🧠 NVDA | Is Nvidia the Most Powerful Company on Earth?

0:00
it's tough to think of a more important
0:01
industry today than computing or when we
0:03
depend on more in our daily lives
0:05
microchips are used in countless
0:07
applications from cell phones and solar
0:09
panels to satellites and self-driving
0:11
cars this year alone the global chip
0:13
industry could bring in around 600
0:15
billion dollars and by 2030 spending on
0:17
artificial intelligence ships alone
0:19
could close in on 2 trillion dollars per
0:21
year the competition for this market has
0:24
been turned up to 11. with multiple
0:26
trillion dollar companies like amazon
0:28
microsoft and google all starting to
0:30
design their own chips but it could be
0:32
that none of these tech giants will sit
0:34
on the computing throne actually the
0:36
number one contender weighs in at just
0:38
under 600 billion dollars of market cap
0:42
which is one quarter of the size of
0:44
microsoft the company i'm talking about
0:46
is nvidia and there are three key
0:48
reasons why it might dominate this
0:49
market over the next decade your time is
0:51
valuable so let's get right into it
0:54
nvidia is known for their graphics
0:55
processing units or gpus when most
0:58
people think of nvidia's graphics cards
1:00
they probably think of nerds playing
1:01
computer games that's for a pretty good
1:03
reason as of late last year nvidia had a
1:06
dominant 83 percent of the market share
1:08
for discrete gpus with amd owning the
1:11
other 17
1:12
that means that nvidia sells five out of
1:14
every six graphics cards if you look at
1:17
all gpus sold last year instead of just
1:19
separate graphics cards amd and intel's
1:22
chip shipments were both down by double
1:23
digits due to a global chip shortage
1:26
ouch but nvidia's overall shipments were
1:28
up eight percent here's the thing even
1:30
though nvidia is dominating that market
1:33
it's not the only one they operate in
1:35
over the past few years nvidia has
1:36
branched out from gaming into areas like
1:38
data science artificial intelligence
1:41
modeling and simulation and even edge
1:42
computing that last one is important
1:44
because the smaller nvidia can make its
1:46
chips the closer to the edge of the
1:48
connected world they can get devices at
1:50
the edge are much cooler than just cell
1:52
phones they're things like drones and
1:54
delivery robots and autonomous cars
1:56
pretty much everything i talk about on
1:57
this channel but you don't win the chip
1:59
market by making the smallest chips you
2:01
win it by making the best solutions to
2:03
real problems for the best price and if
2:06
you want to understand how nvidia is
2:07
doing that you need to know what kinds
2:09
of problems gpus solve that cpus aren't
2:12
really built for in my eight years as a
2:14
researcher at mit i worked on a lot of
2:16
programs and prototypes that use gpus
2:18
for a wide variety of tasks here are
2:21
some of my biggest takeaways gpus are
2:23
really good at something called linear
2:25
algebra instead of working with single
2:27
numbers this special area of math works
2:29
with sets of numbers called vectors
2:30
matrices and tensors if that sounds
2:33
absolutely terrible to you that's
2:35
because it is i still have nightmares
2:37
about this stuff from grad school
2:38
despite being the stuff of nightmares
2:40
this special math is the best way to
2:42
solve many similar problems in parallel
2:44
for example if i'm playing a video game
2:46
and i want to move around every surface
2:48
of every object in the scene needs to
2:50
adjust depending on my avatar's position
2:53
and orientation instead of doing that
2:55
math problem object by object for each
2:57
player in a scene developers can just
2:59
build digital environments then those
3:01
graphics get loaded into a gpu that can
3:04
translate rotate and scale everything
3:06
depending on the positions and the
3:07
pointing vectors that you give it for
3:09
each player well it turns out that
3:11
solving many similar problems in
3:12
parallel this way is useful for way more
3:14
than just gaming think about an
3:16
autonomous vehicle that needs to process
3:18
many cameras and radar feeds as it
3:20
navigates through a real environment
3:22
that's kind of like a video game except
3:24
you have eyes all over the body and the
3:26
danger is very real
3:27
wait is this just an episode about all
3:29
of my nightmares think about all the
3:31
graphical applications this type of math
3:33
can be used for drones in the sky
3:35
telescopes trying to find a specific
3:37
object facial recognition social media
3:40
filters the list goes on and on here's
3:42
the thing tensor math isn't just useful
3:44
for graphics you can use this kind of
3:46
math to solve many kinds of problems you
3:48
can build a matrix of how genes express
3:50
themselves in a wide variety of diseases
3:52
for medical diagnostics or you can solve
3:55
a bunch of valuation models at once for
3:57
every stock on the market how about
3:59
trying to solve the same hard
4:00
cryptographic problem using many
4:02
different guesses at the same time
4:04
that's crypto mining what about scoring
4:06
pieces of content for different users
4:08
based on their interests and showing
4:09
them whatever scores the highest in the
4:11
matrix that's a content recommender
4:13
system like netflix or spotify the
4:15
company with the best parallel number
4:17
crunching solution is going to win the
4:19
lion's share of the computing market for
4:21
all these different verticals that need
4:23
this same type of system one big reason
4:25
nvidia is outselling competitors like
4:27
amd five to one is how they handle more
4:30
complex and power-hungry computing tasks
4:32
like ray tracing reflections and
4:34
upscaling graphics at a high level
4:37
there's a big fundamental difference in
4:38
how nvidia and amd build their chips amd
4:41
puts a physical ray accelerator onto
4:43
each core of their gpus it's kind of
4:46
like a supercharger but for computer
4:47
chips instead of cars i'm just kidding i
4:50
don't know anything about cars nvidia's
4:52
specialized ray tracing cores are
4:54
physically separate from their other gpu
4:56
cores nvidia has another special type of
4:58
core called the tensor core because they
5:01
focus on doing this tensor math very
5:03
very fast at the cost of accuracy then
5:06
nvidia fixes those inaccuracies
5:07
separately using deep learning super
5:10
sampling or dlss which is a major
5:12
selling point for their products for
5:14
dlss they train an artificial
5:16
intelligence model on high resolution
5:18
scans of every video game then apply
5:20
that ai to adjust the output of the
5:22
tensor cores to a player's version of
5:24
that same game nvidia's secret sauce is
5:27
pushing the right limits of what can be
5:28
done with specialized hardware while
5:30
offloading as much work as possible to
5:32
specialized software then using ai to
5:35
optimize the output of both which is
5:37
useful everywhere from video games to
5:39
real-life camera data to simulating
5:41
physics experiments to ai applications
5:43
like speech and object recognition
5:45
nvidia also has cuda which is their
5:48
coding language and development platform
5:50
cuda is how scientists and researchers
5:52
can use nvidia's gpus for their own
5:54
applications like my projects back at
5:56
mit cuda is a massive ecosystem with a
5:59
huge community of developers kind of
6:01
like apple swift for ios apps new
6:03
projects and new developers tend to
6:05
flock to where there's already an
6:06
established developer community well
6:09
cuda is designed specifically for nvidia
6:11
devices and it's up to other device
6:13
makers like amd to make it work with
6:15
their chips if you're finding value in
6:17
this research so far let me know in the
6:19
comments below i think you'll also find
6:21
a lot of value by going to public.com
6:23
ticker symbol u and claiming a free
6:25
slice of stock worth up to 70 when you
6:27
fund your account i'll leave a link to
6:29
that exclusive offer for you in the
6:30
description below nvidia's specialized
6:32
tensor computing chips ai enhanced
6:34
computing software and programmer
6:36
ecosystem are why they dominate the gpu
6:39
market but they're also the fundamental
6:41
building blocks for winning the
6:42
enterprise high performance computing
6:44
market this is a tough capability to try
6:47
and build yourself so it makes sense for
6:49
a business to connect to a pre-built
6:51
scalable solution and customize it to
6:53
their needs for years nvidia has been
6:55
designing its chips for cloud-based
6:57
artificial intelligence applications
6:59
knowing that this demand was coming
7:01
today nvidia's chips sit inside the
7:03
servers of almost every cloud computing
7:05
provider out there including ibm oracle
7:08
amazon web services microsoft azure and
7:11
google cloud even though amazon and
7:13
google both make their own ai chips i'm
7:15
pretty sure this played a big role in
7:17
why regulators didn't allow nvidia's
7:19
acquisition of arm to go through then
7:21
nvidia would be the dominant player in
7:23
gpus and cpus in servers desktops mobile
7:26
devices and a ton of other different
7:28
kinds of companies would end up relying
7:30
on their support speaking of which
7:32
remember the whole specialized hardware
7:33
thing well while other companies are
7:35
making generalized cloud ai chips nvidia
7:38
is building more specialized chips and
7:40
software suites for a wide variety of
7:42
applications like networked
7:43
supercomputing xero trust cyber security
7:46
edge ai medical imaging and diagnostics
7:48
and processing for robotics applications
7:51
nvidia is building maxine which connects
7:53
computer vision natural language
7:55
processing and avatar animation into a
7:58
real-time conversational robot these are
8:00
three very different ai applications
8:02
coming together and interacting in a way
8:04
that behaves more like a human which can
8:07
be used as a talking kiosk in malls and
8:09
restaurants a concierge in hotels and
8:11
robo taxis or a simultaneous
8:13
multi-language translator for video
8:15
conferences this is really next
8:17
generation stuff and it's powered by
8:19
artificial intelligence there's actually
8:21
one more initiative i want to talk about
8:23
before covering nvidia's financials and
8:25
that's the omniverse no not the
8:27
metaverse people actually like this one
8:29
the omniverse is a virtual platform
8:31
built for real-time photo realistic
8:33
simulation architects and engineers
8:35
often create and work on 3d models of
8:38
their projects manufacturing companies
8:40
want to see the real-time status of
8:41
their factory floors and equipment
8:43
scientists want to visualize the data
8:45
they capture whether it's radio waves or
8:47
weather patterns or something else
8:49
entirely so nvidia's omniverse platform
8:52
what's creators designers researchers
8:54
and engineers connect their existing
8:56
asset libraries tools and projects so
8:59
that they can collaborate in a shared
9:00
space and interact with digital twins of
9:03
their projects on top of that software
9:05
providers and developers can build and
9:07
sell extensions apps and microservices
9:09
for the omniverse to expand its
9:11
functionality so knowledge workers
9:13
across different companies don't always
9:15
have to reinvent the wheel here's the
9:17
thing when many different industries can
9:19
share resources like this the creator
9:21
economy booms ecosystems like the
9:23
omniverse will let independent
9:25
developers and micro studios flourish
9:27
because they can focus on building one
9:29
great module or widget without having to
9:31
build the entire infrastructure needed
9:33
to support it if the omniverse platform
9:35
gets big enough there could be a real
9:37
customer base for this kind of work
9:38
encouraging more small developers to
9:40
join the marketplace which kicks off a
9:42
virtuous cycle of more ready to use
9:45
assets bringing in more customers which
9:47
brings in more developers which brings
9:48
in more assets and so on i really think
9:51
that the omniverse could end up being
9:52
something like apple's app store but for
9:54
enterprise businesses now that we
9:56
understand the absolute basics of what
9:58
this incredibly advanced vertically
10:00
integrated high performance computing
10:02
company does we can have a productive
10:03
conversation about its market its
10:05
finances and its future as a tech nerd
10:08
and an investor i think this is what
10:10
really matters today over ninety percent
10:12
of the total computing done on most big
10:14
hybrid supercomputers is done on the gpu
10:17
side and that number could climb to as
10:19
high as 95 percent in the near future
10:21
gpus really put in the work of the 146
10:25
supercomputer systems surveyed 136 use
10:28
nvidia's gpus that's a 93 market share
10:31
of super computers for nvidia and that
10:33
number has been trending up over time
10:36
research suggests that by 2030 gpus
10:38
could be the dominant kind of processors
10:40
for most new workloads growing at a 21
10:43
compound annual growth rate over the
10:45
next eight years as these kinds of
10:46
capabilities get cheaper and get adopted
10:49
on smaller and smaller systems as ai
10:51
systems get cheaper to build and train
10:53
more companies will use them in more
10:55
applications as a result projected
10:57
hardware spend on ai will grow by a
10:59
hundred fold by 2030 from 17 billion a
11:02
year in 2021 to 1.7 trillion dollars
11:05
that's a 67 compound annual growth rate
11:08
over the next nine years meaning this
11:10
market is going to expand a lot for the
11:12
rest of the decade as a result of that
11:14
1.7 trillion dollar spend by 2030 seven
11:17
trillion dollars in enterprise value
11:19
could be created in the ai computing
11:21
space which would generate returns of 57
11:24
at a compound annual growth rate during
11:26
the next nine years of course not all of
11:28
that belongs to nvidia but we should
11:30
expect nvidia to keep achieving
11:32
exponential growth by growing the
11:33
adoption of their ai hardware their
11:35
services and their applications over the
11:38
next decade inside this quickly growing
11:40
market if you're anything like me you
11:42
raise an eyebrow when you see such huge
11:44
projections for growth like this so to
11:46
see if their story checks out let's take
11:48
a look at nvidia's latest earnings
11:49
presentation which came out in the
11:51
middle of february nvidia's gaming
11:53
revenue which comes from their desktop
11:55
and laptop geforce gpus is up 37 year
11:58
over year look how stable this revenue
12:00
growth has been even with chip shortages
12:02
messing up the entire market data center
12:04
growth is even more extreme showing an
12:07
11 quarter over quarter growth or a
12:09
whopping 71 percent year-over-year
12:12
growth that was led by compute products
12:14
and strong demand for nvidia ai in
12:16
applications like natural language
12:17
processing deep learning recommendation
12:20
systems and cloud computing exactly the
12:22
type of stuff we learned about
12:24
throughout this episode and that seems
12:25
to really line up with the crazy 67
12:28
compound annual growth rate in ai
12:30
hardware spend that i just talked about
12:32
nvidia's data center revenue now makes
12:34
up 40 of nvidia's total revenue while
12:37
gaming makes up 46 but their data center
12:40
revenue is growing twice as fast so i
12:42
expect we could see these number flip by
12:44
the next fiscal year and nvidia's
12:46
primary source of revenue could become
12:48
their data center solutions let me point
12:50
out three quick things about the
12:51
metaverse here since you really can't
12:53
talk about the metaverse without
12:54
mentioning nvidia these days first off
12:56
the metaverse needs nvidia but nvidia
12:59
does not need the metaverse today
13:01
nvidia's technologies are found in
13:03
nearly every supercomputer and with
13:05
every major cloud provider and they have
13:07
millions of registered developers around
13:09
the world nvidia is growing at such a
13:11
fast rate today because it's solving the
13:13
computing problems of today and one of
13:16
those computing problems is for the
13:18
metaverse meta platforms unveiled its
13:20
new ai supercomputer called research
13:22
supercluster which is powered by
13:24
nvidia's technologies later this year
13:26
they're going to expand that
13:27
supercomputer to include 21 times more
13:30
gpus something that can only happen
13:32
because nvidia builds its chips with
13:34
cloud computing in mind and third even
13:37
if you think the idea of the metaverse
13:38
is stupid building a connected
13:40
enterprise marketplace for shared assets
13:42
tools and research is certainly not the
13:45
omniverse solves real problems today and
13:47
allows talented individuals and small
13:49
teams to operate in a market usually
13:51
reserved for enterprises it's pretty
13:53
cool to see a free version of the
13:55
omniverse being announced for
13:56
individuals and i'm excited to dive into
13:58
it and share what i learn after all this
14:00
is the channel that invests in you if
14:02
you want to learn more about the
14:03
advanced technologies driving our future
14:05
and the companies behind them this video
14:08
right here is for you
14:09
either way stay long stay strong and
14:11
thanks for watching until next time this
14:13
is ticker symbol you my name is alex
14:16
reminding you that the best investment
14:18
you can make
14:19
is in you