SaaS is over… Why you should build a robotics company in 2026

Okay, in this video I want to do something a little bit different. I want

to convince you why you should start a robotic startup in 2026 because I genuinely think it's the best time ever to do this. And I want to use this video to a explain you why like give you a bit of context, b explain you like the opportunities are currently happening

and then three how you can start.

Quickly about me, my name is Andreas. I

invest uh mainly in hardware, robotics, automation, [music] heavily focused on Europe. Uh and we just launched our

Europe. Uh and we just launched our third fund. So we actively looking for

third fund. So we actively looking for companies to invest in. You can hear more about this on the the YouTube channel actually where we launched the launch video for the fund. So yeah,

check it out afterwards. Okay, cool. So,

so, so let's start like a little bit of context. Number one, Amazon this year

context. Number one, Amazon this year deployed their 1 millionth robot and this is just the start. You can

basically expect this to ramp up like this because they also like want to like automate more and more all of the parts of the logistics and the warehouse and so on and so on and they're just like at the start of this. Another completely

different example like war nowadays is very much dominated by drones which become more and more automated due to jamming and all this kind of stuff. This

is becoming part of robotics fully.

Speaking of classic robotics, automation manufacturing. This is now moving to

manufacturing. This is now moving to what we call like dark factories. So

factories so fully automated that you can run them essentially dark. The main

message here is quite clear like cars are going to be produced fully automated like run by robots front to back.

Speaking of cars, cars are self-driving now. And that's a normal thing that we

now. And that's a normal thing that we just have in society. We have like cars driving next to us. We just on their own. But it's not only cars. It's like

own. But it's not only cars. It's like

literally everything like everything is self-driving now due to open source and multiple progress here. But by the way, this is like a a Chinese uh delivery truck that is like super viral.

Basically a meme in China because they are relentless. they'll have like no

are relentless. they'll have like no sense of self-preservation and they will go through everything just to just to bring you your stuff. Also, humanoids on the horizon and hypernut um they're they're becoming real and we will talk

about this a little bit later. I have

like a whole thesis around them I want to share with you. So, my main message is yes, this is starting to feel like something. Yes, this is starting to feel

something. Yes, this is starting to feel like a hype and yes, this is starting to feel we're seeing like robots everywhere and all this kind of stuff. But this is just the very very beginning. This is

now a wave going upwards and my biggest recommendation is to jump on this wave right now and be in this space in the next five years because so much will change but like why is now the right moment because we had like you know

robotics manufacturing we had this like the last 10 20 years very simple reason tipping points especially in technology like what else are you going to do you know like you want to do a software startup in 2026 your customer will just

use judge DP and most likely have the majority of your features like straight out of the box sus companies used to compete compet against Gmail or like mail and spreadsheets and calendars and if you manage to get above that bar

people were willing to pay for you you know now like marketing department will just use CHDP to do whatever you offer right and if not they can use claw to build like a sus company that's like or like a product a software product that's

good enough let's say you would want you want to do like software AI you know and you're basically competing now with the biggest companies in the world Google can scrape the same data that you get you know but they have way more money to

do so And quite frankly like these hi frontier models uh companies they're not like working to to build better software. They're working to build the

software. They're working to build the industry after software and you're competing in that space. So it's really really crowded. There's a lot you know

really crowded. There's a lot you know like it's really hard to like in that space to compete. This is marketing 15 plus thousand companies in marketing like right now. Like what the hell? Like

let's take the crowd, most crowded, most established, most like automated area that um robotics has, which is warehousing. There's like 700 companies

warehousing. There's like 700 companies of them worldwide. 700, right? Let's

take humanoids, like this big overhyped industry we currently have. There's like

200 companies like worldwide. If you and your buddy start a new company, like you know, we we're very quickly talking singledigit percentage improvement worldwide. Another reason why I

worldwide. Another reason why I recommend like right now to get into uh robotics is because like just the costs are going lower and lower and lower which means you can use automation and robotics in multiple other areas and

don't think only like you know things with like robotic arms like think automation in general because also every component is getting cheaper like we mentioned self-driving before the core components to like have a self-driving car and like all these aspects are

getting cheaper and cheaper and cheaper also the suppliers and the manufacturers actually building these machines then together with you are actually now more open than ever to work with you. For

example, like in Europe, car manufacturing is like had like a little bit of a bump in the last year. So like

they are happy to now work with startups even if it's smaller batch productions to like jump on that next waves. Another

very very strong reason why now in 2026 is a very good moment to jump on this is because we are in the middle of a Will Smith moment. If you remember that meme

Smith moment. If you remember that meme with the spaghetti, it used to look horrible not like like two years ago, you know. And then by the way, the video

you know. And then by the way, the video on the right side is like also a year old. So we are now like already like

old. So we are now like already like photorealistic Will Smith spaghetti eating in video genai and that happened over like 3 4 years. We having this

moment right now in in robotics and like the easiest example is like computer vision like if you remember like hot or not hot dog or not like a few years ago we are now at the point where we can

take full videos annotate in the video in real time segment everything understand everything like you can see in videos reason in videos reason in space reason in 3D space and so on and

so on like that is solved now you know and like Sam is like you can just like download this if you want to use it the robotic brain uh um VA is the is the most common approach to this like vision language action. The way to think of

language action. The way to think of this is like kind of like the equivalent to more or less LLMs or not really but like kind of the biggest issue here is still like reliability. So like being

able to um decide what the robot should be doing and then actually reliably do it correctly even if the environment slightly changed and so on and so on and do this like hundreds of thousands of

times like at the same site so that there is no problem in manufacturing or like wherever the robot is placed so that there isn't any issues. This is

currently the moment like where people are making this starting to work. We're

still like in the early days here. So in

many areas you just don't have enough data and data is super important to a lot of these companies. Think about it like if you if you do like an um a robot like at a special purpose uh uh use case right you can't scrape that data from

the internet like if you're in an for example cleaning robots this kind of data you only get on the site from the customer at the use case and the more data you have the better the models get the better the models get the more capabilities you have and the more use

cases you can have actually offer customers and because of that they will buy more things you have more deployments because of more deployments you have more data so there's almost like an like a runaway loop that if you

manage to get into this case you have like a defensibility and this is like super interesting and there's also like a lot of startups right now only focusing on this because this is like this whole like data issue let's call it

is like a huge problem in this space right now they have a pitch of a data labeling startup twice a week right now you know like it's it's a lot but also like research labs for example here um

Imperial college in London this is called instant policy they they're working on solutions that needs they need almost like no data and can like still reliably do whole processes and

repeat tasks. Even if you mess with the

repeat tasks. Even if you mess with the robot like after one time showing. So

like here they like they move the the things around, they push the robot, they push the thing, they replace the item and so on and so on. And the actual training set you see down here on the right side, it's freaking Blender. This

is like a few renderings in Blender and it's good enough already, you know, like if if these things get into production, like if this is actually turning out from research into actual production, this will change the industry maybe and that will happen in the next years.

Okay? And that's also the reason why every VC right now is like, hey, we need a robotic strategy and so on and so on and so on. And this is not even talking about the big things like macro tailwinds and so on and so on. Like

think about it. If now I can use a robot to be more precise, more build more custom, more more beautiful things, more complex things, more detailed things, can do higher quality, can do like higher quantity, can do smaller

quantity, all of these kind of things.

You need these kind of machines as well.

Otherwise, you as a supplier will be basically out of business. Like you

cannot compete otherwise, right? Like if

the expectation of the customer is now that you do all these details at like high quantity at small quantity that everything that comes out of the of the of the factory for example has been quality assurance test like hundreds of

times you have to keep up. You have to buy these tools. You have to keep up otherwise you're out of business. This

is like a whole pressure wave on the on the market especially with China pushing automation as much as they can. Not to

speak of politics but also like in the west especially if you want to re-industrialize west like aging is a real issue here right build solutions is what I'm trying to say here right? Okay,

this is the context. This is like why now and all this kind of stuff. But

let's talk actual opportunities. I like

this one framing. It's basically robots is the next SS. You can literally look at every industry that you can think of.

Look what they currently do and figure out can I build a robot for this one specific task, one specific problem, one specific and so on and so on and so on.

And not just think like can I replace a human here. also think like if I can

human here. also think like if I can actually do run this around the clock like 24 hours if I can do this like quicker faster in smaller batches or whatever like is this opening up new opportunities that this industry

actually needs anyway another mental model I really like is like structured versus unstructured this is essentially supermarket for robots like if you want to optimize a supermarket for robots this is how it has to look and the

reason is quite simple because in a normal supermarket people run around like chaos is happening everywhere like half the shelf is like the wrong products and so on and so on you cannot do this right now yet in robotics. So a

way to think about this is like how much is this environment meant for robots or meant for humans? And now with the emergence of like the capabilities of like VAS and like all this kind of stuff, it's more and more possible to

have dynamic environments. It's more and more possible to have this like little bit more chaos and human space and entering this human space. It's more and more possible to have like not like every time exactly the same thing but slightly different, right? Like all this

kind of stuff. like structured versus unstructured and more and more of the unstructured space is opening up right now. Another thing that I always

now. Another thing that I always recommend my founders is like look at niches. There's absurd like niches like

niches. There's absurd like niches like like like jobs that should be done that can be done because you cannot like find people to do it like cannot pay people to do it or it's just like extremely

complicated to do it and like very frequently it's like some random small little task that has to be done. If you

can build a robot or like an automation for this one specific problem, this can be used globally by like hundreds if not thousands if not millions of customers.

Another mental model I like to think about opportunities and like how to think about is like vertical versus horizontal. Like in robotics they

horizontal. Like in robotics they usually call it general. A vertical

robot, a vertical automation is like a a washing machine, right? The general

robot is like the humanoid filling the washing machine. And the way to think of

washing machine. And the way to think of this is like how much is this like special purpose for like one use case and how much is it like in general we want like we don't know yet the use case and it will be multiple of them right

for special purpose like there's a form followers function you don't want the robot to to use a tool to clean a pipe you want the robot directly to clean the pipe like the the the robot should be a snake just cleaning out the pipe right

or a better example is actually forklifts like how would you design a forklift a fully automated one would you put like a humanoid in there I you know or would the forklift look look more like this. This is a a company in Munich

like this. This is a a company in Munich called Physices or like tractors like this is the uh John Deere um fully self-driving tractor. It's basically

self-driving tractor. It's basically like their old model and just like patched, you know. You should actually rethink a tractor from ground up and like make it more look like like this one is from Spain company called Walre.

Make it like hot swapper batteries, fully electric, can like like lift four tons, like run around 24 hours a day and so on and so on. This is closer to like an uh ground drone in in in in military

for example, right? Like the idea of this is that you are able to like run it overnight and like just like completely rethink how you manage a farm. The other

thing why I personally like robotics right now is small teams can move really really fast. Like this is a company

really fast. Like this is a company called Rolo. They had their first

called Rolo. They had their first prototype a year ago and was built the prototype was built like by one person with two friends of him helping and this is now already showing at CES like a few weeks ago. They always say like it's so

weeks ago. They always say like it's so slow and whatever. You can extremely fast build and like prototype already.

It's like oh no but the batch production the cost of like the the capex costs and all these kind of things. Yeah. Like

also not true. Like you cannot now like all of the suppliers want to produce with small batches and so on and so on.

It's like yeah but it's so hard and so complicated and so like blah blah blah blah. That's not true. Like all of that

blah. That's not true. Like all of that is changing right now as we speak. Let's

talk humanoids. The weird one, right?

Like is it cringe? Is it hype? Is it

not? This is by the way uh humanoid fighting. Shout out to six like the guy

fighting. Shout out to six like the guy in the middle, a friend of mine. Uh they

are doing fighting of humanoids with VR right now. This is like right now

right now. This is like right now gimmicky. Like this is like now a little

gimmicky. Like this is like now a little bit of a joke, you know, not really like they take it serious but like it's like no it's like a little bit gimmicky. This

will be very much like true in two years, three years, four years as humanoid come into space and want to like show off, right? But like I don't want to encourage you to start a fight club but if then at least don't talk

about it. the counter arguments to

about it. the counter arguments to humanoids which are very much true is a special purpose machine will always be better and there will be hundreds of vertical solutions that that you should build. I think the question I actually

build. I think the question I actually ask people is if you think about a real construction site, do you imagine we're going to build humanoids and then we're going to put the humanoid in a forklift and have it drive a forklift or are we going to just like make autonomous like technology on the construction side that

>> I mean he's right right like would you want a humanoid in the forklift or you want an automated for forklift the counter arguments now like the proarguments for humanoids is there's a certain last step problem yes the

forklift can run by itself but maybe you need to fasten afterwards the packages in the truck or maybe you need to do some other things that normally the human would do, right? And if there's no humanoid anymore and the the the

forklift is just like two bars, like that bars can't do that job and there's like this last step that has to be maybe done. And like in this mixed

done. And like in this mixed environments, humanoids are surprisingly potentially useful. Who knows? But

potentially useful. Who knows? But

that's a very very strong pro argument.

The other pro argument is what I call the iPhone case. Yes, your digital camera is great, but I have an iPhone.

Like I don't care about buying a digital camera. Yes, your special idea for like

camera. Yes, your special idea for like a mobile device for some industry is great. can you just like build an app

great. can you just like build an app because I already have the iPhone and that argumentation case for use cases might be true with humanoids like let's TBD yes your special purpose machine is

great and it's better and blah blah blah blah blah but if if people can buy this for like 5k 10k 20k you know and it's good enough why should they buy your machine that costs like 10 times more so

economy of scale comes into play and just for the context like I'm a personally a little bit like critical about humanoids but I see too way too many people especially investors does also dismiss them. So, uh there's like

hype and there's like overhype and there's like the realities behind is like this is basically my POV on it. But

there's a whole new world that's currently popping up. Like think about it like every robotics company is currently building full stack every problem if they have their own in-house solutions from data collecting to u

navigating in space to like almost everything like there's a few open source solutions but like by default you build a lot of stuff on your own. What

I'm trying to say is that there's currently no AWS for robotics. Don't

think of it like oh the AWS like no think more like what are the different problem f areas that you could just like tackle. Uh there's a good article by um

tackle. Uh there's a good article by um Anderson Horvitz that just came out where they basically argue there there isn't the concept of DevOps right now and that is like slowing down like

development cycles in robotic teams and there's a whole universe of robotics solutions that you could build without even building actual hardware. TDR what

I'm trying to say and I hope I convinced you now is this is just the beginning.

We are at the very much of the beginning and right now is a very good moment to jump on this. to help you with this. We

actually collected like a few links that we recommend you to to to check out especially completely beginner or at least the beginning stages of your journey uh on our substack and also if

you are more experienced with robotics uh do me a favor reply below like what did we forget what would you recommend and so on and so on and also send us your demos because we want to do like a

follow-up video to this and showcase what kind of uh like solutions you guys built and we want to like showcase what the community built and like so like you please send us your demos we want to like feature you. If this kind of stuff

is interesting for you, number one, please like and subscribe because it's super important that to somehow teach this algorithm that like this kind of content is actually useful like uh to to to people like you. And number two, up

here is also another video which is about Sunrise. We mentioned them before.

about Sunrise. We mentioned them before.

We u we talk with them how they build their robots for small batch production, but also like we go with them into the manufacturing side to see how they actually done. Anyway, thanks for

actually done. Anyway, thanks for watching. See you in the next video.

watching. See you in the next video.