Newsletter · · Ashutosh Agarwal

AI Valuations Go Vertical While a Five Person Biotech Raises the Old Way - The Raise - Week of September 7, 2026

The Raise for the week of September 7, 2026: Cognition at roughly 46 billion dollars, Clay at 7 billion and Instinct quintupling in weeks made this the frothiest week of the year, while Infinimmune's five scientists showed that a clean cap table and hitting self-set deadlines still buys an 838 million dollar Merck partnership.

The Raise

Week of September 7, 2026: AI Valuations Go Vertical While a Five Person Biotech Raises the Old Way

The week AI valuations stopped pretending to make sense, and a five-person biotech showed the other way to do it.

On Wednesday (Sep 3), Harry Stebbings opened his weekly 20VC news roundup with a run of numbers that, five years ago, would have been the story of the decade. Instead they were just Tuesday's headlines. Cognition, the maker of the AI coding agent Devin, is raising at roughly $46–47 billion, up from $26 billion in May. Clay, a sales-prospecting tool, at $7 billion. Linear, a project-tracker beloved by engineers, at $2.5 billion. And a barely-known AI assistant called Instinct went from a $500 million valuation to $2.5 billion in a matter of weeks. (The Twenty Minute VC (20VC), "NVIDIA Crushes Quarter and Buys Hugging Face | OpenAI Cuts Off Cursor | Instinct Hits $2.5BN Valuation and The Race for AI Assistants | Cognition Raises at $46BN, Linear $2.5BN and Clay $7BN," 2026-09-03)

I'll say the quiet part out loud: this is the frothiest week I've seen all year, and I've been an AI optimist longer than most. When a company's valuation can 5x in the time it takes the rest of us to renew a domain name, we are no longer pricing businesses, we're pricing FOMO. Cognition reportedly has close to $10 billion of investor interest chasing a roughly $1 billion round, according to Bloomberg's reporting on the raise. That is not a fundraise; that is an auction with a velvet rope. (Bloomberg Tech, "Nvidia Nears $14B Hugging Face Deal," 2026-09-02)

And yet. Some of these numbers are attached to real, fast-compounding revenue. Cognition is said to be running past $900 million in annualized revenue and projecting $1.6 billion by year-end; Linear is at $100 million in revenue growing 100%. So the honest read isn't "it's all a bubble." It's "the price of being right about AI has gone vertical, and the margin for being wrong has vanished." Keep that in mind as you read the rounds below, because the best story of the week wasn't a mega-round at all. It was a five-person team in a shoebox office who raised $12 million by simply refusing to waste anyone's time. More on that after the rounds.

This Week's Rounds

  • Cognition: raising ~$1B at roughly a $46–47B valuation (up from $26B in May), lead investor not yet named. The company behind Devin, the "AI software engineer" that writes and ships code autonomously. Reportedly ~$900M annualized revenue and climbing; investor demand said to be near $10B. (20VC, 2026-09-03)
  • StarCloud: $250M at a $2.3B valuation, led by Manhattan West with NVIDIA (NVDA) and Cisco (CSCO) among the backers. Building data centers in space to run AI compute off solar power in orbit. Notably, that's more than a 2x markup from a $1.15B valuation just five months earlier. (This Week in Startups, "VC experts on why Physical AI funding is heating up | E2333," 2026-09-02)
  • Wonderful: $550M Series C. A "forward-deployed" AI operating system aimed at big, non-tech enterprises doing $750M+ in revenue. Founder Bar Winkler started it in January 2025 and is deliberately chasing legacy industries rather than Silicon Valley darlings. (TBPN, "Jobs Boom, Astra Reactions, Cybercab Roll Out, The Dyson Debate | Hunter Somerville, Bar Winkler, Bridgit Mendler, Carina Hong & Ken Ono," 2026-09-05)
  • HiBob: $166M led by Salesforce (CRM) at a $3.2B post-money valuation. The Israeli-born HR software platform ("Bob") used by mid-size companies. It's now raised ~$700M total and is pushing its IPO out to at least 2027. (The Chad & Cheese Podcast, "HiBob Beast Mode & DOL Dismantles Worker Protections," 2026-09-04)
  • Clay: raising at a $7B valuation from Wellington. A go-to-market data platform that automates sales prospecting and lead enrichment. (20VC, 2026-09-03)
  • Linear: $2.5B valuation in a tender offer (letting employees and early investors cash out). The issue-tracking and project-management tool for software teams; ~$100M revenue growing 100%. (20VC, 2026-09-03)
  • Instinct: raising at a $2.5B valuation, up from ~$500M weeks earlier, with early backing from Bret Taylor. Founded by Noah Shin, it's an AI assistant you run your life through by text message and phone call. (Elon Musk Podcast, "Texting an AI to run your life," 2026-08-31; 20VC, 2026-09-03)
  • Facet: $68M Series B led by SBI (Japan) at a $1B valuation, just three months after a $51M Series A in May, bringing total funding to $119M. An LA-based stablecoin neobank processing $40B+ in annual volume across 125 countries. (Tokenized, "Revolut Launched a Stablecoin. Are Big Banks Next?," 2026-08-31)
  • Felix Pago: $200M in equity and debt (~$85M equity) from Andreessen Horowitz, QED, Castle Island, Swish Ventures and General Catalyst. Stablecoin-powered cross-border payments and remittances. (Empire, "Robinhood Is Proving Distribution Captures Crypto's Value | Weekly Roundup," 2026-09-04)
  • Levanta: $22M Series B led by Volition Capital, taking total funding past $43M. A creator-commerce and affiliate platform connecting 90,000+ vetted creators to brands across Amazon, Shopify and Walmart. (The CPG Guys, "Commerce Riff with Sri & PVSB - September 1, 2026," 2026-09-01)

A quick honesty note: several of the week's flashiest "big numbers" were acquisitions, not funding rounds, and I've kept them out of the list above. You'll find them in Also Heard.

Founder Story of the Week

If the rounds above are the sugar rush, Infinimmune is the actual meal. On this week's Raising Biotech (Sep 3), co-founder and CEO Wyatt McDonnell walked through how five scientists who quit 10x Genomics (TXG) together turned a "junk data" hunch into a $12M seed, then an $838M partnership with Merck (MRK), and, as of last month, a $75M Series A. It's the clearest fundraising playbook I've heard in ages, and almost none of it is about the pitch deck.

The setup: at 10x Genomics, McDonnell had used single-cell sequencing to pull thousands of virus-fighting antibodies out of human blood samples in days during COVID. Digging through public databases afterward, his team kept finding strange antibodies that "were mostly viewed as junk reads." In plain English: everyone else's software was throwing away the interesting stuff. That was the whole thesis.

Here's the part I loved. Most founders agonize over whether to bootstrap on government grants or raise venture money. McDonnell's team debated it for about five minutes:

"I think initially we'd sat down as a team of founders and thought, well, maybe we should do an SBIR grant. And then I said, no, we should go raise a big seed round right now. And my co-founders just sort of chuckled and said, yeah, pat me on the head like, okay, Wyatt. And I'm like, yeah, let's go raise a seed round. And then it happened."

His method for finding investors is one every first-time founder should steal, which is to reverse-engineer the cap tables of companies you admire:

"Trace those back. You know, maybe they're at a series A or a B or a C now, but go back to the seed round and who were the first checks in the door?"

That trail led him to Josh Elkington at Axial VC, who became his "initial center of capital gravity." One meeting tells you how conviction actually spreads in venture:

"I was supposed to get a half hour walk and talk with him. And that turned into a three and a half hour walk throughout most of Berkeley's campus and most of the surrounding neighborhoods."

Ben Kim and Jory Bell at Playground Global ended up leading the round. It closed in October 2022, three months after the company existed. When I asked myself why it moved so fast, McDonnell's answer was refreshingly unglamorous: seriousness is legible.

"We came out of the gate strong and said, we are raising a seed round. We have started a company around this. We have a clean cap table on the Delaware C Corp and a world-class set of technologists, all of whom left their full-time jobs to work full-time on this. Like it's, it's very, it's a strong signal."

He also nailed the psychology of the first check better than most VCs I know, the fear that keeps a lead investor from committing:

"People are thinking about risk-free returns, not return-free risk. And if you're the first money in the door, it can be hard to overcome that fear of return-free risk."

Then the punchline every over-planning founder needs to hear. The team budgeted two years to build its two core technology platforms. They finished the first in three months and the second in four:

"We didn't budget for success, which is an interesting problem to have."

That speed is what eventually won over Merck. And notice how Merck decided to trust a four-year-old company with targets from its own drug pipeline: not with a pitch, but by watching them hit self-set deadlines, over and over, for more than a year:

"Why on earth would we give you high-priority targets from our therapeutic pipeline as a global pharma? We need to know you're going to execute very successfully with conviction and quite aggressively."

The Merck collaboration (worth up to $838M in milestones) was announced in March 2026; the $75M Series A, co-led by Playground Global and Regeneron Ventures with RA Capital and Merck's venture arm joining, closed in August. Two of Infinimmune's own antibody drugs, for eczema, are headed into human trials in 2027. (Raising Biotech, "S4, E8: Infinimmune – From a Scientific Hunch to an $838M Merck Partnership, with CEO and co-founder Wyatt McDonell," 2026-09-03)

The contrast with the top of this newsletter is the whole point. One end of the market is paying billions for momentum. The other end just quietly proved that the oldest fundraising edge there is, a serious team, a clean cap table, and doing exactly what you said you'd do, still works.

Also Heard

  • NVIDIA (NVDA) is buying Hugging Face, the open-source AI model hub, for a reported ~$13 billion (figures ranged from $12.9B to ~$14B across shows, including ~$1B in employee retention stock). Hugging Face was last valued around $4.5B in 2023, so this is a massive markup and a land-grab for the developer community. This is an acquisition, not a round, but it was the most-discussed deal of the week across Bloomberg Tech, Morning Brew Daily and Motley Fool. (Bloomberg Tech, "Nvidia Nears $14B Hugging Face Deal," 2026-09-02; Morning Brew Daily, "Nvidia Buys Hugging Face for $12.9B & Pharma Companies are Bullish on Baldness," 2026-09-04; Motley Fool Hidden Gems Investing, "A $13 Billion Bet on Robots?," 2026-08-31)
  • Stripe acquired OpenRouter for ~$7.5 billion, its biggest AI bet yet. OpenRouter routes AI workloads across 400+ models and was reportedly at ~$160M annualized revenue by mid-August, a wild outcome for a company whose seed round priced it at $60M in late 2024. Its three co-founders reportedly netted ~$1.5B combined. (Market Maker, "Stripe's $7.5 Billion AI Bet: Why It Just Bought OpenRouter," 2026-08-31)
  • NVIDIA is investing $3.5 billion in Taiwan's MediaTek as part of a multi-year chip partnership, pocket change against NVIDIA's $200B+ in free cash flow, but a signal about how it wants to shape the inference-chip ecosystem. (Bloomberg Intelligence, "Nvidia to Invest $3.5 Billion in Chipmaker MediaTek," 2026-08-31)
  • Smaller crypto and fintech rounds from the week's roundups: RQD Clearing raised $74M (Bain Capital); Groma $17.5M (Coinbase Ventures, F Prime, Valor Equity, Castle Island); DiameterPay $10M; Firelight $8M. (On The Brink with Castle Island, "Weekly Roundup 09/04/26," 2026-09-04)
  • Founder note: Greg Krasnov detailed building Tonik, a digital bank in the Philippines, on less than $10M of spend to launch, cheap for a licensed bank, and raising $17M in pre-Series B funding ($44M total to date). A good reminder that not every great company needs a $2B headline to be a great company. (Hustleshare, "The Hustle Behind Tonik Bank with Greg Krasnov | Hustleshare Ep 395," 2026-09-06)