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.
All right.
You're listening to Avery's around
the Desk podcast, where we dig
into markets, companies, and
ideas shaping what is coming next.
I'm Sean Emery, founder and chief
investment officer here at Avery and co.
Today's episode is about the state of
AI through the lens of, you know, big
tech earnings that we learned this week.
You know, Google, Microsoft,
Amazon Meta, and maybe a couple
others that I'll mention here.
You know, this conversation is
for informational purposes only.
Should not be considered
investment advice.
We may hold positions in some
of the companies discussed.
Do your own research before
making any investment decisions.
You know, this, uh, what I wanted to do
today is, you know, step back from the
noise, focus on what actually matters,
you know, capital monetization defense
versus offense, specifically around ai.
And when all of this says about
the underlying economy, you
know, before we get into it.
You know, here's what's
going to be covered today.
Uh, you know, number one is
really, you know, the size and
scope of the AI CapEx build out.
You know, looking at big tech earnings,
who has the lowest hanging fruit?
To monetize ai.
Why some of this spending is defensive,
some of it is also offensive.
So breaking that out slightly
here, why the software panic
tied to AI is likely overdone.
Uh, and then what these companies
behavior, I think says about the economy.
I think all of this is the
biggest takeaways for this week.
You know, the CapEx, you
know, build is signal.
And so that's, uh, you know, step
one here, which is, you know, let's
start with the big picture, you know,
across Google, Microsoft, Amazon Meta.
You know, we're looking at
roughly, you know, $600 billion
of capital being put to work.
I think that matters more than any
single AI feature announcement.
Any headline, anything about, you know,
software, this, software that, you know,
some of that spend, I think is defensive.
You know, protecting moats, maintaining,
you know, relevance, uh, in the eyes of
developers, in the eyes of, you know,
their consumers, and then ultimately
keeping up with competitors as well.
Uh, I think a meaningful portion of this
is, is definitely offensive, just given
you know, what the capabilities are,
uh, and really trying to build capacity.
Uh, ahead of demand.
But, you know, ultimately a lot of
these players that I just mentioned
are utilizing this, uh, ai CapEx
build out for themselves, but also,
you know, reselling it to others.
Here's the point, you know,
you don't need, um, to do this.
You know, you don't deploy
capital at this scale.
If you have.
I think a negative view of the current
state of the underlying economy.
I think, you know, remember that in the
back of your brain when you think about
all this CapEx build out, you know,
these companies I think sit on the best
real time data in the world, better
than any government data out there.
You know, they're running
ads, they see cloud usage.
They, they, they sell in commerce,
uh, they sell into enterprise for.
Underlying software demand.
So I think their willingness to spend
is probably the ultimate signal that I
think anyone that is focused on what is
the true health of this economy, uh, I
think that is the ultimate signal that
conditions underneath are rather healthy.
So again, takeaway there on, you
know, that that section really
is CapEx is not just an AI story.
It's, you know, I think it's an
economic confidence signal that
we should all be listening to.
You know, second part I think
is Amazon, Microsoft Meta, you
know, benefit very differently.
And so I wanted to break
that out a little bit here.
You know, these companies are spending
heavily, uh, but they're not, I don't
think playing all of the same games yet.
Many of these are getting
lumped in the same category.
Um, you know, Amazon is an
infrastructure company at its heart.
You know, when I think back
infrastructure for commerce, when
I think back infrastructure for
technology, AI fits naturally into
that DNA, you know, monetization.
Uh, will likely be gradual, uh, here in
terms of, uh, you know, true ROI spend.
But, you know, I think you have
to build first ramp utilization
thereafter, but they have so many
customers on the enterprise side.
They have so many, uh, ways for
them to monetize it themselves.
Uh, Microsoft.
Yeah, arguably the number
one software company.
Yes.
Software company.
Uh, so I think that means they benefit
from AI, but also very exposed to ai.
I think disruption risk is real.
You know, if you look at Word, Excel,
PowerPoint, I think as workflows change,
you know, those could be impacted.
I about you.
But you know, I don't touch
really Microsoft Word anymore.
You know, if you're starting a
writing, you're usually starting it.
In, you know, something else, whether it's
notion, uh, whether it's something else.
You know, a lot of people are
starting it in their, you know,
preferred AI apps as well.
Um, so for Microsoft, you know, I still
think they control many of the most
important parts of the enterprise,
which are security, authentication,
governance, distribution, all of
that will matter in the world of ai.
Uh, and therefore maybe the
productivity tools that they're,
once, you know, their dominant.
Uh, positioning could actually
become, you know, um, obsolete
and therefore, or enhanced, right?
You know, there's, there's many
reasons why, you know, some of
those tools could be enhanced.
They have time.
I think that's the critical point here.
Um, meta we own it.
Um, I think we think at Avery Meta is the
cleanest winner of ai and I think that is.
Been a contrarian view.
I think it's been a contrarian
view that has been proven right.
Um, if you think about it,
their product is already ai.
It's been ai.
They have recommendation engines,
they have personalization engines.
They have a ma, a massive data set
that, um, is feeding real time.
Data back.
And so you have these feedback loops that
then, you know, point to the advertisers.
So ultimately you have the product
itself gets better enhanced by ai.
Uh, you and I don't feel it, you
know, we're scrolling, we're swiping,
we're messaging, but yet everything
that is being put in front of us is
driven by some sort of engine that
is AI enabled, that is reading.
Um, and.
You know, understanding us, uh,
very much as if we're thinking
about how chat GBT or many of
those solutions, uh, understand us.
So I think that's an important
thing to think about.
You know, a AI is really
accelerating that existing engine.
I don't think disrupting it.
And therefore the CapEx spend there
is going directly into their product,
along with, you know, other, you know,
potential solutions that potentially.
Meta could get into, um, Google.
You know, they sit somewhere in between.
There's narrative risk around
search, but I think the data matters.
Data always matters.
And what does the data say?
You know, Google search is growing.
18%, 18%.
Google search, that's the
fastest growth rate since 2022.
I don't know about you, but people were
saying that Google search was potentially
the most, uh, in the way of ai just given
that we were all searching differently.
And yet Google search strongest
growth since 2022, which makes
no sense to the narrative.
Um, so look hard to argue that
traditional search is, uh, you know,
not in the way, you know logically,
but I think when you look at the,
the data, it's not necessarily,
um, uh, proven to be accurate.
If anything, Google search is evolving.
You have AI mode and AI overviews sitting
around, you know, the top of the search.
So.
Evolution that is taking place.
Um, and people are still searching for
stuff and they're monetizing that search.
So, you know, again, um, sometimes you
have to look at what's, what's what
the data says versus, uh, you know,
how people are feeling emotionally.
Um, so the takeaway really on that is,
you know, same CapEx, very different
monetization paths for many of these.
And risk profiles just, just,
you know, understand these
businesses as opposed to, um.
Um, you know, anything differently?
Uh, the next section I think is,
you know, software or sell off.
Is it overdone?
Is it not?
You know, I think one of the biggest
market reactions we've seen so far,
uh, is that fear of AI breaks software.
I don't think that's happening, you know?
Truly, um, you know, there's a sentiment
gap, you know, for casual users.
I think AI feels brand new, meaning
a new announcement of, you know,
cloud cowork feels brand new.
But I think for, you know, developers
and engineers, you know, some of these
capabilities, uh, are not new at all.
You know, many of these existing
tools have existed in different
forms for I'd say over a year.
Um, you know, we use them.
Internally, uh, you know, there's
a, you know, product called Lindy AI
that has many of the, um, capabilities
or, you know, the input outputs
as, you know, the cloud cowork.
Meaning if I'm trying to spin up
agents and have it, you know, go, go
across, you know, various applications
and, and do things, you know, whether
it's, you know, at that moment in
time or, um, scheduling, uh, you know,
different tasks that can be done.
That has already existed.
It's just I think the casual
observer of AI is not paying
attention, and that's likely why,
you know, we're still seeing, um.
You know, many of these
software companies do well.
Uh, what's changing, I think
is again, accessibility.
People are, are paying attention to it,
not capabilities of these, of these stuff.
Obviously we are progressing.
These, these models are progressing.
They're doing more.
You know, you have, you know, some of
the, uh, the announcements on the gaming
side with Genie three from Google, which
was very, very, uh, I, I would say that
is more new than, you know, some of the,
the cowork stuff that, you know, came out.
Separating those two things, you
know, some of the more frontier, uh,
models that have come out versus,
you know, just an announcement from
a company that you know is worth, you
know, now, you know, valuations of 350
billion for, you know, uh, anthropic.
So, uh, I think it's a little bit
of a headline versus reality there.
Um, the thing about software though is
I think, um, what people are missing
is there's this enterprise reality,
you know, ag agentic platforms.
What they do is they really
introduce real security challenges.
You know, prompt injections
is a good example.
Um, and if you, if you don't really
understand prompt injections, I think
you're, you're pretty far off from
what's, uh, what's happening in ai.
Um, you know, these are like, you know,
I, I would share, share a risk, right?
So hidden instructions, you
know, sitting and embedded inside
of like an email or an image.
That an AI can read, but humans cannot.
Um, you know, that is an example of the
risk of some of these cowork products
where you set it up to have it, you
know, summarize your emails in the
morning and ultimately hackers and
security threats or understand this,
that you're scanning all your emails.
So why not we send an email that
has hidden text inside those emails
that the email AI engine will
read and then inject that prompt
into the ai, and then eventually.
You know, do things that are malicious.
So, um, this is real.
I think that's why governance,
security, compliance all matters.
That's also why I think the, the rollout
will continue to progress slowly and fast.
You know, it all feels fast
and slow all at the same time.
I know that doesn't make conceptually
a lot of sense, but, uh, you know,
we're in year, you know, three and
a half, you know, since chat GPT has
launched and I'd argue many of the
real workflows that people are doing
today in, in the enterprise are vastly.
Um, the same, you know, mostly the
same than they were three years ago.
There's productivity here
and there, uh, for sure.
Uh, and there's disruption happening,
but ultimately, I think, um, it's a
good way to think of it like that,
where there's, you know, disruption
but controlled, uh, disruption here.
Number four, you know, platforms that win.
Uh, tools, um, that could
potentially struggle.
I think this is the best way to
frame some of the software stuff.
You know, not all software
benefits equally from ai.
That's very true.
So it's, you know, my last
section, I don't think the point
was to say that software's okay.
I think, you know, there are places
where you want to be and there are
definitely places you don't want to be.
You know, single purpose
tools are more exposed.
You know, that's, um, you know, our take
platforms, I think with deep proprietary
data are likely strengthened here.
Uh, so I think, you
know, that's the winner.
That's the loser.
You know, I think it's almost that simple.
We've asked executives, I've asked
developers a simple question,
you know, how likely are you to
build your own CRM internally?
How, you know, how likely are you
to build your own payroll system?
I just spoke to somebody today.
I asked the same question.
Are, are you going to build
your own accounting software for
yourself instead of using, you know,
QuickBooks or something like that?
And the answer is always, almost no.
And I think that is, um,
uh, you know, a big deal.
You know, teams want to focus on serving,
uh, their customers, uh, as opposed
to rebuilding, uh, infrastructure.
You are a company, you are
selling to your customers, right?
Your customers are looking
for a service from you.
So are you going to reorient your,
your, your organization to also focus on
building and maintaining and governing
and securing your software stack?
Um, so far speaking to people that
answer is, you know, unequivocally, no.
Um, will there be points in time?
Again, it goes back to are
you a single purpose tool?
Are you a form builder?
Yes, that could be spun up with, um, uh,
you know, rep lit or anything like that.
Um, but, you know, a complex,
multidimensional platform that covers, you
know, your communication, your developer
ecosystem, uh, and everything within
that I think, uh, will be a little bit.
Not a little bit much harder to, to see
a, a, a world where, you know, teams are
completely, uh, replacing that in-house.
Um, you know, so I, that,
that's kinda my take on there.
And also, the, the results that we're
seeing, um, are echoing this, right?
If you look at Atlassian this week,
you know, they, they signed, you
know, strong multi-year deals.
Um.
You saw someone like Paylocity,
um, that saw solid demand,
they beat their guidance.
This is a payroll company.
They beat their guidance and raise
the guidance by more than the beat
signaling, you know, really that,
you know, things are favorable.
That's a, that's also like a
double, that double whammy there.
On Paylocity, you have, um, the, uh, you
know, labor side of that, meaning they're.
A software for labor and they say, you
know, they see things as stable, so
take the economic, uh, you know, uh,
data point from there, but also the
software side of it where, uh, they're
seeing more and more attach rates.
You saw ServiceNow a couple weeks ago.
You know, uh, highlight that they're
seeing improving renewal rates and
signing, you know, multimillion
dollar customers for multiple years.
These customers already know
that AI exists and they're
deciding, Hey, you know what?
Let's, let's sign up, uh, you know, for a
three, five year contract with ServiceNow.
So again, there's a lot of, um.
So there's a lot of reasons, uh,
to believe that, uh, we're seeing
an evolution, uh, and single tool
providers I think are at risk.
But, you know, platforms, uh,
are certainly in a position
to take advantage of, of this.
So AI reinforces that view, that
platforms, um, will, you know.
Have the potential to do well and
not necessarily eliminate them.
Again, it all comes down to execution
like anything else, uh, you know,
last section here is really, you
know, where this ultimately goes.
I do think AI becomes embedded
inside of these platforms.
Users spin up, custom dashboards,
internal tools, lightweight
workflows all faster than before.
And, uh, I still think oversight matters.
I think product direction still matters.
I think human judgment still matters.
So AI is largely an expression of,
you know, the past, the past patterns,
the past history, the past data.
Therefore, it's very hard to, you know.
Run your own software, build your
own software when the software's
gonna be created using historical
reference points of, you know, how
that software worked versus, I think
we all sign up to, you know, software
providers that have a product roadmap.
Some of 'em tell you what
it is, some of them don't.
And therefore, you know, it's their job
and journey to build, you know, new tools,
new offerings, you know, new, um, um.
You know, different types of solutions
for their end customers and not having
the customers do it for themselves.
So, uh, I think that's, that's an
interesting way to think about,
you know, why it would be hard to
build your own software internally.
Um, so that's really it.
So when, you know, again, stepping
back, looking at big tech, AI and
software, I don't see a collapse here.
You know, if anything, you know, we're
sitting here at a pretty interesting
point, uh, from a technical perspective, I
think this is a transition and evolution.
Things will get smacked for sure.
Uh, but ultimately, you know, I think,
uh, massive capital's being deployed.
Uh, you know, we, we had mentioned in
the past that, you know, that was a
risk from the eyes of, um, the markets.
These companies are now going from
asset light to asset heavy on one side.
That likely means they're more entrenched.
They're, they're, they're building much
more of a physical moat as opposed to
a, you know, software centric moat.
But at the same time, what
does that do to cash flow?
So it's that balance, you know, mentally,
I think, um, we're seeing winners.
Or, uh, non losers start to emerge.
I think maybe that's the
best way to frame it.
Um, and then, you know, here's a
market pricing fear faster than, you
know, the fundamentals are changing
and, you know, um, you, you have
to have a point of view on that.
And our point of view is
everything I, I mentioned above.
Um, so that's how I'm
thinking about AI right now.
That's how I'm thinking about, you know,
the earning season so far from some of
the big caps, uh, what it says about the
economy, uh, and what it means for, you
know, potentially software going forward.
So that's it.
Thanks for listening to Around the Desk.
We'll see you next time.