The hand problem is harder than the brain problem

GM. This is Milk Road Stocks, where the next AI trade might have a pair of hands.

The robotics boom is coming. We spoke to a 40-year robotics veteran to find out where the bottlenecks are and broke it down in this edition.

First, a quick detour to private markets.

On Sept 23, Milk Road's Martin sits down with Alumni Ventures to talk about blockchain and private markets. Save your free seat here.

THE NEXT AI TRADE WILL HAVE ARMS 🦾

Yesterday, Jensen Huang said physical AI could eventually be 10x bigger than digital AI.

Why?

Because 90% of the world's work still happens in the physical world.

Source: @Geiger_Capital

Jensen thinks the robotics boom could be enormous.

And that’s why we brought Scott Walter on the pod to talk about this buildout. He's worked in robotics for 40+ years, so he knows a thing or two about robots.

Here's what stood out:

If millions of humanoid robots are coming, someone has to build all the parts:

Motors.
Actuators.
Sensors.
Hands.
Magnets.
The machines that build those components.

Source: @MilkRoadStocks

Most people assume the AI brain is the bottleneck.

But Scott says the brain and the hardware can both be bottlenecks. Sometimes the AI is ready, but the hardware isn't.

And right now, the hand may be the hardest problem to solve.

There are two main camps:

Camp one: The “tendinistas”

These hands use cables that work like human tendons.

The upside: They're lightweight, flexible and very dexterous.
The downside: Tendons can break and are difficult to control precisely.

1X is one of the major humanoid companies pursuing this approach.

Camp two: Direct-drive hands

These put small motors directly inside the fingers and knuckles.

The upside: They're stronger and more durable.
The downside: They're heavier and currently less dexterous.

Nobody has figured out the perfect solution yet.

And that's what makes this interesting.

The robots getting all the attention right now (the ones doing backflips, sprinting down tracks, marching in formation) mostly don't have hands at all.

But the real robotics race may come down to something much less exciting: who can build a hand that actually works.

Tesla famously showed off its 22-degree-of-freedom hand (22-DOF) but Scott says the company hasn't publicly demonstrated meaningful dexterity from it since the original announcement.

If you want to see which companies are working on the robotics supply chain, Goldman Sachs released a report this week mapping out the major players: