Yann LeCun is taking a second swing at venture. But who’s actually pricing these bets? Before we get to today’s news: AI venture money is increasingly clustering around companies that claim a foundational role in model platforms, infrastructure, applied tools, agent security, and sovereign AI. Recent deals include Nvidia’s $75 million Series B extension for Sarvam AI, Obsidian Security’s $85 million Series D, Maximum’s $30 million seed round, and June’s $20 million pre-seed. This is Startup Fundraising. Today: celebrity operators raising money, a giant liquidity machine, and valuation jumps that need receipts. Let’s start with LeCun. Ana-Maria Stanciuc, writing in The Next Web:
Yann LeCun has found a VC fund that will actually launch. The Meta former chief AI scientist is joining 224 Ventures, a new firm that announced Wednesday with more than $100 million in assets under management. His co-partner is Oriol Vinyals, who until today served as Gemini’s co-technical lead at Google DeepMind.
That concentration story has a new gatekeeper: Yann LeCun and Oriol Vinyals are launching 224 Ventures with more than $100 million. Though the core fund is targeting $81 million and has raised about $30 million; the headline total includes SPVs. And they’re not leading rounds. So 224 is writing $1 million to $5 million checks into sub-$100 million startups, then letting somebody else set the price while a room full of AI insiders watches. Cozy. Having the University of Illinois as an anchor helps the institutional story, and LeCun plus Vinyals should see technical talent early. But Vinyals is also co-founding Discovery Loop on his way out of DeepMind—this is a very busy investing vehicle. In two years, they need a portfolio that proves those credentials win deals regular seed funds couldn’t—and companies that can build capital-efficient models instead of just collecting famous names on the cap table. This one's from GlobeNewswire:
G Squared, a global venture capital firm specializing in the private secondary market, today announced the final close of G Squared VII, its seventh flagship fund and largest to date, with $2.3B in capital commitments.
A $2.3 billion secondaries fund tells you the exit line is backed up. G Squared can buy employee stock, tender-offer paper, structured positions—the whole private-company waiting room. It’s their largest flagship yet. GlobeNewswire puts global secondary volume at $240 billion last year, up 48%. Venture is pricing liquidity in as a permanent layer now. Useful, sure. But let’s get the incentive right: when companies stay private longer, somebody has to monetize employee and early-investor paper. G Squared has raised $2.3 billion to own that bottleneck. Here’s the distinction: a secondary price can create liquidity, but it doesn’t prove a company deserves the valuation from its last preferred round. Cap tables can get cleaner while price discovery stays murky. From Lisa Brooks at MedCity News:
On Tuesday, an Israeli startup seeking to shrink that rate picked up $45 million in Series B financing. QuantHealth, which sells a simulation platform that tests clinical trials virtually before running them, has now raised $70 million since its founding in 2020. The Series B round was led by Qumra Capital.
Now this is a better AI-healthcare pitch: QuantHealth says it can simulate a trial before patients enroll, where roughly 90% of drugs eventually fail. Forty-five million dollars is cheap if it kills even one doomed Phase II before it burns years and a nine-figure budget. Qumra Capital led the $45 million Series B, with Sanofi Ventures joining Pitango HealthTech and others. That strategic participation matters—Sanofi has a front-row seat to whether these virtual trial predictions hold up in an actual development pipeline. But “AI plus biomedical knowledge graphs” is still brochure language. QuantHealth has raised $70 million since 2020, so tell me what a pharma company pays per simulated trial—and whether the model changes a protocol decision before the expensive humans get involved. The placement is smart. Drug-discovery AI has attracted plenty of capital at the front end; QuantHealth is aiming at the clinical bottleneck, where failure is painfully measurable. The proof is in prospective calls, not elegant retrospective simulations. SiliconANGLE writes:
HappyRobot Inc., a San Francisco-based artificial intelligence startup that automates enterprise operations, said today it raised $150 million, led by Prysm Capital and co-led by Eurazeo, bringing the company’s post-money valuation to $1.2 billion.
HappyRobot went from a $44 million Series B in September to a $150 million Series C at $1.2 billion post-money. Prysm led and Eurazeo co-led. This looks like a real cap-table repricing, rather than existing investors simply keeping the lights on. “Automates enterprise operations” is still a fog machine. The company says it has 150-plus enterprise customers and grew more than fivefold since September. Okay—now tell me how much those customers pay, how long they stay, and which workflow they can’t live without. The customer list has names with operational mess baked in: DHL, Kuehne + Nagel, Repsol, Uber. If agents are sorting calls, documents, scheduling, and incoming exceptions inside those businesses, that can get embedded fast. Maybe. But at $150 million, they’re funded to prove this works across ugly enterprise systems at scale—not just keep producing polished agent demos. Two years from now, $1.2 billion makes sense only if those 150 customers become durable, expanding software revenue. Here's TechNode:
PaXini has completed a RMB1 billion strategic funding round, bringing its cumulative fundraising to RMB3.5 billion, the company said on Aug. 3. The new round was jointly led by an unnamed global consumer electronics and semiconductor group, BOC International Investment, Kunpeng Fund and Hexin Fangce.
Nearly one million tactile-sensing chips shipped in a year—that’s the number that matters. PaXini’s raising RMB1 billion to add capacity because it has hardware moving, not because somebody drew a robot hand on a pitch deck. BOC International and Kunpeng jointly led alongside an unnamed global consumer electronics and semiconductor group. “Anonymous lead investor” is a curious way to announce a very large conviction check. Beyond exposure, what did that semiconductor group buy—supply access, design wins, or a route to overseas distribution? PaXini is already at RMB3.5 billion raised. Its R&D, capacity, and global expansion had better turn those chips into systems revenue. PaXini can at least point to shipment volume as it scales. That doesn’t answer questions about margins or customer concentration, but it gives this round more physical footing than the usual embodied-AI flourish. If today’s briefing was useful, please subscribe and leave us a review wherever you’re listening. Your feedback helps other people discover Startup Fundraising and keeps these daily updates going.
Links to every story are in the show notes, so take a look at any you’d like to explore further. Thanks for listening, and we’ll be back tomorrow. That’s it for Startup Fundraising today. This is a Lantern Podcast.