From Physical AI discussion

More than 50% of AI startups abandoned after POC. What's going on nowadays in Silicon Valley triggers both excitement and fear. Young first-time founders show success in fundraising, building up teams quickly, creating hype around AI startups. What seems easy to imagine as the bright future of humanity with new technologies might not be so realistic when it comes to generating revenue as a long-term game.

Olivier Club continues its series of private discussions in partnership with AngelList to bring brilliant minds in one room to talk about what really matters not only for startup founders in SF, but for the future of humanity.

Physical AI and how it's different in a way that matters to your cap table.

Jason Lu puts it plainly: "It is a concentrated, deliberate effort on the part of a robotics builder to build a version of your robot that can get revenue today."

The deliberateness is the point. Most founders in frontier tech fall into what Jason calls the long-duration trap: they're so focused on the ten-year vision that they miss the customer who is sitting right in front of them, ready to pay for a version of the product that exists right now.

Jason H. Lu is the Founder and CEO of Flyby Robotics. Raised $4.6M in a pre-seed funding round. Built a team of talent from Yale and NASA. Closed a deal with Palantir. Supplies U.S. military clients. Flyby Robotics has achieved 7-figure revenue.

Revenue Over Hype

The trap most hardware founders fall into: chasing the ten-year vision while ignoring the customer in front of them right now. For Flyby, that first customer was AI developers and researchers — well-capitalized, willing to experiment, and didn't need enterprise reliability. That overlap between what the tech can do today and who urgently needs exactly that is where revenue starts.

Fernando pushes back with the longer view. Amazon lost money for years. OpenAI didn't break even until 2024. The Silicon Valley bet isn't on today's P&L. It's on whether the team can deliver real value over a decade.

Both are right. Holding that tension — revenue now, vision later — is the actual skill.

Why Physical AI Wins

Fernando on why physical AI is more durable than software AI: Pareto's Law. 20% of what we do drives 80% of the impact. The other 80% is repetitive, manual, and nobody wants to do it.

Software can't pick something up. Software can't fly a sensor into a building with no connectivity. Software can't patrol a warehouse at 2am. Physical AI can.

Fernando Lorenzo, Founder and Managing Director of Universal AI Services, an AI and Robotics marketplace that allows anyone to buy, sell, rent, repair, or design robots from scratch. Built a team of talent from Harvard and MIT. Board member of the Harvard SF alumni society.

The Breaking Point

Both founders were asked the same question: when did things actually start to scale?

Replicability.

"We built one, it works. We built five, six, seven, eight — every single one is working. That is the moment," Jason said. Before that, every customer is a risk. After it, every customer is a compounding asset.

Fernando's breaking point was quite human. He ended up in the hospital at CES — three hours of sleep, 180 beats per minute. Not recommending it. But at that conference he saw the real shape of global demand: hoteliers in Las Vegas, security firms in Manhattan, farmers in Kenya — all asking the same questions. The demand for physical automation is unglamorous and everywhere. That's what makes it durable.

How to Get the Deal When You Have Nothing

Jason's framework for landing strategic partnerships early:

"Identify the intersection of what the technology is capable of today and what an end user urgently needs." When alignment occurs and customers recognize immediate value, partnerships solidify through authentic customer research rather than shortcuts.

Fernando's point: surface area. Show up. Ask people smarter than you. Create platforms for others to share knowledge. Luck is real — but its geography is predictable. You manufacture it by putting yourself in more collisions with the right people.

To Every Founder

If you're building something hard, something physical, something that the market doesn't fully understand yet — keep going. The mountain in front of you is real, and so is the person on the other side of it who needs what you're building. Show up consistently, find the customer who can pay you today, build your first unit and then build eight more just like it, and think about the impact you want to leave behind — not just the exit. The ones who make it are rarely the smartest people in the room. They're the ones who didn't stop.

---

This conversation was hosted by Olivier Club in partnership with AngelList at their San Francisco office. Olivier Club runs private discussions for frontier founders and investors.