The Week, Off Sand Hill: The Realities of Physical AI
108 rounds, $8.13B disclosed, and where capital goes when token growth slows down.
Adam's take
Venture capital closed 108 deals totaling $8.13B across the ecosystem this week, with AI & Machine Learning capturing the top sector spot and General Catalyst leading activity across 3 rounds.
Yet behind the headline volumes, a different shift is playing out. As reported this week, token spend growth plummeted to 1.5% across August and September, falling sharply from a 42% pace recorded in July. At the same time, frontier model leaders have pulled back valuation targets. The initial cloud-software land grab of foundational models has collided with basic margin economics.
I think capital is quietly rotating away from generalized chat interfaces and steering directly into heavy real-world execution. The week's biggest capital outlays targeted structural assets: Google acquired an unfinished data center campus in Wyoming with plans to deploy $17 billion, while companies translating machine intelligence into physical execution pulled in massive equity injections. Path Robotics took in $300 million for autonomous manufacturing systems, and Munich-based SE3 secured $30 million to navigate physical environments.
My bet is that foundational models will function increasingly like commoditized compute utilities rather than standalone venture-scale empires. When digital generation becomes cheap, value aggregates around two areas: proprietary access to physical assets or deep industry distribution. Enterprise automation startups like Mercor, which secured $750 million at a $20 billion valuation, show where capital clusters when models are treated as components rather than products.
If I were a founder raising today, I would stop leading pitch decks with model parameters or API wrapped features. The teams attracting checks right now are those anchoring software directly into hardware, proprietary manufacturing, or proven enterprise revenue workflows.
— Adam
The week in numbers
- Rounds tracked: 108
- Disclosed: $8.13B
- Top sector: AI & Machine Learning
- Most active investor: General Catalyst (3 deals)
Five stories that mattered
Mercor raises $750M Series D at $20B valuation
Mercor closed a $750 million Series D equity financing round on Nov. 10 that established a $20 billion valuation for the business. Corporate backers Nvidia and Salesforce participated in the transaction.
Google buys Crusoe data center site in Wyoming, plans $17B campus
Google purchased an unfinished data center project site in Wyoming from Crusoe. The buyer announced plans to direct $17 billion of capital into building out the computing campus.
AI token spending slowed to 1.5% as model builders trim valuations
Spend on artificial intelligence tokens expanded by 1.5% over August and September, dropping down from a 42% increase in July. Meanwhile, developers Anthropic and OpenAI adjusted their forward valuation metrics downward.
Path Robotics raises $300M Series E led by Georgian
Path Robotics completed a $300 million Series E funding round led by Georgian, valuing the company at $2.45 billion. The capital will support development of the firm's physical AI manufacturing platforms.
SE3 raises $30M Series A led by Accel for autonomous navigation software
Spatial intelligence developer SE3 raised a $30 million Series A round on Nov. 13 led by Accel. The Munich-based venture creates autonomous navigation software designed for ground vehicles and airborne drones.
Deals worth a second look
- Mirendil — Kleiner Perkins and Clearlake Capital backed a massive $1B Series A, representing one of the largest early-stage rounds logged this quarter.
- UFORCE — Valor Equity Partners and the UK National Wealth Fund provided $700M in Series A backing, signaling sovereign appetite for deep balance-sheet tech.
- Fractile — Fractile drew $718.6M in Series C financing with an investor group featuring Thrive Capital, Founders Fund, Lightspeed, and Redpoint.
Deadlines next week
- Project ProtoStar · Participation in open science challenges, with access to research datasets and models for reuse. · closes Nov. 13
- Physical AI Growth Lab · Up to €350K in Microsoft Azure credits, plus NVIDIA Inception program credits and hardware discounts. · closes Nov. 13
- Localization Development Accelerator (LDA) 2027 · Nine-month organizational development program with mentorship and practical learning. · closes Nov. 15
- 2027 Data Center Innovation Cohort · Access to technical and industry expertise, investor advisory support, and strategic partnerships for climate tech startups. · closes Nov. 15
- EU Business Hub @ KIMES 2027 · Market intelligence, business coaching, B2B matchmaking, interpretation support, and financial support towards accommodation. · closes Nov. 15
What I'm watching next week
- Applications for Project ProtoStar and the Physical AI Growth Lab close on Nov. 13.
- The 2027 Data Center Innovation Cohort intake deadline arrives on Nov. 15.
- Deadlines on Nov. 15 for both the Localization Development Accelerator (LDA) 2027 and EU Business Hub @ KIMES 2027.
— Adam
Sources
- 1.OSH Rounds data — OSH Rounds
- 2.OSH Programs data — OSH Programs
