni, GoldCast Founder" (June 30, 2026)
Lime went public, with CEO Wayne Ting calling it "the last man standing" in micromobility: a $7.50-per-vehicle-per-day business run at 50%+ cash margins and sub-one-year payback. Proof that operational grind, not the biggest raise, wins the low-margin games. TBPN, "MetaCloud, Bending Spoons IPO... Lime" (July 1, 2026)
Google took its first-ever equity stake in a film studio, a $75M investment into A24 dressed up as a multi-year AI research partnership with DeepMind. Call it what it is: taste-washing to make VEO look like a tool for elite creatives rather than a job-killer. Venture In The South, "E250 The Week In Venture" (June 29, 2026)
OpenAI blinked on its IPO, pushing the listing from 2026 to 2027 rather than go public at a haircut to its $1T-plus target. When the category leader would rather wait than test the tape, take the note. Venture In The South, "E250 The Week In Venture" (June 29, 2026)

Every item above is sourced from podcast episodes published June 29–July 5, 2026. Rounds are reported only where the episode actually stated them; where a lead investor or valuation wasn't disclosed on air, I said so rather than guess.