00:00We are safe to say that we're entering the decade of the robot.
00:03The humanoid market is really tiny right now.
00:05It's worth $2 to $3 billion, but our latest research shows it could be a $200 billion market.
00:10But I think the bigger point and the key picture here is that humanoid is just one element of a
00:15much broader physical AI ecosystem,
00:17which includes drones, it includes autonomous vehicles and other forms of advanced industrial robots.
00:23And so taking everything together, we could hit the trillion-dollar mark by 2045, and hence the investment opportunity is
00:30really sizable.
00:31But what I think is also particularly interesting is that this is a new investment opportunity that comes on top
00:38of what we've already seen in AI's first wave, which was more digital.
00:42This is the next frontier now.
00:44AI is getting physical.
00:45We have intelligence that comes to the real world, and that brings a whole new set of investment opportunities.
00:51And companies, industries, countries are very interested in this trend because the reality is that what we're facing right now
00:59is shortages,
01:01labor shortages that come as a result of aging populations, changing workers' preferences, and urbanization.
01:08And so we don't have the workers where we want them to be, and humanoids could be an answer to
01:11that picture.
01:12So that humanoid deployment, is it happening now?
01:15How do we have humanoids that can operate in factories now that have that intelligence layer that allow them to
01:21do the tasks that otherwise would have been done by humans?
01:24Is that happening?
01:24Is that a now story, or is that a next year or a few years out?
01:28I think it is a now story.
01:29The industry is just beginning to scale up.
01:31We estimate that in 2026, we're going to see about 60,000 humanoid robots deployed worldwide.
01:37And that's a significant step up compared to last year when we only saw 15,000 humanoid robots.
01:42And what's driving this acceleration is actually improvements across what we call the freebies, the bronze, the brain, and the
01:50batteries.
01:50These are the free critical bees that you need to produce a humanoid robot.
01:54And as all of these freebies get better and more efficient, the costs are coming down.
01:58We've seen a 34-fold decrease in costs over the past decade.
02:02And I think we're going to see definitely more.
02:05And we estimate that by 2035, we might see about 13 million new humanoid robots deployed annually.
02:13And that's a very large number compared to where we stand today.
02:16And we touched on the physical data and that being a constraint.
02:19Is that being solved?
02:21How much has been advanced on that front?
02:23I think that people often forget that humanoids is as much a physics problem as it is an AI problem.
02:29And the challenge that the industry faces right now is that we need to work towards general purpose intelligence
02:34so that these machines are really efficient, they're really useful in the real world.
02:38To do that, we need to take a step back and accumulate physical AI data.
02:43And the paradox right now is we just have a big shortage of physical AI data.
02:47The world has an abundance of digital data.
02:51But the challenge with that is that it's a bit tricky to translate that into real world actions that we
02:57need to train humanoids on.
02:58So, for example, YouTube has about 280,000 years of video footage.
03:03But only a tiny fraction of that is actually useful to generate the insights that are needed for these machines
03:09to operate and to navigate the real world.
03:11And what we're seeing right now is that companies are working towards solving that problem
03:16by accumulating the data needed, being through deployment in the real world.
03:20Like we send the humanoids and internships out there, we collect the data,
03:23and then the next generation of humanoids gets better.
03:27Does China end up dominating the robotics space?
03:30Do they end up leading and winning this market like they've done with EVs, like they've done with renewable energy?
03:35So, I think the short answer to that question is China has data right now.
03:39But I think that the U.S. and Europe are still very, very much in the race.
03:42The reality is that China is the world's powerhouse in terms of robotics.
03:47It installs nearly half of the industrial robots of the world, more than 90% of the humanoids.
03:51And that did not happen overnight.
03:53It's a result of a decade-long state-guided push along the entire value chain from critical minerals,
04:00energy, and key components such as batteries.
04:02But China does not hold all the cards.
04:05We've done analysis.
04:06We're using millions of patents and innovations.
04:08And what you see is, yes, it's true.
04:10China leads when it comes to kinematics.
04:12That's kind of the physical form behind the robot.
04:14But the U.S. leads when it comes to AI models and advanced chips.
04:19And now, you may ask, where does this leave Europe?
04:21Yeah.
04:21Right?
04:22And I think that Europe's opportunity lies in high-precision components and high-precision manufacturing.
04:28Because humanoids are, in many ways, like cars in miniature.
04:33They rely on the same set of very sophisticated components, in particular actuators and motion systems.
04:38And now, if you take a step back and look at where these components are produced on the world's map,
04:44well, Europe lights up.
04:45Europe produces about a third of the world's high-precision actuators.
04:49Germany alone, 10% of the world's supply.
04:52So I think that if you put these numbers together,
04:54Europe and the region is very well placed and positioned to leverage its automotive heritage,
05:00its industrial past, and to become a leader in the space.
05:03So if you ask me the question, would you reproduce the majority of that 13 million humanoids
05:08that we expect to deploy by 2035?
05:11Probably not.
05:12But does it have the potential to become a key supplier to that robotics industry?
05:16I would say absolutely not.
Comments