Right now, somewhere, a company is making a move in plain sight.
The street reads the headline, "Autonomy AI sees the signal."
We deliver it before the market catches up.
Autonomy AI, your models, our intelligence.
Request a private strategic briefing at AUTNMY.AI.
The autonomy economy is real.
Commercial Robotaxi operations.
Humanless freight runs.
AV infrastructure investment measured in billions.
The question is no longer whether autonomous mobility scales.
It's who advises the company is building it.
KPMG works across the full autonomy ecosystem.
Operators, OEMs, suppliers, insurers, and investors.
At the moment, it matters most.
When the road ahead is uncertain, we've already been there.
KPMG, accelerating what's next in mobility.
I'm out in the field doing field work right around in Ohio.
My phone buzzes its wall.
He's like, did you see it?
Did you see it?
I said, see what?
I'm in the Ohio.
I see Ohio.
He's like, no, the Tesla's doing stunts.
I said, oh, my goodness, gracious.
Perhaps the Tesla stunts are the show, or the showcase,
or Elon Barnum is a reference to the great PT Barn
and Racing Bridge for Connecticut, one town away from my hometown.
We can talk about the signals next week.
But we've got a great show of signals here.
As Rob, as I've been busy on the road, you and Omega,
you're becoming best friends now.
You're keeping track of the signals.
And this week, the signals are as follows.
Too far for the school bell.
Federal regulators put Tesla's robot taxes on probation.
Uh-oh.
And then say hello to yo-yo.
Expand moves from EV Maker to Robo Taxi Provider.
And signal three.
A autonomy on aisle four.
That's right.
We're not calling Jim Carey.
We're calling autonomy.
Supermarket humanoids hit the night shift in Denmark.
Lots of signals to break down this week.
And before we do is this, every week,
a huge thank you to KPMC for being a wonderful sponsor.
Where do we want to start?
Rob, I think I know we have to start with signal one.
But that little show we saw was very, very interesting.
I don't quite know what to make of the show.
Flying cars will look like a giant hot wheel ramp.
And, you know, special effects flying everywhere.
It's a spectacle, at least, right?
It's never boring.
So you gotta give them credit.
It's, you know, come see the egress type of stuff.
You never know what's gonna show up.
And thank you, Grace, for being on the road
and being in our road warrior and trying everything out.
It's great to get calls from you.
And as you're hopping from one ride to another to the next,
it's good stuff.
I'm getting up to the minute updates,
even faster than Omega's getting them, which is impressive,
because Omega gets this stuff fast.
And what it found for us today is signal one, right?
Earlier this week, Nets have launched a safety inquiry
into Teslas 2026 Robotaxies,
both the Model Y and the CyberCab.
After receiving numerous reports and submissions
of video evidence of the vehicles speeding
in active school zones.
As you and I have always talked about it,
there are a couple special categories of folks
that always draw a regulator's attention.
If you put cops or you put kids in danger,
that is gonna draw attention.
And so Tesla has drawn itself some attention.
The potential of investigation.
And it's potential because this is a public submission
to Nitsa and Nitsa is opening an inquiry based upon
the evidence submitted.
Doesn't mean that they're gonna take any action,
but it's something that they have to take seriously
and they are gonna take seriously.
And the next 90s or days come back
with the decision whether or not this leads to a recall
or something even grander.
But the potential investigation covers an estimate
of 1,000 autonomous vehicles in the Tesla fleet,
which are currently operating across X's,
where you are, Florida, where you came from,
and the Bay Area, where I am.
Imagine that.
No incidents though to be clear.
No incidents of traffic violations or crashes
or injuries or fatalities have been recorded
in these video evidences.
So this is just, hey, I'm catching this vehicle speeding,
not that there was a ticket given,
but we think this vehicle is going too fast in school zones.
All of this adds to the mounting regulatory pressure
that the Tesla is facing, right?
It's got the petition, not the petition,
but it's audit query going for it's steering wheelless,
pedalless vehicles in the cybercab to see if that certification
meets the standards of the law.
It's also recently had to rename its full-site driving
to Tesla's sister driving in Europe.
And now we have this potential inquiry as well.
For those who follow the indexes closely,
the road to autonomy, Robo Taxi Index,
which I do every day, twice a day, as it updates,
Tesla currently ranks sixth.
It's fallen a little bit,
and we'll talk about who's overtaken them at number five
in Signal Two, which is X-Ping,
four of the top five spots in the road taxi index
are now held by Chinese companies.
The Tesla currently ranks sixth
on the road to autonomy, Robo Taxi Index.
In that index, it's got a manufacturing score of 95,
one of the highest of all the companies.
This just goes to show, though,
production strength is not the same as showing
that the cars can and will-be traffic laws on their own.
Traffic laws are important.
You call it kids, I'll call it children,
because then you get CNC, cops and children.
Even if you've got a CNC's music factory,
so we can get a little music reference here early in the bod.
I own a Tesla, as this audience knows,
and I use FST when I drive my daughter to school,
and when it goes from 30 to 15 in the school zone,
I take over it manually because it doesn't stop,
and it does not slow down.
So that's my personal experience.
Is this Robo Taxi? No.
I've not experienced Robo Taxi in the school zone.
I've experienced FST in the school zone,
and it doesn't work.
I've seen just being a parent here for a moment.
It gets out of the car pool lane,
and it tries to go around,
and I'm like, "No, no, you can't be that person.
That's not good. I'll get my hand slapped."
Look at my hand slapped.
We can't do that.
So there is a lot still for this technology to work,
and I was talking to Spencer on my way back from dinner,
and we had this whole conversation about routing.
And so that's the last word.
There's a lot of things to work through,
but Nissan is clearly taking a proactive approach
to safety.
The administrator Morrison is doing a fantastic job
of promoting safety and putting safety first with all this.
So there's a lot of moving parts here with Tesla.
Why does this matter, this specific inquiry,
is there going through the FMVSS approval right now?
It matters in particular,
because there is a lot of scrutiny on Tesla's self-certification process,
and what I'm saying, and this is not what NITSA is saying,
but if you look at what happened with the ZUX petition,
they'll grant it a different process, right?
ZUX went through that petition process,
it got its petition granted,
and then separately, as part of the process,
in what was fairly unique for NITSA,
it's set an operational letter saying, "Yes,
we've granted your petition,
but here are some operating conditions
that you have to work under,
and there are certain milestones that if you meet,
we can expand these operating conditions,
which included geographic locations for the vehicles,
it included certain speed limitations for the vehicles,
a certain fleet size for the vehicles.
If I'm Tesla, I'm concerned,
because one, it's never good to be,
it has to be explain yourself
why potentially you're putting children in danger.
And two, if the certification perhaps can be tied
to a similar operational letter saying,
"Hey, we're going to perhaps bar you
from operating in school zones
until we have further proof
that you can do this correctly."
Now, you have to remember, right,
these Tesla operates on a mapless system,
it doesn't pre-map its operating areas,
so there are not pre-loaded maps
that have perhaps speed limit signs already
in the map on the vehicle.
And some of those signs, right,
they can be conditional,
they're conditional where I live,
which is like if children are present,
it's 15 miles an hour,
if children are present,
it's up to 25 miles an hour,
or sometimes they come with between 3 and 7 pm,
or 7 and 12.
And you have to wonder if,
without the pre-loaded map
that already has this information
for the vehicle to retrieve,
as it's traveling in that area,
if they're having certain issues
with either perceiving or understanding
what it's perceiving as it goes
by a speed limit sign or something like that.
So I would say the big concern here
is that one,
on the investigation itself,
or the inquiry itself,
that it leads to a recall of some kind,
saying, hey, Tesla,
we're not satisfied that you're software operator.
rates up to snuff, fix it, or don't operate in these areas,
or, too, somehow, this gets tied into a conditional operation
letter when the self-certification issue is resolved.
Because these timelines are starting to line up
all towards the end of the year beginning of next year.
And we talked about this a couple of weeks ago
with the update for the NHTSA rules as well.
Things are really going to center around
that December/January period for Tesla.
And so if I were Tesla, I'd be concerned
that because of the unique nature of the violations
accused of here, particularly in danger in children,
speeding in schools' zones when children are present in them.
It's not just speeding when they're absence.
That's speeding at 2 a.m. in the morning.
It's speeding at 8 a.m. drop-off time
that you could potentially see either a recall
or a conditional letter saying Tesla.
Until we've approved your self-certification,
but we're going to put some conditions on the way you operate
because we're concerned about the way you operate in schools' zones.
Could we see potentially or hypothetically
these conditions applied in one market and not another market?
Or do you think this could be blanket across all 50 states?
Given that there seems to be nothing unique in terms
of the evidence and the request that was sent in,
nothing particularly unique to Texas,
I would be worried that it would apply everywhere
if the federal government, if it were to put a condition on.
Now, Texas has taken an approach
where when it comes to these types of safety issues
and safety performance around the software
that Tesla is using or anyone else is using,
they have taken a step back and said that squarely lands
with the federal government.
So we're not expect Texas to take any individual action here.
Now, if an incident should occur, if somebody should be injured
or a really close near miss, you could see Texas step up,
especially if the federal government is not moving quickly
and they feel like it's an emergency situation.
But generally, Texas has said when it comes to the safety
of the vehicle and the safety performance of the AES,
we're going to leave that to NITSA.
Now, if and when California that cybercabs are allowed
or permitted to go without a driver in them,
you could see California perhaps take
a more aggressive approach than Texas, which I'm not sure anybody
on this call would be surprised to think
that California would take a more aggressive regulatory approach.
So we could see it differ by state action
if the federal government is not moving quickly enough
according to certain states.
But given that Tesla has not yet received permission
to go driverless in California, I think we're not going
to see Texas or Florida, both states
where Tesla is currently operating driverless,
take any action and they're going to defer any action
to the federal government.
Is there precedent or history has this happened before
to another company?
And if so, what was that outcome?
And could that outcome potentially repeat itself
for Tesla if there's a proven potential fix that did occur?
When it comes to active school zones,
there have been a few inquiries about performance
in school zones, both questions were raised about WAMO's
performance in school zones, particularly
as it was called on camera multiple times,
potentially not eating the stop signs that pop out
from school buses or the stop arms
that sometimes pop out from school buses.
That required a recall if I remember correctly
and WAMO did take some actions to update their software
for better performance around school buses.
It did not require WAMO to stop service, just like here.
Tesla is continuing to perform its services in Texas and Florida.
It's actually growing the number of vehicles in Texas.
So that kind of justifies what I was saying
about Texas taking a hands-off approach with this.
It's still allowing more cybercabs to be added
to the Texas DMV registry.
I believe a couple of years ago,
it wasn't with a Robo Taxi per se,
but it was with an autonomous vehicle operating
in Jacksonville, Florida that had an issue with transporting
as like a mini bus that kids used.
I'm trying to remember the company.
You might remember it.
Do you remember the name of that company?
So I was there.
Oh, wow, you were there.
Yeah, look at this.
I was there.
I was not involved with the operation.
I was a guest of the owner.
There was Mr. St. Kitson, St. Kitson is the developer.
Kitson and partners developed Babcock Ranch.
Punta Gorda, Florida.
And the operator was transdev.
And it was an easy miles shuttle.
And if you've ever spent time in Punta Gorda
in Babcock Ranch, it's an eco-friendly community.
I believe the top speed limit is 10 to 15 miles an hour
and wanted to move kids in the schools
to the homes in this master plan community.
Little fun fact, the school was actually
designed by Governor Jeb Bush.
And so very focused on academics.
And they wanted to bring the future to these children
and put a lot of really great solar solutions,
a lot of great things around the community
for children to interact with.
And then as soon as they wrapped it in a school bus wrapping,
they got a letter from Nitsa.
And I think the administrator of that time
was Dr. Mark Rosekind.
You got to fact check me on that.
But I was there the day the shuttle got shut down.
And it sat in a parking lot.
And all it was was wrapped in for the record
because we might get somebody to cry.
And oh, safety, there was a safety operator on the shuttle.
And I think it was going 10 miles an hour
in this master plan community trying
to expose children to this technology.
And I think my daughter might have been six or seven
of the time she wrote in the shuttle,
not the one that was wrapped like a school bus,
but she wrote in the shuttle.
And she was perfectly fine.
And she was happy.
So it's the wrapping.
And you want to bring that full circle to today, Rob?
I visited with a company because we visit
a lot of companies and they manufacture school buses.
And this individual that runs the plants
said to me, what if we made this autonomous?
And then I told that story today.
So very good timing to bring autonomy signals
to the field work that we're doing.
So well done.
And since done a great job building Babcock Ranch
in Putecorda.
Yeah, Babcock Ranch is great, it's great.
But yeah, it just goes to show there's
a lot of evidence that when it comes to children,
they're in like a special category, right?
My legal background breaks me to there
are different kind of constitutional categories
that it's called scrutiny.
What kind of scrutiny is brought to a constitutional law case,
for certain protected groups.
You have strict scrutiny.
Then you have this kind of nebulous area called
intermediate scrutiny that applies to certain cases.
And then you have kind of your regular baseline scrutiny
that goes to most cases that involve issues of constitutional law.
So you can put cops and kids, or as you say cops and children,
because you like the illiteration better,
cops and children into the strict scrutiny standard
for federal regulators.
And so your story about the school boss, that was just wrapping.
But anything that potentially raises the question of,
is there potential danger?
And it's very much potential danger here, right?
Again, we should remind folks that the evidence
that was sent to NITSA, that for them to open up this inquiry,
did not involve any injuries or damage or close calls
to hitting children.
But just the potential that it's there
raises that scrutiny level to that strict scrutiny level for NITSA.
That's the key we're here, potential.
Nothing has occurred, which is a very good thing.
But it is potential, which brings us to Omega's risk.
Omega's then covers some really good risks here, up.
Risk one.
School zone, temperamental and contextual recognition issues
produce a defect determination and a recall.
That is a good risk, because in my humble opinion,
there's two kinds of recalls.
There's a software recall and a hardware recall.
But yet, they share the common denominator.
They're called recall.
And then our friends at the meeting,
oh, they love to run with things.
Recall, recall.
No, it's not total recall with our Schwarzenegger, OK?
It's a software recall.
So just stating that there for the record.
For sure, right?
And this risk goes to exactly what I was talking about before.
There's the temporal potential issue here, right?
It doesn't understand the times at which the speed limit
supplies.
And then there's the contextual issue here as well,
which means it doesn't understand that what the sign says,
if it's like one of those signs that has just specific hours
or that will only when children are present line to it.
And so these are potential perception issues, right?
As well as planning issues, things of that nature.
And so this could lead to a defect determination and a recall.
And again, that's not necessarily a death knell for Tesla,
But it does continue to pile on more and more
concern about Tesla from the federal perspective as it's looking to finish its
audit inquiry into the Tesla self-certification as it's exploring how it's
updating the rules around steering wheelless and pedalless vehicles that clearly
affect Tesla and then obviously trying to determine the safety performance of
the vehicle ADS separate and apart from the certification question and the
safety issues that come with certifying a vehicle regardless if it's that ADS
driven vehicle or non-ADS driven vehicle so it just it just it piles up I'm
sure who's ever working the Tesla department at NHTSA you know is staying
up late but at least they have some job security they're not going anywhere
for for the next couple of months now and what we saw tonight and you know
there's somebody in NHTSA this I'm used the old term from Phil Rosuto the
scooter holy cannolis or holy cow is somebody sitting there waiting this thing's
not going on a public road you and I both know somebody probably said those
exact worries this is not going on a public road so they have the work cut out
for them and let's get to the risk too that Omega has uncovered here additional
adverse crash publicity prompts Texas or federal action limiting the
geofence or authorizations that's true that is true and as we see here I'm
currently in Houston right now just to put this in perspective and this is
Houstonian traffic so I want to put that in perspective from where Tesla is
operating in Caddy to where it is downtown 55 minute drive but 55 minutes and
is operating in a very small area is downtown Houston necessarily the busy
smoke populous now when the Astros plays a lot busier but still it's
operating if you want to call the suburb so this is another really good risk
Omega's pointed out it is and it makes me think back to my time of crews right we
had a series of incidences right with with cars that stalled or outages that
caused the vehicles to stop in the middle of the road for a period of time you
know things of that nature that you know these little things as they start to
pile up what it does right is when the big thing happens any big thing will
happen it will happen to all of these players eventually because it's just
that's just maths the law of large numbers right the the uniqueness of the
things we see on the road something bad will happen again and if you've had
all these kind of little things that have really chipped away at the
regulator trust and the regulator patience there's less patience for you when
you trip into that big issue and I think Waymo is the counter example I think
Waymo did a really good job and still does a really good job of addressing
issues very quickly of getting ahead of them of even kind of slowing their roll
outs down to ensure these issues don't occur I would say you know I love crews
very loyal crews in GM we we were moving at a faster pace but it created these
kind of small cuts at the regulators trust and ultimately when the big
issue comes the regulators just out of patience with you and so Tesla may
seemingly be headed in that direction that's not a direction that that can't
it's not a one-way door as they say here in Silicon Valley they can reverse
it but I think they're they're going to need to address these issues kind of
forthrightly and as quickly as possible particularly around the issue of safety
around children because I think regardless of where you stand on autonomous
vehicles if you're pro autonomy you're concerned about it you you want it in
your city you don't want it in your city there's zero tolerance there's strict
scrutiny applied across the board when it comes to performance around children
and so you know this risk is spot on I call it a death by a thousand cuts right
you continue to have these issues it it erodes that trust and that trust is
really critical in that moment of need when it went a much bigger issue occurs
and so I say kudos to wemo I think you know Tesla I think it's like it's not a
one-way door they're not on that's kind of like inevitable downhill slide
towards regulators you know not trusting them but I do think it's this type of
issue that you really want to see them jump on top of and get ahead of and
really address the issue proactively you have to address it proactively and you
have to manage risks I mean risk management risk of payment whatever term you
want to use they're going to have to do it and building trust takes time takes
decades and it takes a minute to crush decades to build so we'll be watching it
and we wish all parties involved success and you know for the children out
there and autonomy for the children like there's that you know there's that
great ex account who tang is for the children all this go on here and say
autonomy for the children because this happens a child runs after a ball in the
street the robot actually is going to stop when there's a 50-50 chance of
human stock and stop and unfortunately we see those statistics so
autonomy is for the children that's my PSA today let's bring us to Omega's
take Tesla now faces two separate gates to scaling whether the vehicle
design is lawful and whether the automated driving behavior is lawful
having permission to operate in a state does not settle federal deficit
questions the developments to watch are the petition decision any recall
filing and any dated restrictions on unsupervised FSD or cyber cab
operations those would change the success trajectory of cyber cab
operations more than any other city launch what that is a good take and it is
concise and Omega's been in her week she has right Omega is for the children
since we're we're talking about who's for the children Omega's for the children
and for all the adults out there if you want or need to talk to us or use
Omega or have access to Omega reach out to us at you know Rob and Grayson you
can reach us at Alpha at AUT and NY dot AI or reach out on LinkedIn or wherever
you find us or we're ready and willing to engage and you know support you as
well as support the children thank you and I will say our data is ground-tooth
verified and we've uncovered some very interesting things which all T's here
for our subscribers will get access to it and just Omega the number is
publicly reported by the Texas DMV or not the actual number of vehicles on
the roads so tease that for you we got the real numbers they're in Omega and
their first subscribers Rob signal to X-Pan yo yo I remember yo yo we were kids
I can never get that stinkin thing to go up and down up and down by yet X-Pan
they've got the yo yo robo taxi brand what do we know about this signal yeah I
was terrible I always try to I want to walk the dog or whatever that trick was
and I think I smack myself in the forehead more times than than was
necessary and I may show in some of these episodes you know I have been
spacked in the head a few times so be patient with the audience but here X-Pan
is officially named its robo taxi service X-Pan yo yo and opened online
registration to the public in China transitioning the service from technical
testing to commercial operation X-Pan yo yo is China's first fully in-house
developed mass produced robo taxi it relies on four in-house touring AI chips
delivering 3,000 tops of compute and the vision language action 2.0 vision model
operating without HD maps are lighter sound familiar right X-Pan is very much
taking a Tesla-like route vehicles rode off production lines in Guangzhou in May
2026 began automated pickup and drop off in July and secured driverless road
test qualifications in late August or early September and initially they have
started to run paid internal trials before opening registration to external
users so about the vehicle itself go online take a look at it features
customizable cabin pre-settings including fragrance and temperature I have
enough seen one that allows you to set the fragrance in the vehicle I thought
that was pretty cool zero gravity massage seating and a second row entertainment
screen I mean you have to go online check it out Google X-Pan yo yo these
screens are huge huge I mean it is like going to a mini iMacs as you sit in a
robo taxi and so it's fully integrated with rich content services no X-Pan
selected the short name yo yo for global brand recognition and plans to
generate revenue as a technology provider via hardware sales software
services and commission sharing while partnering with local fleet operators
globally and we'll speak a little bit more about you know what this means but
the name yo yo was chosen because it appeals they believe to foreign markets
right? It is not an aim chosen because it is an aim popular
in China it is a name chosen for Europe in mind and for the Middle East in
mind and for the US and North America in mind. So I think that's very interesting.
We're going to hear a lot more about X-Pang in the coming year I believe.
Now X-Pang ranks fifth just ahead of Tesla who we just talked about on the road
to a Tommy Robotaxi index. They also have a manufacturing index in the 90s.
And so they are they're moving up the ranks fast and the yo-yo like I said
beautiful vehicle so check it out online I can't wait to get more.
Yeah no they're doing beautiful things and our friend Spencer and David
Moss were recently in X-Pang vehicle so we're releasing field reports on
those and what they experienced there over in Paris but
it's very clear and you and I have talked about this for a long time
offline and online here in the show China's autonomous belt and road
initiative is real and you're seeing it play out in Europe and now you're
seeing the Chinese EV is going to Canada and it's my
prediction here I'll make a prediction here it's just a matter of time till
Chinese Robotaxi companies start operating in Canada
and it is globally globally expanding. So what is this signal
in the PR release here for X-Pang because you and I we put on our inspector hats
I don't travel with costumes so I don't have the inspector hat I have the field
inspector hat that comes out in the and the field
but I don't have the general inspector hat on right now
but press releases tend to get bolstered on headlines so what is the real
signal in these announcements as I was paying has made?
Yeah there are two things that really drew by attention out of this press release
from X-Pang. The first is that it is now qualified
to do testing without a safety operator on certain roads in China
so we always talk about how the key thing in determining progress
is do you have a safety operator in the vehicle or not? Are you truly
driverless? And X-Pang has now qualified to go driverless
on certain roads and so that's a big step for them and I think because of that
once they go driverless and you see some of those results made public
you're going to see them rise in our Robotaxi index
I'm pretty sure coming up and the second and I think this is probably
more valuable for our audience. The second is the business model and how they
spoke about how they're going to make revenue.
X-Pang says it will not operate the fleets. It won't be a fleet operator.
It will act as a technology provider and ecosystem enabler
burning from hardware sales services and commission sharing
and a handing offline operations to global partners
right and so that's really interesting and I think it's really interesting
because you mentioned the spread of Chinese vehicles
if you look at the U.S. right now right we have Chinese vehicles in the U.S.
We don't have Chinese fleet operators. We have the O.I.
Nemo is running a Chinese vehicle in the U.S. So is it too long before we see
an X-Pang maybe not run by X-Pang as a fleet operator
but maybe you have a different fleet operator you could see them partnering
and perhaps with an Uber or a Lyft or somebody else
right who wants to be the fleet operator with their vehicle
and kind of take that approach that Gile did with getting its vehicles into
the United States in the Waymo operations and so I think these are the two
big things that I walk away with which is the going driverless and their
business model is a little bit different than say a buy-to-apalgo
or a pony AI some of the bigger competitors or we ride some of the bigger
competitors that we see in China. It's very interesting
it seems that X-Pang and this is my personal opinion is studying the market
they're taking the best of Tesla with no LiDAR no maps
and then they're taking the best of Waymo with having third parties
managed to fleet and putting it together and it's called
X-Pang Yo-Yo because that's really what it is and they're putting all the
pieces together and they're studying the market
and if they are successful in this it's a buckle up and watch out moment.
Great it's it's going to be really interesting and what you're starting to see is
where can folks want to add value to the chain but two where can they
offload cost from the chain and we're seeing people take different
approaches and I think this is what makes the robotaxi
market so interesting who wants to own the vehicles who doesn't want to own
the vehicles who wants to operate the vehicles who doesn't want to operate
the vehicles who wants to be the least holder on depot who doesn't want to be the
least holder on depot and what you're seeing is different approaches
taken by some of the bigger players and it'll be
fascinating because it's unclear we have a sense
and Omega certainly has a sense to help us inform it and we're happy to share
that with everybody so I'm clear how this all plays out
but in the meantime there's opportunity everywhere for everybody in the chain
to find value and I think that's what's really interesting
about what X-Ping is doing here is they think they've found value
in a different spot along the chain and perhaps the Tesla or OEMO
or even a buy-do from their own home country so I think that's what makes it
great we're going to see different approaches
different business models different opportunities for people from outside the
autonomy economy or the traditional kind of
operators as to state but inside the autonomy economy to find
opportunities to engage and to make money. There's a lot of opportunities there's
a lot of competition you use you said earlier you have
we ride you have buy-do Apollo go you have pony AI now you have X-Ping all
all competing on the mainland in my opinion soon to compete in Hong Kong
and then the next competition market is going to be Europe it's going to be
very interesting to see how that resonates on the European continent
and you said the most important thing earlier brand
whose brand is going to resonate most on the European continent
I think that's something to really truly pay attention to
in my opinion for the Chinese robot exercises it's not going to come out of
technology it's going to come down to brand and brand awareness and that's
something to pay attention to in really well. I think you're right I think the
other vector that I look at is just what is the service like what what is the
expectation of a robot taxi service right I think
it's unclear to a lot of consumers that haven't engaged with them
is this a premium product this is a luxury product this is a regular everyday
product I think it's you know when I talk to my family and sure this is anecdotal
but they don't live in Silicon Valley they've never engaged with a autonomous
vehicle they're not quite sure what's make of it is it is it for just the rich
and the famous is it for everybody and I think
it'll be interesting how the brands start to identify
with the consumer at that price point and you know this X-Pang looks pretty
luxurious you know it it could be a luxury vehicle but the yo-yo I think is meant to
signal that it is a kind of fun open type vehicle right that it's not just a
luxury item you don't call something a yo-yo knowing what the that means in
different languages if you're if you're looking for
a premium type of brand so it's it's it's really interesting
we should also mention just real quickly here um as to get you know back to more
than nuts and bolts instead of just branding um X-Pang does have some
exposure to the USAI training chip restrictions
right and we didn't mention that it has its Turing chips but that's for the
vehicle edge in inference right that's protected by their Turing proprietary
Turing silicon but it's cloud training right that remains exposed
because they they are they're cloud training relies on high performance
Nvidia anchored in kuda optimized compute so why I think it's
it's pretty unique they're Turing proprietary silicon but that's for
inference in edge compute is not for cloud
for the cloud training which is where their foundation model is is going for
for high performance in pre-training and post-training things of that
nature so a lot of the work is being done in the cloud
and so um I think anytime we talk about a Chinese company we have to
understand and where has its advantages and then where it might be hamster on
a little bit and so in this sense um X-Pang as I said it's it's not fully
insulated from the USAI training training chip restrictions
and so that's something to watch as well um because they may have hit a certain
level where they feel that their cloud compute can now move to the edge
and the edge compute is strong enough and the inference strong enough
operate driverlessly um but that's only for certain roads
and they haven't done it yet and so uh that that signals to me a little bit
that while they're getting
ready for commercialization with this brand chains that they're not quite their technologically,
they're close, but they're not quite there.
They're trying it's the Wu-Tang clan saying, "I'm on the Wu thing now," is like, "Dala
Dala Billio," and for X-Pang, it's like, "Dala Dala Chips show, Dala Dala Chips show,"
or if you want to get really technical, "Dala Dala GPU show," and that's truly what's
needed.
They need gents and to get gifs, obviously, you have export restrictions, which are going
to slow this down into Huawei and others in the Chinese market catch up.
We don't know what the black market is, I'm not gonna comment on the black market there.
Do we have any inclination or any public announcement of when X-Pang potentially could
look to export the robot technology to the global market from the mainland?
I think, yes, Omega has surfaced from its vast array of intelligent sources, including
other Wu-Tang members, Rizza, and Method Man, that may begin overseas robot taxi trials
in 2027.
So, I think that's the date to look for is sometime in the first half of 2027 for an overseas
trial.
Omega is the genius digit, because I was listening to liquid swords on my drive from San Antonio
to Houston.
But, you know, when you get in a sword fight, there's risks and sword fights and challenge
shadow boxing.
There is risk, which brings us to Omega's Rizza Wu-Tang style.
Driverless daily robot taxi operations without safety officers slip beyond early 2027.
Okay.
That's a possibility.
That is.
And that's what I mean.
They are announcing that commercial operations are coming, right, hence the branding of the
yo-yo and all of the campaign around it, through our attention, through Omega's attention,
and rightfully so.
It's a big step for one of the major players in the robot taxi race.
However, what we didn't see is an actual driverless operation.
And so they're still working towards that.
They have the authority to do that in certain roads.
They haven't done it yet.
And this risk is all about, okay, what's your money where your mouth is now?
Are you going to meet those expectations?
It's one thing to say you're ready to go driverless.
It's another thing to actually go driverless.
And I have that experience in multiple AV operators that I worked with.
It's a big deal to take that final step.
You get closer, you get closer, you get closer, you feel the excitement, you feel the
worry, and then you actually have to commit to it, which is a big thing.
It's a big step.
And so we'll see if whether internally they are delayed for some reason, because they
don't feel properly ready yet, and that's okay, right, we want people to be safe and feel
like they're ready to put their physical AI on the road in a safe manner.
Or it could just be that they haven't solved everything that they feel like they need to
solve, whether that's an environmental condition, or they just want to make sure that as they
go overseas, the vehicle can make that transition to maybe a different hand drive or to different
weather conditions or different road grades, things of that nature.
We'll watch and see how it evolves, which brings us to risk 2 here.
Tightening U.S. restrictions on advanced computing chips to late X-Pang and VL, a 2.0 model
training.
That's real.
That X-Port risk is very, very big, especially when you see the massive sell-off NAI chips
in the whole sector here in the U.S.
Yeah, it's interesting, so I've been doing a lot of talking to folks that are using different
types of models, right?
This is a VLA, Vision Language Action model, which is a little bit different than some of
the other models out there, right?
You have your large language models, you have your world models.
I think what will be interesting is that the U.S. chip restrictions are geared towards
like the language models in terms of parameter counts and the power of the model capacity,
which now runs in the trillions of parameters, and so those really advanced chips are needed
to run those super huge parameter models.
What we're seeing from Vision Language Action models, of which China is a leader in this,
right?
They consume and they have so much access to video data, right?
That's not only data from the digital hydromes or from TikTok or from asking factory workers
to wear phones on their head and video, what they're doing, it's just been part of their
culture to be a much more head on ingestion of video data.
And as those models start to scale, meaning you put more data in and you add more compute
to it, the scaling laws for VLA models are potentially steeper than they are, meaning
like they will achieve higher outputs in capacity and achievement at lower parameters than
what the large language models are doing.
So they'll be interesting to see as X paying in others really start to turn up the volume
on their compute.
So to speak and get those parameters up, if they even need to compute in the chips that
are under restriction to get a really efficient output from their model, meaning they could
potentially use chips that are not under restrictions to achieve the advances they want
in their VLA model, which could then put the vehicle on the road with a higher efficiency
and capacity to do things as being asked to do and ultimately with powerful inference
chips on the computer itself.
So there's going to be really interesting to see how that all plays out.
It's potentially an issue for X paying, depending on how VLA models go up the scaling
laws with the number of parameters and needs and the compute needs to achieve those, but
it's possible that they don't run into it as well and that will be a really interesting
thing.
That will be a moment I think that kind of really resonates in the US when the US government
makes up and says, "Hey, we put all of our tariffs on preventing China from running
ahead on certain models and you're going to be like, "Yeah, that's great, you did it for
large language models, but there are all these other models out there as well."
And perhaps they can achieve the types of gains in physical AI at lower parameter accounts
with less powerful chips than we originally thought.
In you and I know, the world knows China is not going to sit around and wait for a restriction
to be lifted.
They're going to try and figure out another way to do it.
Point blank.
I said that quite a part out loud.
And as Rob Elligan said, there's a lot of moving parts, so we're going to have to watch
and see how it evolves and we've all known that China is not going to stop.
And they're going to continue to try and build a product to move ahead in which way that
they see a competitive advantage.
And X-Pang is surfing that wave in moving there.
We're looking forward to testing their technology when it scales in European content, which
brings us to Omega's take.
X-Pang's launched the X-Pang Yo-Yo.
Every time I hear the Yo-Yo Rob, I'm going from hardcore hip-hop to Yo-Yo Ma, the great
celloist.
And that's why I think of Yo-Yo, with Yo-Yo Ma here.
Rob Otaku, Service Mars, a strategic shift from EV manufacturing to becoming an ecosystem
technology provider that partners with global fleet operators rather than managing fleets
directly.
By leveraging proprietary turning, Silicon and the VLA 2.0 vision only model without
lidar or HD maps, X-Pang is pursuing a Tesla-style approach to autonomy.
Key constraints include heavy reliance on Nvidia, anchored cloud training infrastructure,
and the execution hurdle of achieving driverless daily operations by early 2027.
That is a very good, all to use in term, compact summary of everything that we packed into
this segment.
It's great.
I'm going to need to get like an agent with Omega's brain to summarize everything
I have to say in such a compact and wisdom-filled manner, because I think everybody who talks
to me would benefit from it.
My conversations would be a lot shorter, and maybe I'd miss out on a few things like
all the Wu-Tang references, but we'll leave that to the listeners to decide if they'll
miss that in the future.
I think they will, because I love the banter back and forth personally.
I think it adds character to this show.
If you compact it, I read the great book, Ari Manu, Wolf's book, which is phenomenal,
by the way.
He's the founder of Endeavor, or co-founder of his that term, Roll the Calls, and he
just got the call off and dropped four letter words, which were children's rated program.
I'm not going to drop the words, but I love the banter.
So the banter is going to stay there. I have one vote. You have one vote for my votes to get the banter stay if you don't agree
Okay, well then we have a tie all little stickers. We have an audience of everything. Have a vote
But hopefully the audience agrees that the banter stays to as really great in color. It's a signal three
This one's awesome. This one's going to get banter
Now I'll hire four humanoid in Denmark. This is cool. Lights out humanoid house. Okay. There we go banter time
Why don't we know how it humanoids in the grocery stores in Denmark after hours. This is great
That I love these type of stories
Uh a photo text supermarket near Copenhagen, Denmark is testing humanoid robots to handle repetitive time-consuming retail tasks like picking up items and stocking shelves
The primary focus of the test is to assist staff during night shifts and perform tasks away from customer areas
The robots currently work at a slower pace than human employees and still require close human supervision
Uh, and under Danish regulations the humanoid robots are not permitted to operate autonomously around shoppers during regular store hours
But I mean this this is a credible retail automation experiment, right? This is great
Uh, the real signal is that a major retailer has given an actual workplace to test data collection and integration
Um, and so I really think this is the beginning of
Where we can see some utility at a humanoid robots. I mean stocking shelves after hours
Um, and checking barcodes is a great use case for them
And yes here they're slower than humans right now and there's a lot of kind of one-to-one
supervision going on but this is a real use case. So this could provide actual utility to people and to help leave
You know, maybe even lower grocery prices as the cost of operation
Look, I by I guess first job when I was 13 years old was at an AMP in Brewster, New York
And I they did not allow me to work to cash register, which is a good thing because uh, I'm functionally illiterate at math
Um, so they had me stocking shelves
Uh, and that was awful
I was in the in the back room of a grocery store, which is dark and and dank just you know, moving things in the boxes, moving them out in the shelves and
You know, I was 13. I was fairly spry, but I was sore after a long day
Uh, and uh, it was also super tedious
It's I think the point of my parents getting me that job is was to show me that like it looked there
If you don't work hard at times like you're going to work hard um, and so um
It was it was it was difficult job and would I have loved a humanoid to do it for me? Yeah, of course
You know, I think uh, and particularly if you're talking about after hours late at night and the store is closed
This is a real
Real use case and I love to see
The things that we talk about and people banter about whether or not humanoids have a real utility or not
Here we're seeing one in practice. Yeah, what it did for you was quick and character
I mean that's truly what it did and
Hopefully the regulation at some point I'll get say gets toned down where
Uh, humans in the grocery store is excuse me humanoid
What aisle could I find the sugar one? What aisle can I find the crackers on because you forget things different grocery stores
Different aisles different layouts you can ask questions really really helpful there
Do we have any insights or signals to how big this potential deployment could get there? Yeah, it's small right now
And that's intentional so intentionally so but
Fotex is is owned by the sailing group which is Denmark's largest retailer which has more than two thousand stores across six countries
Um, and so it's a trial run at the Fostex store now
But it could it could expand to many more stores in the future
And you know something to call out here because folks are good and they're gonna want to know like what kind of who's
Uh, humanoid is being used in this case. It's a unitary
G1 human weight
um, and the Danish startup hive robots supplies the hindale intelligence platform
Whereas unitary then supplies the hardware
So the shape of this is a European customer potex a local integrator in the hive robots with their hindale intelligence platform
And a Chinese hardware supply chain
And I think this is just goes to your
Code earlier about the the belt road initiative. This is we're going to see more and more of this because China's lead in robotics and humanoids
um is growing
The number of humanoids deployed in China. The number of humanoids being built in China and the way humanoids and being exported out of China
Um is unparalleled right now
And what you're seeing is you know, yes, maybe you partner with a local integrator
But you know these European stores and retailers are not having any issue
um
Using the Chinese hardware and the other thing with China and robotics they have the raw materials to make them
And that is a key thing and obviously the part of the raw materials when the most important things that will make it when we first started this journey
You and I had a deep dive on was magnets the importance of magnets
And in China also controls that
Do we have any in in-site or projections around when the humanoids might go unsupervised
In the grocery stores without having to have a human all use in term overlord
I'm sitting here to to monitor them. They'll be able to stock the shelves fully atomically unsupervised on their own
Yeah, they have a target date the goal is independent overnight restocking by early
2027 with the store closed and so um you know with that
They're they're hoping that they can reduce as you said the human operator
supervision um and
and the um kind of staffing ratio needed if you were to put additional humanoids into
service there
But they are targeting by early 2027
Independent overnight restocking with the store closed um now the data slot would have to be changed if if the humanoid were to work
autonomously among the shoppers
Uh, I don't anticipate that changing by 2027
Um, but I do think overnight use will will happen in it just a you know what so two quarters now three quarters from now
Uh, and I will say I'm sorry. My Danish is not that good. I'm saying I'm doll
Uh, uh, and that may be pronouncing it incorrectly. I'm recalling this from Marvel movies
And so that may not be the correct pronunciation
So please forgive me if for all of our Danish speaking friends out there
Uh, by the way, we're talking about Copenhagen earlier beautiful city. I enjoyed my time there
wonderful
Place to go and visit so I encourage folks to go there and go to the phone tech store and ask them
As you humanoid doing ask how all the humanoid is doing
I'm really curious when you're in that store could you find out how much it costs?
I love to know that do we have the insights on what the operational cost are
And the insights into the economics here. So yeah, we do have a little bit uh of insights
You know, Omega has uncovered that it believes the hardware price so the the price for the unitary
G1 was a little above $50,000
And that the year one total cost of ownership is going to exceed over $100,000 once integration
Licensing safety engineering and tooling are included
And so with the throughput being slower than human speed right now
Um and the continued human kind of oversight for it uh, this is not a super profitable
Uh, venture at this point, but I don't think that's the point of it. I think the point of it is really to understand
Um, as the robotics get better in terms of you know, things like gripper and navigation and need less supervision
As the cost of the actual hardware goes down and the ability to
Um, do more tasks right other than just stocking shelves. Let's maybe check
Expiry dates right empty is shelf detection finding misplaced goods
Things of that nature things that you can do in a store when when it's closed overnight
Um, I think you'll see the economics improve, but I think the
Real economics here are are not really the goal of this uh, because they don't
They don't suggest that this is something you'd want to scale at the moment
But what they're trying to do is learn as much as possible
Because a lot of those costs will come down
Pretty quickly over time cost are going to come down and
Both the green this problem choice always agrees that humanites are going to scale they're going to scale globally
But as they scale there are risks and omega has surface two quick risks here
Risk one Danish regulation continues barring autonomous operation around general shoppers
That's a real risk because we know the Europeans love regulation all you got to do is dial 1-800-GDPR
That's all I'll say they do and but in this case it might not be a bad thing to have somebody at least put a marker down about how
Humans and robots should interact in certain environments
Right, I think
That may not be where the market set forever
But I do believe this conversation
needs to happen and it needs to happen quickly in many places because the human
weights are advancing, whether that's in the home, in the factory, in grocery
stores, and a better understanding of what our expectations are, of what
liability is, where that liability falls, right? Because even in this instance,
you have the folks that make the hardware, you have the folks that are
actually putting the intelligence into it, and then you have the folks that
are operating it. There's three different potential places along that
chain that you'll fault-to-lie. And so I think I would credit any state that
takes a first stab at that. It may not be right, and it's certainly not going to
be where I think it needs to be 10 years from now or five years from now, when
there are many more human weights out there. But I would love to see this
discussion get going because I do believe that regulation is going to be one of
those long poles for human weights and their consumer use. We'll continue to
watch it because it's an involvement. Marco, which brings us to the final risk of
the show here. One-to-one robot ratio and slower than human speed limit
productivity gains. That is true, it's something to watch. And then, as Elon
said, and he is correct, it goes back a lot of it to the dexterity. So that's
something to really watch. And what is the key with dexterity that we highlighted
earlier? For sure, that is clearly an issue, right? And then secondly, it is
the learning and the training of the robot for those situations, right?
Understanding how do I recognize that a shelf is empty? How do I recognize what goes
into that shelf? How do I recognize that? You know, what I'm picking up is a can
which can be dropped, you know, from a certain level as opposed to a glass that
can't be dropped from that same level. And so a lot of this has to do with, you
know, is the is the training of the robot and what's where is that done? Is it
based on the physical world? Is it based on a visual world? Is it a large
language model? We saw Astra, right? Come out and show that it had done some pretty
unique things with the robotic arm and sorting. So a lot of this will
actually, in addition to the production potential for slowdown with the
actuators and the magnets, how are these embodiments of physical eye being
trained and are they succeeding in the tasks that they're being trained for? Or is
it the task specific, right? Or is it a more general-purpose training that allows
the human way to maybe respond and react to different situations that might
find in the store? Well, that's going to be a really fascinating thing because I
do think personally, and I'll make a prediction, I do think in the next 12 months
we're going to see just unheard of advances in terms of the capabilities of
humanoids because of so many smart folks working on those types of issues.
Humanoids are going to scale. They're going to do repetitive tasks. They're
going to do dangerous tasks. They're going to make our world safer. They're
going to make our world more efficient, and it's just part of the overall autonomy
economy. Each and every week, Rob and I are going to be here breaking down the
signals in the autonomy economy. In addition to the work that we do, our
ground truth intelligence, our Omega servers, which are general subscriptions that
were offer and just subscribers, where they get constant real data, proprietary
data, reach out to us at
[email protected]. The future is bright. The future
autonomous. The future is great shows with musical references. Rob, I don't know
how we're going to top this next week, but we're going to have to get something
else cooking. Oh yeah, I mean, I'm going to go put on some
moutain right now. You know, I've got my 17 year old in the house. He's a big fan,
and I think we're just going to have a little dance party. Rob is going to
enter the 36 chambers. Great show. Opinions here are as of the recording date
may change without notice, and do not necessarily reflect the views of autonomy AI
incorporated, its affiliates, or any speaker's employer.
Content, including the road to autonomy indices, is for informational and
educational purposes only. Autonomy AI is not a registered investment advisor
or broker dealer and does not give investment legal accounting or tax advice.
Nothing in our public or gated content is investment advice, a recommendation,
an offer, or a solicitation to buy or sell any security or financial product,
and should not be relied on to evaluate any transaction. This program may contain
forward-looking statements, investments involve risk, including loss of
principle. Past performance is not indicative of future results.
Hosts, guests, and affiliated entities may hold long or short positions in
securities mentioned. Content is provided as is without warranty.
We disclaim liability for reliance on this information.
Consult your own investment, legal, and tax advisors before making any financial
decision.
PodcastTranscript AI