Over 200 subscribers
The narrative on venture has been negative for the past couple of years:
Mega firms are concentrating all the capital in the market
Exits take longer, companies stay private forever
No more cash exits from large tech companies, who are investing / inventing internally
Massive competition for smaller funds, as mega funds write lots of pre-seed and seed tickets as options for later rounds
Consolidation inevitable
and so on.
But this week, it seems like the narrative reached another bigger, more specific breaking point.
It seems almost like VC is changing forever.
Many people think that VC will turn into a version of PE, raising larger and larger funds, aiming for 2-3x on each deal, investing massive amounts on each company and trying to “king make”.
The article that lit everything on fire is: Lightspeed is latest firm to shift away from classic VC model (and will be raising around $7B for their new funds).
So Lightspeed followed all the big ones (Sequoia, A16Z who is raising a $20B fund, GC) into becoming a RIA. But that’s not all:
Thrive launched a new $1B fund called Thrive Holdings which is a pure PE M&A fund, buying and re-selling AI infused companies.
VCs are searching for liquidity wherever they can, with this week all the VCs in Royal Match exited to a PE firm
Coatue is launching a new public-private fund with, listen to this: 1.25% management fees, 12.5% carry, a 5% hurdle rate, redemptions (!!!) and a $50k minimum entry ticket.
Coatue’s fund, especially, should really be a wake up bell for the market. When a market participant starts lowering prices and dramatically adapting different models, it clearly shows that the market is not providing the expected demand for their more traditional prices.
So now it really seems that the top firms either believe there is no more alpha in early stage venture, and the money is to be made in massive capital aggregation with the goals of becoming tech’s Blackrock - OR that outcomes will be so stupendously bigger in the new AI era, that venture math is still going to work with such sizes.
What’s a market observer or player to do with all of this?
The sands are definitely shifting underneath our feet, and as active investors we constantly have to evaluate what the market for capital at the later stages looks like.
It’s easy to think that a number of people might have just completely lost the plot. $20B venture funds? $10B vals for pre-product companies? $3B acquisitions for VS code forks? $9B vals for Cursor? Allocations being offered to every HNWI in the world? Smells euphoric and it’s very, very hard to make the math, math.
Over here, we’re thinking that easy thought . Playing with the venture math model, we don’t really see the math of funds in the 100-500M range breaking for top performers - but we think that we will continue to see increased experimentation with models and fee levels (as it should be!) as the current market can remain irrational for longer than players can remain solvent.
What’s clear is that the 2021 euphoria has brought too much capital into the market and distorted everything (the 2025 AI distortion in our view is doing exactly the same).
We recommend listening to Josh Kopelman on Uncapped who explains venture math very clearly.
The biggest talk of the town this week has got to be Benchmark’s investment in Manus at a $500M valuation. The announcement has sparked controversies along the US-China tech nexus, with Delian not mincing words and calling Gurley a traitor.
What is happening here is that potentially Benchmark is thinking in terms of geopolitics math, and taking a massive bet on valuation arbitrage with geopolitical risk.
We understand more Benchmark doing Manus at $500M than the $10B pre-product vals over here.
More on the AI bubble:
Cursor hits $9bn valuation: $900M from Thrive and A16Z.
And:
OpenAI released a massive PDF with their guide to building agents. Also rolled back their overtly sycophantic 4o update, and published an explanation.
deepseek-ai/DeepSeek-Prover-V2: DeepSeek-AI is gearing up for a big leap in AI development, inviting contributors to join its GitHub project, which hints at collaborative innovation that could redefine how we approach artificial intelligence in practical applications.
Is it a race to the bottom for the model providers / consumer plays, when you can just export most of your data and run it in the best model available?
Eric Scmidth believes that AI will consumer 99% of the world’s energy. There’s also a must watch video interview with him, where he speaks about China and a piece by Schmidt and Selina Xu on what China used to be, what it is now, and what to expect.
Stripe just launched a payments foundation model (and a bunch of other stuff)
As missiles start flying across the India-Pakistan border, our usual brief on technologies and spies.
CIA trying to recruit Chinese spies, in a bit overly open way. Interesting to see alongside China’s efforts to strengthen their soft power and build up their sphere of influence.
Nvidia CEO Jensen Huang Sounds Alarm As 50% Of AI Researchers Are Chinese, Urges America To Reskill Amid 'Infinite Game': Jensen is rallying U.S. leaders to prioritize AI investment and workforce reskilling, warning that without proactive measures, the nation risks falling behind in the global tech race, especially with half of the world's AI researchers based in China, while also hinting at a future where robots could fill a massive impending labor shortage.
And as if Jensen wasn’t clear enough: Harvard’s former chemistry chair takes new position at Chinese university AND Microsoft researcher moves to China as AI assistant professor.
Back-of-the-envelope calculation of the bill of materials for humanoid robots within and without Chinese supply chains.
The Techno-Industrial Policy Playbook - an excellent reports with strategies and recommendations on how to strengthen America’s technology and industries, from the NIH to naval shipbuilding and biological threats.
Is China Set to Dominate the Future of Fusion Energy? We’ve been reporting on China’s efforts in fusion energy. Quote from the article: “Even if China is not ahead right now, when you look at how quickly they build things, and the financial willpower to build these facilities at scale, the trajectory is not favorable for the U.S.”. And some updates on BEST.
Full walkthrough of the Shanghai Auto Show 2025, it’s 3+h. If you want to know what a mid car looks like, you can also just watch a 30sec Youtube short, there’s a million of them. EU carmakers are absolutely done. No innovation in 30 years, cars look all the same, no consumer focused features, exorbitant prices, no status symbol anymore, it’s over.
On the same topic, How BYD foresaw itself overtaking Toyota back in 2007: fun book review of a BYD book that’s out only in Chinese. The journey from a humble battery manufacturer to the largest electric carmaker globally through pivots to EVs and integrated supply chains. Investors yelled at Chuanfu Wang, saying, “You don’t even have a driver’s license — how can you make cars?”, absolutely classic.
The founding partner on 8VC on how China eats everything
Hyundai bets $21B on Atlas humanoid robots for US car assembly: Hyundai is deploying Boston Dynamics' Atlas robots at its new Georgia plant, poised to handle up to 40% of vehicle assembly tasks. Looks like they really believe it.
RobustDexGrasp shows how RL-based closed-loop dynamic control can make a robot grasping task generalizable to hundreds of unseen objects.
VCpedia - Startup Funding Intelligence: Probably the most useful startup news site nowadays, all AI built.
FutureHouse Platform: Superintelligent AI Agents for Scientific Discovery: FutureHouse is unleashing a squad of AI agents designed to break the data bottleneck in scientific research, dramatically speeding up literature searches and synthesis.
In a similar vein, the Agentic BioML Challenge
Who will be the first to die at the hands of a robot?
VersionOne’s thesis on robotics
Universal Antivenom May Grow Out of Man Who Let Snakes Bite Him 200 Times: Tim Friede's extreme self-experimentation with snake venom could pave the way for a universal antivenom, potentially saving millions from fatal bites by neutralizing toxins from various species with just one treatment.
Rainmaker raises $25m series A to scale their cloud seeding technology.
If you enjoyed reading the Colossus piece on Matt Huang a few weeks ago, you’ll love this new profile on Neil Mehta, early backer of Coupang and legendary founder of Greenoaks.
Fertility on demand: Emerging technologies in reproductive health promise to extend women's fertility into their forties and fifties, potentially leveling the playing field in the workforce by allowing women to pursue careers without the strict time constraints of motherhood.
Programmable Power: Energy Is Becoming a Software-Defined Network: nice thesis around software redefining energy and how that changes the whole grid.
Let us know what you think of this format, and if you have anything else we should read!
The article is a testament to the value of in - depth reporting.
I like this format more than the previous one - good job