World Models, Wild Money - from first checks to $6–10B in just a few years 🤯
While Anthropic preps for world domination in the enterprise, physical AI had an incredible week. the connection? the last mile - turning impressive models into systems that actually work.
first, Anthropic. it’s committing $100M to train 10,000 “Frontier Deployed Engineers” by the end of 2027 through its new Claude Frontier Academy. The first cohorts include engineers from major consultancies and enterprises. Let’s call that brute forcing the last mile problem with dollars moneybag💰.
This is a smart move by Anthropic as by default it has to build a product for everyone, but making agents work inside an enterprise requires customization, context and a deep understanding of workflows. Someone still needs to connect it to the business, figure out what good looks like, and make it reliable enough to use.
And yeah - you know my fox-in-the-henhouse question 😂. how much of your operational know-how do you put in the hands of a platform that also serves your competitors? Where are the boundaries? And once your workflows, integrations and evaluations are built around that platform, how easy is it to leave? Brilliant strategy for becoming deeply embedded.
Now let’s take that last-mile problem into the physical world. An enterprise agent has to navigate your systems and processes, but a robot has to navigate a changing environment, decide what to do, and deal with the consequences. Different challenges for sure but both push AI beyond producing an answer and toward taking useful action.
With that, world models are having an insane moment this week.
World Labs joining AMD in an $8.2B deal. General Intuition raising another $220M at a $6.2B valuation. Field AI reportedly raising $700M at $10B. extraordinary numbers - and the timelines below make them even more striking.
These companies were founded in 2023 or 2025. Their first checks were already $50 to $100M and within a few years they are worth $5-10B. The speed from company creation to strategic relevance is absolutely mind blowing 🤯.
World Labs is the headline as an $8.2B exit with a few years and no revenue is simply remarkable. Yes Fei-Fei Li is an absolute legend wind will be joining AMD as EVP and chief scientist. But this is more than a world model as Patrick Moohead highlights - its a much broader bet than a single physical AI application and the team is intended to become a nucleus for attracting talent.
General Intuition brings another approach - learning from billions of action-labeled videos to build models that can act in unfamiliar environments. World models help predict what happens after an action; action models help decide what to do. Understanding the relationship between action and consequence is the big idea here.
and Field AI brings the story into robot autonomy, with its foundation models built around a belief world model. the reported $10B valuation is another signal of how much capital is chasing the ability to operate reliably in the physical world.
We’ve seen the frontier labs establish this playbook. OpenAI started in 2015; Anthropic in 2021. exceptional teams building foundational capabilities became enormously valuable - and strategically essential. Now spatial and physical AI are arriving with that conviction already in place. World models aren’t the only approach here. But investors are clearly betting on physical AI as the next big wave - and moving fast.
And yeah - capital moves faster than deployment. These valuations don’t tell us when robots will become reliable and widespread. But from Anthropic putting $100M behind enterprise deployment to physical AI teams attracting billions within a few years of founding, the message is clear. Everyone is racing to turn model capability into useful action. Solving that last mile is where a lot of the value will be created.
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.
#hard agree with Shaun - we have our own office in SF as well and while I love to spend time there I also like to not be there to think on my own, away from the echo chamber
#how to extract clear understanding and value from LLMs from Andrej -i’m a huge fan of creating dynamic HTLM visualizations for example to remove AI slop and better understand or explain concepts - read below - our job comes down to managing our AI and taste
#this a16z state of markets chart deck is fantastic, pulled a few slides for you…full deck here
and yes this chart is what many of us are banking on 😀
# congrats to Topos Bio a boldstart port co…this is a big one. 🔥 We’ve spent the last few years watching AI make huge leaps in language, images and video. Now we’re seeing the same kind of step-function progress in biology.
Proteins don’t sit still. They move constantly and that’s one reason many of the most important disease targets have remained “undruggable.” ToposBio just introduced Topos-2 + Topos-Bind to model that motion.
And the benchmarks are 🤯: Topos-2 ranked #1 across 10+ models and came nearly 4x closer to experimental ground truth than the next-best model.
If this works at scale, it opens up an entirely new class of targets for drug discovery.
#wow, didn’t see this coming…OpenAI enterprise customers can now run competitive open models inside Codex itself and bill against the OpenAI spend they already committed, with no new vendor contract.
This turns Codex into a multi-model router rather than a closed OpenAI-only tool
#the real breakthrough here 👇🏼 -wiring a quantum chip straight into @nvidia GPUs means AI can catch and fix qubit errors as they happen - error correction is the whole ballgame for useful quantum and AI x quantum is going to move a lot faster than people think
#first multiplayer/team bot from Grok - shared AI teammates, so easy to build and use, shared skill/context improves from whole team interaction and always up to date
“He who has the gold that gets the compute makes the rules,” said Ed Sim, founder and GP at Boldstart Ventures, an investor in agent identity company Keycard”
#congrats Reco (a boldstart port co) - The best signal in enterprise security is a customer who writes a check. AT&T went from Reco customer to investor. Agents are getting keys to every app, and Reco maps what each one can touch.
