Latest / Connecting the Dots / AI’s Energy Debate and the Rise of Robotaxis

Transcript

AI’s Energy Debate and the Rise of Robotaxis · Connecting the Dots

0:00
Today's tech news moves fast, but understanding
0:03
it shouldn't be hard. That's why Snarfl Solutions
0:06
Group brings you Connecting the Dots, where we
0:09
break down what's happening in tech and why it
0:11
matters to you. Now, let's dive in with your
0:14
hosts. Good morning, everyone. I'm Alex. And
0:19
I'm Morgan. And welcome to another edition of
0:21
Connecting the Dots. We are brought to you by
0:23
the Snarfl Solutions Group. Yep. It is Monday.
0:28
February 23rd, 2026. And we're coming to you
0:31
live from Sacramento, California. We are. And
0:34
I got to say, it is a really beautiful start
0:37
to the week here on the West Coast. It really
0:39
is. But before we get too smug about our sunshine,
0:43
we probably need to address the elephant in the
0:45
room. Or well, the blizzard on the map, I guess.
0:48
Right. Yeah. If you're tuning in from the mid
0:50
-Atlantic or the Northeast right now, you definitely
0:52
know what we're talking about. Yeah. The national
0:54
weather map just looks incredibly intense today.
0:57
There's this major winter storm that's just hammering
1:00
that entire region. We're seeing all these reports
1:02
of heavy snow, possible blizzard conditions,
1:06
and even some coastal flooding. It sounds pretty
1:08
rough out there. Yeah, it's definitely a stay
1:10
inside and drink cocoa kind of day if you can
1:12
manage it. For sure. I mean, the National Weather
1:14
Service is literally warning people about those
1:16
blizzard conditions along the northeastern seaboard.
1:19
So if you have travel plans or a commute in that
1:22
corridor today, please be extremely careful.
1:26
Absolutely. The roads are looking pretty much
1:27
impossible in some of those spots. They are.
1:29
And meanwhile, You know, the Pacific Northwest
1:31
and our neighbors up in North Central California,
1:33
they're getting another round of that wet, windy
1:35
weather, too. Right. But here in Sacramento.
1:38
Here in Sacramento, we are looking at a high
1:40
of 67 degrees and partly sunny skies. It almost
1:45
feels a little unfair. A little guilty to say
1:47
it out loud. It does. But, you know, we'll take
1:50
the vitamin D where we can get it. We certainly
1:52
will. So let's get a really quick snapshot of
1:55
the financial world before we jump into the main
1:57
topics today. Let's do it. It was a pretty solid
2:00
end to the week last Friday, February 20th. The
2:03
market seemed to kind of shake off some of those
2:06
mid -month jitters. The Dow Jones closed up about
2:08
.61%, sitting at 49 ,625 .97. Not bad at all.
2:14
Yeah. And the S &P 500 actually had an even stronger
2:17
showing. It closed just a hair under 1 % up,
2:21
finishing Friday at 6 ,909 .51. Right. It's interesting
2:26
to see that resilience, especially with some
2:28
of the volatility we've been tracking. Yeah,
2:30
exactly. But if we look at this morning, crypto
2:32
is taking a bit of a breather. Bitcoin is trading
2:35
Monday morning around $66 ,088 .21, which is
2:40
down about 2 .26 % from where we were. Seems
2:44
like just a little bit of profit taking to start
2:46
the week. Yeah, a little cooling off is healthy.
2:48
But speaking of things that definitely should
2:51
not cool off, let's talk about how we are actually
2:54
powering through this Monday morning. Oh, definitely.
2:56
Because, you know, I absolutely cannot function
2:58
without a serious caffeine kick. I am right there
3:01
with you on that. And that is exactly why we
3:03
are powered by Strong Coffee. Yes, we are. If
3:05
you haven't tried this stuff yet, you really
3:07
need to listen up. Strong Coffee is a premium
3:09
instant coffee, but you have to just completely
3:12
erase the memory of that terrible instant stuff
3:15
your grandparents used to drink. Oh, this is
3:17
completely different. Really? It tastes exactly
3:19
like a fresh traditional brew, but it's engineered
3:22
for performance. It's packed with 15 grams of
3:25
protein and MCTs, which you need for firing up
3:28
your brain first thing in the morning. And the
3:30
absolute secret weapon in this is the adaptogens,
3:33
specifically the ashwagandha. The ashwagandha
3:36
is a total game changer. It balances out that
3:38
caffeine spike. So you get this really clean,
3:41
sustained energy. Right. So there's no jitters,
3:44
no anxiety, and most importantly, no terrible
3:47
crash into in the afternoon. Exactly. It's basically
3:49
this functional latte that's ready in seconds.
3:52
And because we want all of you operating at peak
3:55
efficiency today, you can get 20 % off your entire
3:57
purchase. Such a good deal. All you have to do
4:00
is go to strongcoffeecompany .com and use the
4:02
promo code SNARFL. That's strongcoffeecompany
4:05
.com promo code SNARFL. It is genuinely the best
4:08
way to start your week. All right, let's get
4:11
into the good stuff. It's time to open up the
4:13
Morningstar, that's our internal Snoreful Tech
4:16
article report, and really connect the dots on
4:18
today's top three stories. We have a really fascinating
4:21
lineup today. It kind of feels like the theme
4:24
connecting everything is redefining efficiency.
4:26
I like that. Yeah, we're looking at everything
4:28
from the biological energy cost of a single thought
4:30
all the way to the capital efficiency of running
4:33
a robotaxi fleet. Well, let's jump straight into
4:36
our first story then. This one comes out of the
4:39
India AI Impact Summit 2026, specifically an
4:43
event there called the Express Adda. Right. And
4:46
OpenAI CEO Sam Altman was speaking and he made
4:49
some comments that are really stirring the pot
4:51
right now. He decided to push back pretty hard
4:54
against this whole narrative that AI is some
4:56
sort of energy villain. Yeah. And this is a debate
4:59
that has been raging for years now. We all know
5:02
data centers consume massive amounts of electricity,
5:05
but Altman is trying to completely reframe the
5:07
comparison here. How so? Well, he's arguing that
5:10
comparing the massive energy it takes to train
5:12
an AI model against a single human thought is
5:16
fundamentally unfair. He actually called it intellectually
5:19
dishonest. Okay, I see where he's going with
5:21
that. It's sort of the difference between building
5:24
the entire library and just checking out a single
5:26
book. You wouldn't count the cost of pouring
5:29
the concrete for the library every time someone
5:32
reads one page. Yeah, it is exactly the distinction.
5:34
He's making it was between training and inference
5:37
Training is that massive one -time energy expenditure
5:40
to create the digital brain? Inference is just
5:44
that tiny amount of energy used when you ask
5:46
chat GPT a question and it answers you But he
5:49
actually took it a step further than that, didn't
5:52
he? He didn't just compare it to a computer process.
5:54
He compared it to human biology. He did. And
5:57
this is where it gets really provocative. He
5:59
argued that if you want to talk about the training
6:01
cost of intelligence, you have to look at the
6:03
training cost of a human being. Which is wild
6:06
to think about. Right. He literally said it takes
6:08
like 20 years of life and all the food you eat
6:11
during that time before you get smart. That is
6:13
just a staggering way to look at human existence.
6:17
He's essentially amortizing 20 years of breakfast,
6:19
lunches, and dinners as the energy bill for human
6:22
intelligence. And he didn't even stop there.
6:25
He went totally cosmic with it. He referenced
6:27
the widespread evolution of a hundred billion
6:30
people who had to live, learn, and avoid getting
6:33
eaten by predators just to produce the genetic
6:36
baseline for a modern human. Wow. His whole point
6:40
is that biological intelligence has this astronomically
6:43
high sunk cost that we just completely ignore
6:46
when we criticize the power draw of a data center.
6:49
It's a very Silicon Valley engineer way of viewing
6:52
the world, isn't it? Just boiling human existence
6:54
down to caloric inputs and outputs. But if we
6:57
put the philosophical part aside for a second,
6:59
just look at the math. Is he actually right about
7:01
the inference part? Strictly speaking, in terms
7:04
of joules per answer, He might be. He claims
7:06
that if you compare the energy of a single chat
7:08
GPT query against the calories a human brain
7:11
burns to answer that exact same question, AI
7:14
is likely more efficient. I guess if you think
7:16
about it, if I spend an hour scaring at a blank
7:18
screen trying to write an email snacking on almonds
7:20
and burning metabolic energy, a two second chat
7:23
GPT query is objectively less wasteful in that
7:26
specific moment. Exactly. And he used this platform
7:30
to debunk some of the wilder myths out there
7:32
right now. Like, there was this viral statistic
7:36
claiming one single query uses as much energy
7:40
as charging an iPhone one and a half times. I
7:42
saw that one going around. Yeah, Altman shut
7:44
that down completely. He said there's no way
7:46
it's anything close to that. He also addressed
7:49
the water usage concerns too, right? Those headlines
7:52
about AI drinking up all our local reservoirs.
7:54
He called those totally fake, at least in the
7:56
context of modern infrastructure. He explained
7:59
that those scary numbers, like the stats saying
8:01
it takes 17 gallons of water per query, those
8:05
were based on older data centers that used evaporative
8:07
cooling. Ah, okay. And modern centers have largely
8:11
moved away from that technology entirely. So
8:13
he's saying the per query efficiency is actually
8:15
quite good now. But, and there is always a, but
8:18
with this stuff he did admit that the aggregate
8:21
total is a problem. Even if one query is incredibly
8:24
efficient, a billion queries a day is still a
8:27
massive amount of power. Yes, he completely acknowledged
8:30
that the total load is massive and it's only
8:32
growing. But his solution isn't to slow down
8:35
the development of AI. Right, of course not.
8:37
His solution is to speed up energy production.
8:40
He is calling for a massive acceleration in nuclear
8:43
wind and solar power. He effectively said, the
8:47
problem isn't the AI. The problem is we aren't
8:50
building enough clean power plants to support
8:53
it. It's a bold stance to take. Basically saying,
8:56
don't turn off the computer, just build me a
8:58
bigger reactor. Pretty much. And he capped the
9:01
whole thing off with this totally casual prediction
9:04
that artificial super intelligence is not that
9:06
far off. Just threw that in there. Yep. So the
9:09
energy demand isn't going down anytime soon.
9:11
If anything, we are just getting warmed up. Well,
9:14
while Sam Altman is out there trying to optimize
9:16
the energy of thinking, our next story is about
9:19
a company trying to optimize the business of
9:21
moving. Let's talk about Uber. This is a story
9:23
about a massive strategic pivot. If you've been
9:26
following the autonomous vehicle space for a
9:28
while, you probably remember the Uber ATG days.
9:31
Oh, yeah, the advanced technologies group. Right.
9:34
Years ago. Uber was burning billions of dollars
9:36
trying to build their own self -driving cars
9:39
from scratch. They had the test tracks, the thousands
9:41
of engineers, the modified Volvo's, the whole
9:44
nine yards. And that turned out to be a total
9:46
financial disaster for them. They eventually
9:48
had to sell that entire division off. For a while
9:50
there, it really looked like Uber was just going
9:52
to be a bystander in the robotaxi race while
9:55
Waymo took over. It did. But now they've launched
9:58
something called Uber Autonomous Solutions. And
10:01
this is such a brilliant move. They realized.
10:04
They shouldn't try to be the manufacturer of
10:06
the robots. They should just be the operating
10:08
system for them. They're visiting to become the
10:10
mission control for everyone else. Precisely.
10:13
They are targeting third -party AV companies.
10:17
We're talking about players like Wave, Wear Ride,
10:20
Neuro, and Wallaby. Okay. Now these companies
10:23
have fantastic driving software, but they don't
10:25
have the massive consumer infrastructure that
10:28
Uber has built up over the last decade. Right,
10:29
they don't have the app on everyone's phone.
10:31
Exactly. So Uber is offering them fleet management,
10:35
vehicle financing, and regulatory assistance
10:37
to get them on the road. It's basically the Intel
10:40
Inside strategy, but applied to ride hailing.
10:42
But there's one specific feature they're offering
10:45
in this package that I think is really crucial
10:47
and that's the remote assistance. Oh, this is
10:50
the absolute key to the whole thing. One of the
10:52
biggest hurdles for autonomous cars right now
10:55
is the edge case. The weird stuff. Yeah, the
10:58
weird stuff the unpredictable construction zone
11:01
with confusing signage or a police officer using
11:04
complex hand signals in an intersection Those
11:07
are things the AI still really struggles to interpret
11:09
safely, right? So uber is providing this layer
11:12
where if a robotaxi gets confused a human agent
11:16
sitting in a control center can remotely assess
11:19
the situation and help it out. No, we should
11:21
probably clarify for everyone, this isn't a guy
11:23
with a steering wheel and a joystick driving
11:24
the car remotely like a video game, right? No,
11:27
not at all. Because the latency, just the delay
11:29
in the cellular signal over 5G, would make real
11:32
-time steering incredibly dangerous. Exactly.
11:35
You cannot steer a two -ton vehicle in real time
11:38
over a cell network with zero lag. It doesn't
11:40
work. Instead, this remote assistance is about
11:43
high -level guidance. Okay, how does that work?
11:45
The car... basically stops, assesses the confusing
11:49
obstacle, and pings the center asking, should
11:52
I go left or right around these cones? The human
11:55
looks at the cameras, clicks go left, and then
11:57
the car actually executes the physical steering
11:59
and braking maneuvers itself. That makes a lot
12:01
of sense. And Uber providing this specific service
12:03
at scale significantly lowers the barrier to
12:07
entry for all these smaller AV startups. It solves
12:09
a massive economic problem for them. They don't
12:12
have to spend millions building their own human
12:14
command centers to monitor their fleets. Right.
12:16
And CEO Darcos Rashahi is clearly playing the
12:19
long game here. His quote was, we want to work
12:21
with all the software providers. It's a survival.
12:23
play. Absolutely. Just look at San Francisco
12:26
right now. Waymo is already eating significantly
12:28
into the traditional ride share market. If Uber
12:31
doesn't have robots on its network soon, they
12:33
just become an overpriced legacy human taxi service.
12:37
By aggregating all these different autonomous
12:39
providers, Uber ensures that when you open their
12:42
app, you get a ride, whether it's a human driver,
12:46
a Waymo, or a WeRide bot. They honestly do not
12:49
care who owns the car. They just want to make
12:51
sure they own the customer relationship. It's
12:54
classic platform economics. And just to prove
12:56
that this isn't merely some theoretical slide
12:58
deck presentation, we actually have a real world
13:02
launch happening right now that utilizes this
13:04
exact model. Which leads us perfectly into our
13:06
third story for today. We are heading over to
13:09
Dubai. Yes. This is the Uber Autonomous Solutions
13:13
strategy in action. Uber and WeRide have officially
13:16
launched RoboTaxi passenger rides in Dubai. And
13:19
Dubai has always been incredibly aggressive with
13:21
future tech adoption. So how does this actually
13:24
look for a user on the ground? Can you just hail
13:26
a robot to take you to the Burj Khalifa right
13:28
now? If you are in the right zones, yes. It's
13:31
currently available in tourist hotspots, specifically
13:33
areas like Sukhum and Jumeirah. OK. You literally
13:36
just open your standard Uber app. You select
13:38
the autonomous option from the vehicle list and
13:40
a rear ride vehicle rolls up to the curb. Now,
13:43
the word autonomous can mean a lot of different
13:45
things depending on the company. Is this truly
13:48
driverless right now? Or is there a safety net
13:51
built in? For this initial rollout phase, there
13:54
is still a vehicle specialist sitting in the
13:57
front seat. Gotcha. They aren't actively touching
13:59
the wheel or pedals during the ride, but they
14:02
are there specifically to take over if things
14:05
go sideways. Makes sense for a new market. But
14:07
the roadmap they've laid out is very aggressive.
14:09
They are targeting fully driverless service,
14:12
meaning absolutely no human in the front seat
14:14
by early 2026. Which is incredibly fast. I mean,
14:18
we're already in early 2026 right now, so they
14:20
really mean any day now. Exactly. And this connects
14:23
perfectly back to that Uber strategy we were
14:25
just discussing. Because this fleet in Dubai
14:28
isn't actually run by Uber, and it isn't run
14:31
by WeRide either. Right. It's operated by a local
14:33
company called Tawasul. So Tawasul owns the physical
14:37
cars and cleans them. WeRide provides the artificial
14:40
intelligence to drive them. And Uber provides
14:43
the network of paying customers. It's a completely
14:46
decentralized supply chain for autonomy. And
14:49
it's highly necessary because Dubai has a massive
14:52
overarching goal right now. They want 25 % of
14:55
all transportation journeys in the city to be
14:58
fully autonomous by the year 2030. That is a
15:01
quarter of all trips. That's not just a cute
15:03
pilot program. That is a fundamental shift in
15:06
city infrastructure. It is huge. And it proves
15:09
that this technology is really starting to scale
15:11
globally. We ride already has something like
15:13
150 autonomous vehicles operating in the Middle
15:16
East. This isn't just a San Francisco or Phoenix
15:18
science experiment anymore. It really circles
15:20
back to that theme of efficiency we talked about
15:22
at the top of the show. Uber is making the market
15:25
more efficient by connecting all these disparate
15:27
players. Dubai is trying to make city transport
15:30
more efficient by removing human error and driver
15:33
fatigue and Sam Altman is trying to make intelligence
15:36
itself more efficient by moving it from biology
15:38
to silicon and Sitting right in the middle of
15:41
all of this our very definition of work and value
15:43
is fundamentally changing whether you are a taxi
15:47
driver navigating traffic or a writer drafting
15:50
an email the machines are just getting vastly
15:54
better at the doing part of the job. It's definitely
15:56
a lot to process on a Monday morning. It is,
15:59
but before we leave you to ponder the existential
16:02
dread of being less efficient than a server rack
16:04
full of GPUs, we want to help you with something
16:07
a little more fun and practical. We do. We want
16:09
to take a second to talk about our friends at
16:12
Old Glory. Yes. If you have that one friend who
16:15
is just completely impossible to shop for, maybe
16:18
they're huge techie, but they also have this
16:20
secret love for classic rock or NFL football.
16:24
This is absolutely your answer. Old Glory is
16:26
the ultimate destination for iconic music and
16:29
sports fan merchandise. I was actually browsing
16:32
their site the other day and the depth of their
16:34
catalog is just wild. They have everything. They
16:36
really do. They have everything from official
16:38
ACDC hoodies to incredibly obscure sports team
16:41
jerseys. And it's all high quality gear. It's
16:44
not that cheap knockoff stuff you find at street
16:46
stalls. It's the perfect place to find a gift
16:48
that says, hey, I actually know what you like.
16:51
And of course, we've got a hookup for our listeners.
16:53
Always. You can visit OldGlory .com and use the
16:57
promo code SNARFL at checkout to get 15 % off
17:00
your entire purchase. That's OldGlory .com promo
17:03
code SNARFL. Go get yourself or that tricky techie
17:07
friend of yours something really cool. So let's
17:09
quickly recap what we covered in today's Morningstar.
17:12
First up, we had Sam Altman challenging the world
17:15
to rethink energy consumption. He argues that
17:18
comparing AI training to a single thought is
17:20
unfair, and we should instead be considering
17:22
the massive biological training cost of human
17:25
evolution. And he's pushing heavily for a nuclear,
17:28
wind, and solar -powered future to support this
17:31
AI -driven world. Then we looked at Uber's incredibly
17:33
smart pivot to Uber autonomous solutions. They
17:36
stopped trying to build the hardware themselves
17:38
and started selling the infrastructure, things
17:40
like fleet management, financing, and that crucial
17:42
remote assistance to enable other AV companies
17:45
to scale up on their network. And finally, we
17:47
saw that exact model go live in Dubai. You can
17:51
now hail a WeRide RoboTaxi via the Uber app in
17:55
certain tour zones, which is a major step toward
17:58
the city's goal of making 25 % of all transport
18:01
autonomous by 2030. It's a world. where the very
18:05
infrastructure of both intelligence and physical
18:08
movement is being rewritten right in front of
18:11
us. You know, it really leaves you with a lingering
18:13
thought. If Uber succeeds in becoming the underlying
18:17
operating system for all cars and OpenAI succeeds
18:20
in becoming the operating system for all thought,
18:23
where does that actually leave us? That really
18:25
is the question. Are we going to remain the active
18:27
users of these systems? Or are we just the biological
18:30
bootloaders that got the machines running soon
18:32
to be viewed as legacy hardware? That is definitely
18:34
something to shoe on while you finish that cup
18:36
of strong coffee today. Indeed it is. Well that
18:39
is it for us today. We hope you have a fantastic
18:42
and safe start to your week. We're here to help
18:44
at Snarfl, so reach out on our website if you
18:45
have any more questions. That's a wrap on today's
18:51
top stories. Remember, tech doesn't have to be
18:54
complicated, and we're here to help connect the
18:56
dots. If you've got questions or want to learn
18:59
more, visit us at snarful .com. Thanks for listening,
19:03
and we'll catch you next time.