Monetizing AI: The Uber Analogy
Key Points
- The tech industry is pouring unprecedented amounts of capital into AI, yet there is still no clear model for how that spending will translate into sustainable revenue.
- The speaker likens the current AI hype to Uber’s early‑stage, heavily subsidized growth, noting that massive upfront investments can reshape consumer habits but may require years of higher pricing and ancillary services to become profitable.
- Analysts forecast AI‑related expenses could reach a trillion dollars in the coming years, implying companies will need $5‑10 trillion in AI‑generated revenue to achieve a typical 5‑10× return on investment.
- So far, the primary profit generators are “picks‑and‑shovels” firms like Nvidia, which sell the chips needed to train models, while downstream AI‑application companies have yet to demonstrate significant earnings.
- The growing “AI revenue gap” highlighted by firms such as Sequoia underscores a widening short‑term shortfall, raising the critical question of how long AI investments can be sustained before tangible profits emerge.
Full Transcript
# Monetizing AI: The Uber Analogy **Source:** [https://www.youtube.com/watch?v=Gb6FJWnnkLE](https://www.youtube.com/watch?v=Gb6FJWnnkLE) **Duration:** 00:14:16 ## Summary - The tech industry is pouring unprecedented amounts of capital into AI, yet there is still no clear model for how that spending will translate into sustainable revenue. - The speaker likens the current AI hype to Uber’s early‑stage, heavily subsidized growth, noting that massive upfront investments can reshape consumer habits but may require years of higher pricing and ancillary services to become profitable. - Analysts forecast AI‑related expenses could reach a trillion dollars in the coming years, implying companies will need $5‑10 trillion in AI‑generated revenue to achieve a typical 5‑10× return on investment. - So far, the primary profit generators are “picks‑and‑shovels” firms like Nvidia, which sell the chips needed to train models, while downstream AI‑application companies have yet to demonstrate significant earnings. - The growing “AI revenue gap” highlighted by firms such as Sequoia underscores a widening short‑term shortfall, raising the critical question of how long AI investments can be sustained before tangible profits emerge. ## Sections - [00:00:00](https://www.youtube.com/watch?v=Gb6FJWnnkLE&t=0s) **AI Monetization: Lessons from Uber** - The speaker warns that massive AI spending lacks a clear revenue model, compares it to Uber’s subsidized growth strategy, and suggests future profits will stem from price hikes and ancillary services like AI‑driven platforms. ## Full Transcript
I want to break down the problem we have
around how you monetize Ai and we're
going to do it in depth and the reason
we care is because there is nowhere that
we are spending more right now in Tech
than AI That's why Nvidia stock is going
through the
roof and the thing that I want to call
out is that we don't have a clear
picture of AI monetization we don't know
where AI money is going to come from we
have a lot of hopes and dreams
but it's not clear yet and I want to
actually paint the picture because I
have my gray hairs I earn them I want to
go back a decade or more into another
era of tech and I want to talk about how
this moment is reminding me a little bit
of uber and the hyper subsidized ride
hailing experience that we got in the
2010s so Uber's investors spent
approximately $30
billion over a 10-year period on
subsidizing the ride hailing habits of
Americans and they paid so much money
that they changed people's habits people
moved away from taxis and the BET was
once you change people's habits they
will stick with them even if you raise
prices and by and large taxis have not
come back and Uber is making due with a
combination of increased prices up like
100% in some places and with Uber Eats
which is uh very profitable for them and
actually they're their sort of biggest
driver of profit margin going
forward and I'm thinking about that
because at the end of the day what we
have with AI is similar except the
numbers are way way way bigger we're
talking about potentially a trillion
dollars in expenses over the next few
years uh at least according to a Goldman
Soxs research not that came out uh I
think it was this week last week
something like that and if you're going
to spend a trillion dollars on things
and you're looking for a conventional
five or 10x return on investment that
means that you need five to 10 trillion
dollar in revenue for the company's
investing in that Capital expense to
make this work and so far the only ones
who are making an enormous amount of
money on AI are companies like Nvidia
that are as we put it selling these
picks and shovels in the Gold Rush
they're selling chips that other
companies can use to train large
language models and that is moving right
like they are selling bucket loads they
can't keep them on the
shelves but if you go a level F farther
up in the stock and you look at what are
the companies that are using those chips
selling with AI you don't come up with a
lot yet and that is one of the biggest
question marks in t right now so for
example Sequoia treats this as a $600
billion Revenue Gap and they've updated
that that by a fair bit since just the
start of the year I think they they
doubled or tripled it uh this was
according to a research note they put
out asking where the revenue for AI was
going to come from and that was only
last month that Revenue Gap keeps
growing every month as companies invest
more in AI on the hope of future revenue
and so my question is the Uber question
how long will
investors tolerate this kind of capital
expenditure how long will the markets
tolerate if you're a publicly traded
company and where do you expect the
revenue to come from I want to put
forward three different options that I
think are historically plausible for
where AI could pull that revenue
from number one is a massive lift in
productivity and the reason I put that
at number one is because I think that
large language models most useful and
interesting use cases for lack of a
better term are
around how you can be creative we we
made these large language models with
the assumption that they would be
logical that was our default assumption
for artificial intelligence in movie
after movie after movie after movie The
Narrative in science fiction was that
way the way we talked about AI even
before 2020 was that way in TCH it's
just not that that way that large
language models that became popular
across the globe are highly creative we
invented creative AI
first and because we did that what we
are getting is an opportunity for
productivity growth that we are all
coming to terms with and so if you want
to look for like where is a place where
you could get a surprising amount of
money I think one of the
options
is breakthrough productivity driven by
AI
creativity and that is not going to look
like the efficiency gains that most
companies are banking on it's going to
be much more growing the top of the
funnel it's going to be much more
focused on how you can uh grow your
sales your Revenue how you can as a
company invent a new product I know that
people are using large language models
for research applications because open
AI has talked about it those are the
kinds of things that could yield
disproportionate benefits imagine the re
line impact if a large language model
leads to a breakthrough that unlocks a
new drug class right and now it's a
hundred billion $200 billion and there
you start to attack that Revenue line
right there that is not how Wall Street
typically thinks about the potential of
new technologies because Wall Street is
typically geared toward thinking about
it in terms of efficiency gains and so
one of the things that I think is an
inherent tension is that Wall Street is
not really buying this larger vision
that Sam Alman and others in the AI
futurist uh movement have embraced W
Street doesn't see the dollars and cents
adding up and Sam is basically saying
and his friends and and his colleagues
and the rest of folks in Tech who are
close to Ai and who believe in it are
saying just be patient we will get there
we will see what happens one of the
analogies that comes to mind for me as a
former Amazonian is Wall Street also
didn't buy Amazon web services for a
long time I Vivid remember picking up
the magazine where uh I think it was
Time Magazine and it was like Jeff you
should keep the store was the headline
and it was like an entire article
dedicated an entire issue dedicated to
this idea that Amazon was branching out
inappropriately from retail and should
not get into cloud computing and we all
know how that went
right at the end of the day AWS is doing
pretty well but the monetization wasn't
obvious similarly for prime
if you look at the dollars and cents
case for prime it does doesn't add up
it's actually a fairly well-known story
like if you look at whether or not it
made sense to launch Amazon Prime the
cost of two-day shipping way outweighed
the expected monetary value and what
people didn't realize was that by taking
away that mental block that came with
calculating shipping you were going to
unlock a ton of new Demand on the
internet for e-commerce and that's what
Amazon ended up doing uh it was a case
of uh what we call uh jv's par jens's
Paradox uh it's basically a case where
if you drop cost down you see demand go
way through the roof in this case Amazon
drops the cost of shipping they remove
this mental walk demand Rises so where
does that us at the end of the day AI is
sort of in the place where Amazon was
when it was deciding on Prime where
Amazon was deciding on AWS where Uber
was when they were trying to figure out
how you monetize off these cheap ridu
subsid AI as a collection of companies
as a movement as a vertical intact is
looking for
monetization they need it otherwise the
capital expenditure doesn't justify so
my first case was AI is creative maybe
we find ways with AI creativity like a
drug breakthrough or something like that
and we start to chip away that Revenue G
that my second is the efficiency gains
but
I want to ask where that comes from and
why because I think people tend to
assume efficiency gains mean job losses
and going back to Jen's Paradox I
actually don't think that's true here I
think efficiency gains are likely to
lead to more demand for the work that is
done efficiently and that's something
companies are missing right now and so I
think instead of looking at it as we can
get stuff done with lower costs yes
maybe the unit economics Dro maybe it
costs less to run a marketing campaign
but you can now do cooler marketing
campaigns and so maybe the dollar costs
don't drop and that's something that
again Wall Street may not be calculating
in as much like they tend to look at it
as very
linear and I think that's one of the
things that will be interesting to see
is do we see these anecdotal reports
where people are saying I am getting
more done but there is always more work
to do and therefore even though AI has
helped me a lot I'm just getting through
more of my infinite to-do list it's not
like the infinite to-do list has
actually gotten shorter and I think
anecdotally for a lot of us in Tech it
is an infinite to to-do list and there's
always things we could be doing and AI
is simply helping us to get to like 60%
of it versus 40% of it and so even if
the productivity gain is Big it may not
come through in like reduced salary
costs or layoffs which by the way is
something that is really interesting to
think about there's some anecdotal
evidence that that there are layoffs
associated with AI that are happening
they're mostly anticipatory it's mostly
around we think we can get away with
this and it's mostly associated with
companies that have been struggling
anyway and are kind of looking for
something that gives them a narrative to
turn the company around it's less about
hard productivity gains in practice
driving hard layoffs that one doesn't
happen as much at least not yet we will
see so to me that second one around
efficiency gains is actually more about
how can a company get more done cover
more territory given the same Workforce
for and we do see like I I see a lot
more conversation around people delaying
hiring or slowing on hiring because they
think they can get more done with the
same
number that one may not be true in
practice in the sense that people may be
overd delaying they may be taking that
productivity lift for granted too much
people may get tired they may burn out
they may switch rolls who knows but
there's more to it there than there is
is from a hard edged productivity cut
perspective so I want to pause there so
wrapping up the efficiency thing I think
there's two ways we play this right we
have hard Edge productivity Cuts uh and
job losses from AI I don't see much
evidence of that there's a few companies
that are doing that that are already in
trouble and they're anticipating I don't
think they're actually like facing it on
hard edged facts of productivity in
their company for the most part and
there's also the softer ones where
they're slower ring higher ing because
they see more efficiency and my general
Point here is that if we are getting to
more of our infinite to-do lists you're
not going to see that show up in the
dollars and cents of the company because
well we all have infinite to-do list
right even if we're seeing huge
efficiency gains so what is the third
way we could monetize I talked about the
massive lift of uh creativity the
example was what if we find a new drug I
talked about efficiency gains a couple
of ways we can play that and the last
one I want to talk about is New Markets
and new devices one of the things that's
really interesting is that we've mostly
interacted with large language models
through a chatot experience well what if
it's not just a chat imagine a world
where you start to pair a large language
model with an inhome household robot
that is a device class that doesn't
exist yet that is a device class that
people have speculated may be worthwhile
since the 50s and we certainly have
well-known entrepreneurs going after
that device class right now building it
right now with the explicit goal of
inventing a new home device that people
will pay a lot of money for because it's
such a labor saver around the
home I don't know the future we will
have to see what happens but one of the
ways you pay off on llms is if suddenly
they are effectively the operating
system that powers the most ubiquitous
roll out of a new device since the
iPhone and then layer on top of that
what happens if they're also powering
iPhones essentially because Apple's in
the middle of powering up their iPhones
to do AI it was a big factor in
WWDC and they may be more cautious than
many but they're still going after so I
want to wrap this up and give you a
sense of how I feel about it and where I
land on this I agree we have a big
Revenue gap on AI it is not trivial I
also agree that we are spending at a
massive rate and the people who are
winning right now in this space are the
people who make chips which is mostly
Nvidia we do have monetization that we
need to find if we are proponents of AI
if we feel like there's value
there I think that the three options we
have are essentially find ways where we
can be more efficient and I don't see
that as equivalent to job losses find
ways to use lm's creativity to find
breakthroughs and unlock
disproportionate gains and I see a play
for devices as well does that add up to
5 to 10 trillion
I don't know this isn't a math podcast
we will have to see how it plays out but
if I were to pull on those threads and
say where is the use case that companies
who are investing hard money are
anticipating those are the three that I
would go after tell me what did I miss