Silicon Valley is finally remembering that atoms matter as much as bits.
The Summary
- Felicis hired Graham Littlehale as partner to invest in hard tech — aerospace, defense, energy, manufacturing, robotics — from Point72 Ventures where he backed reusable rockets and military logistics software
- Felicis already holds CoreWeave (GPU infrastructure), Crusoe (AI data centers), and Skild AI (general-purpose robot brains) — this hire deepens their physical world thesis
- Hard tech is hot again: robotics and physical AI raised $16.3 billion in Q1 2026 alone, driven by labor shortages, reshoring pressure, and AI moving into machines
- The shift matters because Web4 can't run on cloud servers alone — agents need bodies, chips need power, and supply chains need to come home
The Signal
Felicis backing Shopify and Notion made sense in the Web2 era. Commerce and collaboration tools scale cheap and fast. No factories. No supply chains. Just code. But Aydin Senkut's firm is now staffing up for a different bet: the physical world is about to get the same software-style leverage that made digital products print money.
Graham Littlehale isn't coming from a consumer app fund. He was at Point72 Ventures backing Stoke Space (reusable rockets) and Rune Technologies (military logistics software). That's not Instagram for dogs. That's steel, fuel, and government contracts. The kind of work that takes years to pay off and requires understanding physics, not just product-market fit.
"The physical world is entering its own software moment."
Here's what that actually means. For 15 years, venture capital chased zero-marginal-cost businesses. Apps that could scale to a billion users without hiring a factory worker. That model breaks when AI agents need to interact with physical reality. A robot loading boxes doesn't scale like a chatbot. It needs hands, sensors, motors, power, and something to load. Crusoe's AI data centers need megawatts. CoreWeave's GPU clusters generate heat that could warm a small town. Skild AI's robot brains are useless without bodies that can open doors and pick things up.
The $16.3 billion that flowed into robotics and physical AI in Q1 2026 isn't hype money. It's infrastructure capital. Three forces are converging:
- Labor economics: Warehouse workers, truck drivers, and factory operators are expensive and scarce. Robots are getting cheaper and more capable.
- Geopolitical reshoring: Supply chains that ran through Shenzhen for 20 years are coming back. That requires automation, because American wages don't pencil otherwise.
- AI embodiment: Language models can write emails. Vision models can read X-rays. But neither can change a tire or assemble a battery pack. The next wave of AI value is in the physical world.
Felicis already owns pieces of this. CoreWeave provides the compute. Skild AI provides the software. What Littlehale brings is the defense and aerospace lens. National security money moves differently than consumer VC. It's patient, it values reliability over speed, and it's willing to fund deep R&D. Stoke Space isn't trying to get to product-market fit in 18 months. It's trying to make rockets reusable, which takes physics, not pivots.
The tell here is the shift in what "hard tech" means. Ten years ago, it meant solar panels and batteries — capital-intensive, low-margin businesses that VCs mostly avoided. Now it means robots, defense software, and space infrastructure. The margins are better. The customers (DoD, logistics companies, manufacturers) have real budgets. And the technology finally works well enough to deploy at scale.
The Implication
If you're building software that assumes abundant, cheap labor, your model is about to break. If you're building agents that only live in Slack or email, you're solving yesterday's problem. The next decade of value creation is in the physical economy. Watch where capital is hiring, not just where it's deploying checks. Felicis hiring a defense and aerospace investor is a louder signal than most funding announcements.
For founders: hard tech is no longer a hard sell. The question is whether you understand atoms as well as you understand APIs.