Subscribe
Share
Share
Embed
Around the Desk: This is where we at Avory think out loud, challenge narratives, and look for signal through the noise. Each episode, the Avory & Co. team dives into what’s moving markets, how companies are performing, and where opportunities may be forming. We break down earnings, macro trends, and investor sentiment — all from the lens of a concentrated, high-conviction portfolio.
*** The views expressed on Avory Podcast: Around the Desk are those of the hosts and guests and do not constitute investment advice. This podcast is for informational purposes only and should not be relied upon to make investment decisions. All investments involve risk, including potential loss of capital. Avory & Co. may hold positions in the companies discussed.
Speaker: All right, you are listening
to Avery's Around the Desk podcast,
where we dig into markets, companies,
and ideas shaping what is coming next.
I am Sean Emery, founder
and chief investment officer
here at Avery and Company.
You know, here we manage, you know,
various, uh, equity strategies, uh,
in-house, uh, really focused on,
you know, concentrated portfolios
that focus on where the world is
headed, but doing so with discipline.
You know, today I wanted to talk
about momentum, not just, you
know, stock market momentum, but,
you know, narrative momentum.
And narratives have, you know,
essentially been at the forefront
really the last several years.
You know, increasingly it feels like
we're in a market where capital is
flowing towards certainty that I don't
think, uh, you know, really exists yet.
You know, this conversation, you
know, before I get into anything,
is for informational purposes only,
should not be considered investment
advice, and we may hold positions
in some of the companies discussed.
Please do your own research before
making any investment decisions.
So let's get to it.
You know, look, over the last
two and a half years, one trade
has dominated almost everything.
It's kinda been like AI on, AI off,
uh, you know, really, you know,
predicated on the fact that we've
had multiple policy-induced issues,
tariffs, war, inflation, all kinda in
this flywheel that, uh, has, has been
in, uh, existence really the last,
like, two and a half years, and it's
made it, you know, difficult for,
you know, disciplined fundamental
investors that aren't chasing momentum.
You know, before, uh, anyone, I
guess, like, misunderstands that, you
know, AI is at-- is central to this,
you know, we're coming from a place
where we're incredibly bullish on AI.
We write a lot about it.
We know it inherently well.
I'd say we probably use
AI more than most people.
You know, it's embedded in many of the
things we do on the operational side.
You know, it's changing, you
know, how we should think.
It's changing how we should
potentially research.
You know, really focus definitely on,
you know, your old school fundamental
research, but still at the same
time, you know, it helps and assists
and increase productivity across
the board in areas and surfaces
probably we weren't able to do before.
Um, you know, it's changing
how, how we communicate.
It's thinking how businesses should
operate, and I think the opportunity
around AI is, is frankly enormous.
Again, we've wrote articles about this.
You know, the question
isn't whether AI matters.
I think the question is really where
the durable value ultimately ends up.
And again, um, keyword there is
durable Um, that's where I think,
uh, you know, things get interesting
because when I, when I look around
the market today, I see investors
making, you know, very large bets on
outcomes that are far from certain.
You know, you think of the infrastructure
build-out, the power build-out, the
data center build-out, the model
layer, you know, the energy layer.
Um, what else?
Yeah, you know, the hardware layer.
You know, all of these are receiving
massive amounts of capital because, you
know, the market is assuming today's
winners will remain tomorrow's winners.
Uh, maybe they will, you know, maybe
they won't, but the durability is
far less clear than current, you
know, valuations in those spaces,
uh, you know, suggest, I think.
You know, take the model
layer, for example.
You know, one assumption, uh, embedded
in many, uh, AI discussions is that,
you know, frontier models will maintain
a large, uh, advantage indefinitely.
Uh, this is why they're
getting, uh, you know, roughly
a trillion-dollar valuations.
You know, maybe that's true, but, you
know, if we're still or, you know,
three, two, four years out away from,
you know, what some would consider true
AGI, which is what DeepMind and Google,
you know, has continued to suggest.
Anthropic, uh, you know, Dario has
also, you know, mentioned that If we're
truly, uh, you know, two to three years
away from AGI, doesn't that also imply,
you know, more time or some time for
then open source to catch up there?
Uh, right now, I think we're
roughly like six months away.
You know, some with the new Fable launch,
which is, you know, kind of a, a fork from
Mythos, uh, uh, out of the Anthropic lab.
You know, the idea is that, you
know, po- potentially we're twelve
months out, um, in terms of open
source catching up to closed source.
But we're already seeing signs that I
think AI labs are concerned about this.
You know, they're spending so much time
here recently preventing distillation.
Uh, and if that's not a threat, then why
are we-- they spending so much time there?
And again, for those that don't know,
distillation is essentially, you know,
taking the existing models, essentially
querying them in a very, you know, basic
way of framing it, querying them, uh,
getting all the inputs, outputs, uh, and
doing so and creating a model far cheaper,
uh, than the original trained model,
uh, in these multi, you know, gigawatt
clusters, let's say, or megawatt clusters.
Um, so really, why create barriers around
model replication if, if that gap, um,
is, you know, somewhat impossible to close
or, you know, the perception is that,
you know, you can close these things?
So, you know, I, I just see
what I see, and I think that's
ultimately, you know, a signal there.
You know, the reality is, is that
technology has a long history of
becoming more accessible over time.
Companies will try to slow
that process down, obviously.
You know, but stopping it, uh,
entirely is another matter.
So eventually, I-- you know, it's, it's
pretty clear that many models will begin
to look, you know, more and more similar,
uh, as, as we, we get closer and closer to
kind of a model that can do many things.
Um, I think performance differences,
you know, will start to narrow.
I think costs will continue to decline.
Capabilities will, will begin
to converge more and more,
and we're already seeing that.
And when that happens, I think the
model itself becomes less valuable
than, than what sits around the model.
Again, almost like the internet,
you know, as that continued
to, uh, develop and form.
Think of, you know,
telecom carriers today.
Uh, there's other, you know, examples
and analogies that you can use where,
you know, the central technology that
was developed, you know, became kind of
in the background, and then it's, you
know, what do you build on top of that?
Um, you know, side note, obviously, Cash
App just launched a, uh, you know, a,
a mobile service, you know, for, for,
you know, your, your mobile device,
uh, and it's built on, you know, AT&T.
So, you know, that's a good example
of, you know, the, the communication
layer being somewhat commoditized and
then everyone building around that.
And then, you know, who would have
thought that a payment company
would have a, um, you know, carrier
product, uh, attached to it as well?
You know, but w- much like we think about
the current state of technology and you
go back to cloud, you know, we think most
enterprises eventually operate in kind
of a multi-model world, um, much like
we, we sit in a multi-cloud world today.
Uh, and, and not even just cloud,
uh, in the public cloud, but also
private cloud, so inside, uh,
private data centers as well.
You know, most people thought that would
be obsolete, but we're sitting here with
companies not exclusively using AWS.
N-no one's exclusively or s-
many aren't exclusively using
Microsoft Azure or Google Cloud
or Oracle Cloud or Alibaba Cloud.
You know, the list goes on.
Um, you know, the best tool, I think,
for a specific task is essentially,
you know, the protocol here.
You know, we think AI likely
evolves similarly there.
You know, one model could be, you
know, best for reasoning, another for
coding, another for image generation.
You know, uh, another can just be really
around, you know, task usage and, and
making sure it's dramatically cheaper.
Uh, but I don't think the
winner will be one model.
Uh, the winner will likely be an
ecosystem built around them, and
that's important because I think
that signals that most of the value
created, uh, will be, uh, far further
than just the infrastructure layer.
And so far, you know, most of the value
is accrued to power, semiconductors,
networking, compute, data centers,
really the picks and shovels and,
you know, that narrative goes--
gets thrown around and around.
Uh, and if you look at many of the AI
application companies this year, you know,
performance has been actually very weak.
So again, infrastructure has gotten
all of the, the, the, the capital,
all of the, the, the, the, the money,
um, uh, all of the attention and
narrative, you know, for good reasons.
Clearly, it needs to be built out,
but nobody knows the depth, um, and
whether people are actually underwriting
these things for compounders over
the next, you know, several decades
versus just trying to ride a wave.
Um, and in general, look, the capital
continues flowing towards this
infrastructure stack because these
investors view it as, you know,
the most direct way to participate.
I think that does make sense, but
that concentration itself creates,
you know, some layer of risk.
Um, because again, you don't
know fully what the long-term
economics look like here.
I don't think anyone with a, with
a You know, a real conscious could
say that, you know, the build-out is
happening at extraordinary speeds.
You know, debt is being issued
like we've never, uh, seen before,
specifically in some of the companies
that you wouldn't have thought of.
Billions are being
committed, uh, outright.
You know, depreciation schedules are
being stretched and, or narrowed,
depending on who you're looking at.
Yet technology overall, uh, is
moving so quickly that it's difficult
to know what remains valuable,
you know, five years from now.
You know, look at energy.
Energy has become one of the
biggest bottlenecks in AI.
But at the same time, some of the smartest
people in the world are trying to solve
that bottleneck, you know, outright.
You think of, you know, nuclear, you know,
renewables, grid optimization, you know,
some of the advanced cooling we're seeing.
You know, new chip architectures.
Uh, even, you know, discussions around,
you know, space-based infrastructure,
which obviously SpaceX is getting, you
know, a lot of that attention here.
Um, and in general, a lot of
that is, you know, exciting.
It's very exciting.
Um, but at the same time, you
know, I don't think most people
can truly underwrite this stuff.
And when you're a concentrated manager,
you need to be certain, more certain,
higher probability outcomes versus lower
probability outcomes and spray and pray.
And I think that's the challenge in
general today's constraints don't
necessarily equate, uh, to, you know,
tomorrow's constraints, uh, at all.
Um, so, you know, the question though
then comes back to What is durable,
but not what is just exciting, uh,
you know, not what is just popular,
um, not what, what's working, you
know, in the cycle, let's say.
Um, uh, it's really what is durable
because as investors, again,
you're not trying to surf the waves
and jump off before they crash.
I don't think that's the, uh, the mindset.
I think, uh, ultimately that is trading
and not, you know, truly investing.
Um, we're trying toâ¦
And I was on a podcast the other day,
you know, we're really trying to,
uh, identify businesses whose value
compounds, you know, really regardless
of where the cycle in AI goes from
here, and that's where names like, you
know, a Zoom become interesting to us.
Obviously, it's our top
holding, so full disclosure.
You know, most people think about Zoom
as a, you know, video conferencing
company, but the story today is
much larger than that, much more
interesting, to be quite frank.
You know, we just put out a
piece that shared, you know,
the breakdown of Zoom, right?
And, you know, I think the summary
there was it's trading at, you know,
five times cash flow if you back out
their, their stake in Anthropic, right?
They have about a, you know,
l-latest was around a two billion
dollar stake, uh, in Anthropic.
That was at the round, uh, that was three
hundred and roughly sixty billion dollars.
We know their latest round was three
X that, so we can take the two billion
and multiply that by some factor,
and you get probably closer to f-
you know, four, five, six billion
dollars depending on dilution.
Then we know they're
gonna go public this year.
So if they go public this year, then
ultimately you have likely a higher
than a trillion dollar valuation.
So there, you know, there's a
reference point there where, you
know, Zoom's stake in Anthropic is
somewhere around, you know, five,
six, ten, eleven billion dollars.
Again, depending on dilution and, and
whatever, you know, price they go public.
But ultimately, you take that plus
the cash, they have eight billion
dollars of cash and zero debt,
a market cap, you know, closer,
you know, mid-twenty billions.
You're talking about almost sixty percent
of the value of this business being, uh,
consumed by cash and a stake in Anthropic.
Uh, and then you have a business where
their revenue is re-accelerating, and more
importantly, uh, a company with no debt.
And perhaps the most, uh, important
aspect here is an enormous
amount of communication data
flowing through its ecosystem.
So when I think of AI,
you think of context.
Context m-meaning meetings,
messages, contacts, phone
calls, workflows, documents.
Uh, what else?
Like just conversations
that everyone's having.
And if AI runs on context, Zoom
possesses, you know, arguably one
of the most, uh, you know, largest
troves of, of context to do that.
So that doesn't automatically make them a
winner per se, uh, specifically within AI.
But it does create a foundation that is
increasingly valuable in an AI-driven
world, and then a, a business, a core
business that isn't, again, uh, dependent
on AI, uh, but benefits from AI, has
a large stake in AI, and therefore,
you know, you could argue a direct and
indirect play into the AI ecosystem that's
not being, uh, accredited, uh, for there.
Uh, and so that's what
makes it interesting.
Another one is a company like Roblox,
a, a new investment of ours, but
one that's been pretty volatile.
It's one of the largest
consumer engagement platforms
on the internet, you know.
The-- We ultimately think AI
ultimately makes creation easier.
And if creation becomes easier, more
developers can build experiences.
And if more experiences are built,
uh, you know, engagement rises.
And then again, if engagement
rises, monetization improves, and
that flywheel becomes stronger.
To me, that makes a lot of sense.
Um, you know, there's some things going
on there with, uh, age verification.
Uh, and I think ultimately that
proves to be a positive given
verified users are much more
valuable users than unverified users.
Um, and, you know, this doesn't
require economic backdrop.
This doesn't require,
you know, AI or non-AI.
Uh, it benefits from it.
Uh, but the core business is, um,
one of the best, uh, communication
engagement platforms out there.
Um, then there's the
macro backdrop itself.
And over the last several years, we've
repeatedly seen markets, you know, begin
broadening beyond AI and technology.
Then uncertainty emerges.
Uh, first it was like, you
know, inflation in '22.
You know, obviously that led
to rates rising, you know,
people questioning the economy.
Uh, we, you know, we were out
there suggesting the consumer
and the economy was fine.
That proved to be accurate.
Then you had tariffs, you know,
another, uh, situation which led to
fears around the economy once again,
fears around inflation once again,
which, uh, you know, our view was,
you know, pretty straightforward and
documented, you know, everywhere.
Uh, and again, I think we were fairly
accurate on that, that the economy
and consumer was strong enough.
Um, all of that, you know, and then you
have the, the, uh, the war here, and,
uh, again, same, uh, output, different
inputs, and, you know, same reaction.
So investors become
concerned about inflation.
Inflation creates uncertainty
around growth, and then capital
moves back and forth between
what they perceive as safety.
And safety is, you know, safety in
the quality of the assets, let's
say, but also more importantly
is like, what has momentum today?
And there, you know, there's
a stat going around where
we're in like the ninety-ninth
percentile of the momentum trade.
And so at some point that will
break, and, uh, I think, you know,
peace deals and some of those other
things will be the catalyst there.
And assuming we don't get any geopolitical
tensions thereafter, you know, it
could set up for, uh, again, the trade
that we've been, you know, honestly
sitting on for the last two years.
Um, you know, you have the consumer.
Consumers continue to spend.
Employment remains, you
know, fairly healthy.
Travel remains fairly healthy.
Experiences, you know, continue to, you
know, produce when you look at, you know,
the earnings of many of these companies.
You know, but the markets don't trade
on today's reality alone, let's say.
Um, they're trading on, you know,
what uncertainty potentially could
happen, even though, you know, uh, I
think a lot of that, again, is well
documented on our outside You know, at
some point, uh, what's interesting as
well is, uh, many of these AI, private
AI companies will also become public.
And when they do, I think investors
will be forced to answer questions
that I think are easy to ignore today.
You know, how profitable
are these businesses?
What are the long-term
margins of these businesses?
How much c- ongoing capital is required?
How much pricing power actually exists?
If profits are strong, you know,
perhaps the economics justify
current investment levels.
But, you know, if profits are weak,
you know, which is likely what they're
going to show, perhaps, uh, it means
models are becoming commoditized.
Um, look, if prices fall or have to
fall to, to drive adoption, you know,
perhap- that tells us something, uh,
you know, about commoditization here.
And those questions, uh, you know,
that public markets eventually force
companies to answer, you know, are you
profitable, and can you do it sustainably?
And that's why we remain patient here.
You know, we're excited about AI.
We believe AI will transform
nearly every industry.
You know, we believe, uh,
adoption is still early.
Um, we, we also believe that
optimization is very early as well.
You know, everyone is, uh, you know, token
maxing they call it, and using, you know,
AI sporadically without understanding, you
know, how much it costs or where it costs
and, you know, how to apply it really.
And so that will come down and optimize
eventually, but you'll have it move
into the economy at the same time.
So likely that's bigger than
the optimization effect.
But look, a lot of excitement, a lot of
valuations, you know, fairly excessive in
many of those areas Uh, plenty of areas,
you know, that are trading at, you know,
thirteen, ten, twelve, fifteen times
earnings, um, with, you know, double-digit
growth, which makes those interesting,
but they're not getting the love.
Um, and again, everything I just mentioned
I think is, uh, explanatory for that.
Um, you know, when we look around
today, again, we see a lot of momentum.
We see a lot of capital.
Uh, most importantly is we see a lot
of narratives, uh, that people are,
are, are chasing and, and, and, you
know, um, surfing on, let's say.
Uh, and what we're trying to find
are those businesses that still have,
um, you know, durable advantages,
make sense after, you know, some of
the excitement fades, uh, and the
businesses that don't require, you
know, perfect assumptions, let's say,
for the valuations to make sense.
You know, the businesses that benefit
regardless of, you know, which
hardware wins, which model wins.
You know, the businesses that benefit
regardless of, you know, which energy
solution, you know, comes out on top.
You know, if models get commoditized,
you know they win, if anything, you
know, more so than not because the cost
of, you know, AI generation becomes,
you know, ultimately the Internet.
And, uh, you, you essentially
want to feed that with context,
again, using the Zoom example.
And, you know, eventually momentum
slows, narratives change, and
then those cycles, you know, turn.
And then economic reality,
I think, reasserts itself.
And so, you know, obviously we're
hoping for a peace deal here.
Uh, we think, uh, clearly there's,
uh, incentive to do so, so that, uh,
you know, energy prices don't stay,
uh, elevated prolonged, uh, basis.
And historically, that's where some of
the best, uh, investment opportunities
emerge is, you know, staying patient.
That's what we're doing, uh, very
much aligned with our investor base.
And, uh, you know, we're just sharing
some of the context around, you know,
what's happening in the markets, um,
you know, where there's, you know, areas
of opportunities, where there's not.
Uh, and that's what we're
spending most of our time doing.
Uh, that's it for this week,
uh, for Around the Desk.
We will be back next week.
Every Friday we're doing
this, uh, so stay tuned.