What’s 🔥 in AI/Infra/VC #522
What’s 🔥 in AI/Infra/VC #522
Market Notes From Our boldstart AGM 📈: The Consensus Machine Moves to Inception
Fortunately, I didn’t have much time to watch the markets over the last couple of days as we hosted our AGM, our annual general meeting for LPs. But wow, the AI trade is unwinding quickly. Earnings from Amazon, Google, Intel, and others came in below expectations, and the selling accelerated from there.
Impacting companies like OpenAI as many others.
And based on what I’m hearing, plenty of other deals are getting repriced as we speak.
That brings me to a few slides from our AGM showing just how concentrated capital has become.
It starts with the funds.
That’s just a different scale and kind of investing…
40% of this capital went to just Anthropic and OpenAI and the top ten deals accounted for almost 70% of the dollars! Rounds of $50M or more are now at 90% of all capital raised. Bigger funds write bigger checks!
There’s a reason, of course. Revenue is scaling at a pace we’ve never seen before.
The outcomes are jaw-dropping, too. The problem comes when every company is priced as though it will deliver one.
This has now reached inception, or company formation. While it’s not a perfect dataset, it shows how far the market has moved. Almost 40% of priced first rounds in US AI/ML companies are now $10M or more, up from 11% in 2023. And 17% exceed $20M, up from just 3%.
But smaller rounds haven’t disappeared. Roughly a third are still under $4M. The market is becoming a barbell, with less in the middle.
Jumbo has gone mainstream. And valuations have gone jumbo, too.
The median valuation has doubled since 2023, but at the top of the market, it’s more than five times higher.
At the 95th percentile, post-money valuations went from $57M in 2023 to $300M in 2026 YTD. That’s almost 10 times the median, compared with 3.8 times in 2023. The top of the market has entered a different stratosphere.
As round sizes grow, bigger funds enter the fray. Among inception rounds of $10M or more, nearly half now include a $500M-plus fund. Which also means that focused inception or seed funds with the proper sizing are also continuing to win their fair share of opportunities.
Our response continues to be earlier conviction, before the company becomes an auction. Fortunately, we anticipated this shift and armed ourselves with the capital to keep leading larger inception rounds, with ball control on price, terms, and ownership.
This is what bigger funds and consensus look like from day one. Yes, it’s a different game, and yes, Jeff is a rockstar. But for many early-stage funds, an exit at that $10B entry valuation would deliver enormous fund returns.
This all feeds the consensus machine, with the biggest funds converging on the same founder profiles and names.
AI kingmaking has now arrived at inception in full force. The biggest funds aren’t waiting for winners to emerge. They’re trying to crown them from day one.
The good news is that it leaves room for firms like us at boldstart to keep backing exceptional founders before it’s obvious. The next Cabbage Patch Kid or Labubu, before everyone’s fighting over one. That’s where a ton of alpha still lives.
There’s a lot more I’ll share from our AGM on inception, the tech, and robotics, but this feels especially relevant as the AI trade unwinds in public markets and investors start questioning how long they can keep funding companies that need ever more capital. For us, inception investing means getting in early enough, with the fund size (not too big, just right), ownership and entry price to generate meaningful returns without needing a $10B outcome, although we’ll always welcome one. That matters even more when capital gets tighter.
We also discussed how quickly venture funds are deploying and coming back to market, some just 12–18 months after their last raise. How long can that last? In a world changing this fast, time diversification matters more than ever, and so does backing founders who can adapt.
As always, 🙏🏼 for reading and please share with your friends and colleagues!
Thanks for reading What's Hot 🔥 in AI/Infra/VC! This post is public so feel free to share it.
Scaling Startups
#for all those VCs raising
#yes, this is what we at boldstart do!
Enterprise Tech
#one person and agents can just do things 👀
#great read - as intelligence gets cheaper, where does durable value accrue - gutenberg invented the printing press and went broke 📜 - value never stays where the tech is born. it settles with whoever owns the rails - “one either owns the rails or pays those who do”
#$6B stolen from Binance crypto exchange by North Koreans but not the wallets but by getting access to a single DevOps engineer’s laptop 🤯
#the internet as we knew it is gone - a world built for humans fading fast as evidenced by Reddit and even declines in Google search - worth a read 🤔
#major scientific and math breakthrough from OpenAI easily explained - means a lot of computer science ideas that said “this kind of puzzle is basically impossible to solve quickly” might be wrong, because an AI found a cleverer (still very slow) way to solve them
#huge breakthrough: Interactive RL in games and simulators is creating transferable reasoning that solves hard math problems. Expect faster scientific breakthroughs, stronger research agents that can accelerate AI itself, and better long-horizon computer-use agents. Public models will lag internal ones as progress shifts toward agentic training in rich environments.
#the labs are already working on the next frontier - everything I’ve been saying, solving super hard problems as AI gets easier and easier…chips, bio, physical sciences and engineering
more here as well from a few weeks ago
#i would say that most enterprises don’t even know where to start…
#🍿 databricks and snowflake CEOs battling out in public…
#inference economics and why even lesser than model cos with compute are becoming inference providers - tens of inference providers popping up every week and the whole game is utilization + software stack... same chips, wildly different margins, squeezing more out of every GPU is the margin 📈
#what the largest public security company CEO has to say about the recent security attacks
#big statement from Elon - robot memory will far exceed anything?
#stunning…
#time to build
#own your data, own your own AI
#congrats to Systalyze, a boldstart port co
#the new gold - GPUs and compute - also why hackers are attacking inference providers for API keys
Markets
#Anthropic S-1 dropped this past week but this is one of many surprising stats i read
#not just the infrastructure cos getting hit this past week - AI is making it easier and easier to rewrite many of the static rules built into ServiceNow’
What's Hot 🔥 in AI/Infra/VC is free today. But if you enjoyed this post, you can tell What's Hot 🔥 in AI/Infra/VC that their writing is valuable by pledging a future subscription. You won't be charged unless they enable payments.
© 2026 Ed Sim
Miami, FL 33134