TRANSCRIPT

Q3 2025 — Follow-up call transcript

Q3 2025PDF · 10 pages96 KB
1
Meta Platforms, Inc. (META)
Third Quarter 2025 Results Follow Up Call
October 29th, 2025
Operator: Good afternoon. My name is Krista, and I will be your conference operator
today. At this time, I would like to welcome everyone to Meta’s Third Quarter
Results Follow-Up Q&A Call. All lines have been placed on mute to prevent any
background noise.
After the speaker’s remarks, there will be a question and answer session. If you
would like to ask a question, please press star one on your telephone keypad.
To withdraw your question, again press star one. We ask that you limit yourself
to one question, and this call will be recorded. Thank you very much. Kenneth
Dorell, Meta’s Director of Investor Relations, you may begin.
Kenneth Dorell: Thank you. Good afternoon, and welcome to the Follow-Up Q&A call. With me
on today’s call are Susan Li, CFO, and Chad Heaton, VP of Finance.
Our remarks today will include forward-looking statements, which are based on
assumptions as of today. Actual results may differ materially as a result of
various factors, including those set forth in today’s earnings press release and
in our quarterly report on Form 10-Q filed with the SEC. We undertake no
obligation to update any forward-looking statement.
During this call, we will present both GAAP and certain non-GAAP financial
measures. A reconciliation of GAAP to non-GAAP measures is included in
today’s earnings press release. The earnings press release and accompanying
investor presentation are available on our website at investor.atmeta.com. And
now, I’d like to turn the call back over to Krista for the first question.
Operator: Thank you. We will now open the lines for a question and answer session. To
ask a question, please press star one on your touchtone phone. To withdraw
your question, again press star one. We ask that you limit yourself to one
question. Please pick up your handset before asking your question to ensure
clarity. If you are streaming today’s call, please mute your computer speakers.
And your first question comes from Stephen Ju with UBS. Please go ahead.
Stephen Ju: Great. Thank you. Hi, Susan, Chad, Ken. So, from the outside looking in, I think
there’s been some reports of some headcount changes, especially within some
of your AI headcounts. So, I guess, in this era where that kind of talent
continues to be pretty difficult to come by, it seems a little bit at odds with
what’s kind of going on out there.
So, can you talk about where you’re prioritizing, where you’re reprioritizing,
what you might be deprioritizing, et cetera, in terms of what you might be
looking to do?
And for next year around the OpEx commentary, the D&A expenses and the
energy expenses, that’s all straightforward. But is there anything you can add

-- 1 of 10 --

2
from a headcount expense standpoint as we think about the modeling for next
year? Thanks.
Susan Li: Hi, Stephen. Thank you for the question. So, a bunch of points to that, and I
will try to get to all of them.
So the first, I think you were referencing reading about some of the press
reports about the MSL reorg. That was effectively first. Over the course of the
last one to two quarters, we’ve been really focused on building out that team.
Mark described it, I think, very eloquently in his remarks as the most talent-
dense team in the industry.
Now that that team has been largely built out, we are making changes to
optimize the org structure. It has three key areas, research, which is led by
Shengjia Zhao, who is our Chief Scientist, and Rob Fergus is leading the longer
term research efforts within FAIR. Then, there’s product, which is led by Nat
Friedman, and Infrastructure, which is led by Aparna Ramani.
And the reorg that we did to optimize the org structure, we did that in
conjunction with sort of the headcount reduction. We’ve really set up our new
research lab to be the most talent-dense team, and we’re trying to carry that
principle over to other areas of MSL as well.
And the team size reduction is really to ensure that each person working within
the lab has more scope and autonomy. We find that smaller teams can move
more quickly.
Also, to your point about the difficulties in AI recruiting, generally, the majority
of the people we believe moving out of MSL will have the opportunity to find
new roles elsewhere in the company. And we continue to focus on hiring
industry-leading, AI native talent. So that’s on the first part of your question.
The second part of your question, just in terms of working through where the
2026 hiring will be, we expect that we’ll be growing headcount in 2026 in a
very targeted way, again, across the company’s priority areas, monetization,
infrastructure, AI, and that the talent will generally be concentrated in technical
roles, which, again, have the dimension of carrying a higher compensation
profile, so that accrues to compensation expenses growing faster than
headcount.
But, having said that, we are not done yet. We are very focused on making sure
that the company at large, including the priority areas, is operating efficiently
and leveraging, frankly, the AI tools that we have available to ensure that
people continue to operate with increased with increased productivity as those
tools become more effective and performant over time.
Operator: Your next question comes from the line of Benjamin Black with Deutsche Bank.
Please go ahead.

-- 2 of 10 --

3
Benjamin Black: Great. Thank you for taking my question. So, Susan, can you talk a little bit
about what you’re seeing across the advertising operating environment? How
is the macro backdrop holding up as you sort of enter the holiday shopping
season?
And then, I think you touched on this in the public call, but could you talk a bit
more about your initiative to leverage LLMs to drive content understanding,
which I guess feeds into your recommendation models. How do you think
that’s going to impact engagement and what hurdles are there to rolling it out
more broadly across the Family of Apps? Thank you.
Susan Li: Yes. So on your first question, we continue to see that our revenue growth has
been broad-based across all user regions, across most advertiser verticals. We
think that the macro landscape as it is today is strong, and obviously that is
something that can change and there are many factors outside of our control.
But we think that the revenue guidance range that we give helps capture the
likely span of possible outcomes there.
We also believe in addition to the macro backdrop, again, being strong as it is,
that we’re also quite happy with the pipeline of revenue initiatives that we have
internally that we believe will help us deliver continued strong revenue growth
in 2026. Yes. Chad, do you want to take this other part of the question?
Chad Heaton: Yes. So I think the second part was how we’re going to be using LLMs to
improve the engagement and our recommendation systems. So, we’re
continuing to work on utilizing multimodal LLMs more into our
recommendation systems across our apps. We expect this is going to be
particularly useful for improving content understanding as LLMs can more
precisely label the keywords and topics within videos and posts.
This is going to help us classify content with more precision, which then helps
our systems develop deeper intuitions about why a piece of content is
interesting to someone, and then, also, better identify content that matches
those interests.
So, we’re expecting to leverage LLMs for content, understanding more broadly
across Facebook and Instagram over the coming months.
Operator: Your next question comes from the line of Shweta Khajuria with Wolfe
Research. Please go ahead.
Shweta Khajuria: Thanks a lot for taking my question. I have one on Meta AI Vibes. So, that has
shown since launch some good traction from the data we see. So how are you
thinking about differentiating the product itself?
You certainly benefit from distribution and that’s what Mark talked about on
the call. But from the product perspective, how do you think it’s going to
differentiate durably versus competitive products? Thank you.

-- 3 of 10 --

4
Susan Li: Yes. I think we’re pretty early in understanding how this is going to evolve.
Now, as you noted, we have seen strong increases in usage of Meta AI since we
launched Vibes. We’ve also seen that media generation in the app is over 10
times higher in the app than since we launched it and we’re seeing DAU growth
inflect.
So, it’s still early, obviously, but we’re seeing some clear signals that there are a
lot of people who are interested in using AI tools to create content.
I think that, given the degree to which we are seeing strong growth, even with
the limitations of the models today, I think the fact that the models are only
going to get better from here is going to enable us to tap into a lot of the
strengths of our platform as we introduce the next models.
The fact that we have a massive ecosystem of content already produced by
people today should mean that having AI editing tools that reduce the barrier
to editing content should enable even more and faster content creation.
We find that it’s easy for content on our services to pick up traction and
distribution as we add more AI capabilities. Those capabilities can get surfaced
quickly and we think that it’ll leverage our distribution platform in that way.
We also have strong capabilities in messaging. What you see with a lot of AI
generated content is that you may create something that you just want to
share with a small audience pretty easily, and then you can message it to them.
And then, finally, we think that as these tools make it easier for people to create
content, we should expect content in the ecosystem to grow meaningfully, and
that’s where having a strong recommendation system that can then help
match the right content to the right person can be a big advantage.
So, I think it’ll be very interesting to see how this evolves over time. But I think
the combination of continually improving models, and then the ecosystem
distribution and recommendation advantages that we have, I think will really
enable us to tap into kind of a lot of our core strengths.
Operator: Your next question comes from the line of Deepak Mathivanan with Cantor
Fitzgerald. Please go ahead.
Deepak Mathivanan: Great. Susan, a couple questions. So, first, the time spent on the platform
continues to see very good growth based on the gains that you’re seeing with
recommendations. But curious if the growth of platforms, like ChatGPT, and
also Sora recently as -- it was kind of like a weight on engagement, or maybe
specific types of engagement in certain cohorts or geos of users that you’re
seeing. Broadly, how do you view the competition from AI tools that are
scaling to the time spent on your apps?
And then, second one, if I can, can you give us an update on where the
penetration of GEM and Andromeda across sort of the various surfaces, be it
Feeds, Stories in Facebook and Instagram currently is? Just trying to

-- 4 of 10 --

5
understand how additional capacity coming live can take this penetration
higher in the next few quarters. Thank you so much.
Susan Li: Yes. I mean, again, I think we’re able to measure our own engagement directly,
and what we can measure is, again, that we are seeing very healthy trends
broadly that the product improvements and AI recommendations are driving
incremental engagement as we would expect.
And we talked about some of those things on the call. The AI recommendation
system is driving 5% more time spent on Facebook and 10% on Threads, and
generally growth in family DAP was healthy at 8% year-over-year in Q3 and
daily usage continues to grow across each of Facebook, Instagram, and
WhatsApp.
So, we feel like the engagement trends we see are quite strong. And for us, it
isn’t sort of any single launch. We see that there are many optimizations that
we make to the recommendation systems over the course of the quarter that
help us make more accurate predictions about what will be most interesting to
each person.
So whether that is adding new engagement signals to our ranking models on
Reels, such as seeing how people save content or interact with re-shared
content and using that to better personalize the posts we recommend, or on
Facebook video, we integrated more engagement signals that we were able to
use to better align our content recommendations with people’s preferences,
and the increased emphasis on ensuring that we are increasing the freshness of
recommended content, so people see more recent posts when they begin their
sessions.
So, we feel like, again, the core engagement metrics that we look at are
generally healthy, and we feel like we’ve seen the recommendation
improvements that we’ve made pay off, and we also feel like when we look
forward, is a good -- we see a lot of headroom to improve recommendations in
2026, whether that is around scaling up our models and increasing the amount
of data we use to improve recommendation quality, whether it’s around
evolving our recommendation systems to help people discover more diverse
and niche content.
We’ve made some of these changes already on Facebook and have seen good
results and are evolving Instagram in the same way to consider posts you’ve
already seen when deciding what to recommend to you.
We’re also focused on making the recommendations even more adaptive to
what a person is engaging with during their session, so that the
recommendations we surface are more relevant to what they’re interested in at
that moment. And then we have longer-term research bets around developing
foundation models that will support recommendations across multiple
surfaces, and some of what Chad alluded to earlier around incorporating LLMs
more deeply into our recommendation system. So, there’s, I think, a good
pipeline of work to do there.

-- 5 of 10 --

6
Chad Heaton: On the second part of your question about the penetration of Andromeda and
GEM, so starting with Andromeda, so, in Q3, we focused on improving
Andromeda’s performance by consolidating retrieval and early-stage ranking
runtime models into a single model. So that drove a 14% increase in ads quality
on Facebook ad surfaces that we introduced it to, which were Feed, Reels, and
Marketplace.
So later this quarter, we’re planning to roll out Andromeda to additional
surfaces, including Instagram Feed and Reels, and we’re also working on
further consolidating early-stage ranking models using the Andromeda model
architecture to drive additional gains.
In terms of GEM, in Q3, we made improvements to GEM’s model architecture
that doubled the performance benefit we get from adding a given amount of
data and compute. That’s on top of the prior 2x efficiency improvements that
we made earlier in the year. So now, GEM is actually four times more efficient
at driving ad performance gains for a given amount of data and compute than
our original recommendation ranking models.
So, going forward, we’re planning to extend GEM to the rest of Facebook Reels
in Q4. At that point, GEM will cover all major surfaces across Facebook and
Instagram. We’re also exploring training GEM on even larger cluster in the
future, and we’re focused on improving how we transfer the learnings from
GEM foundation models to our smaller vertical-specific models.
Operator: Your next question comes from the line of Tom Champion with Piper Sandler.
Please go ahead.
Tom Champion: Hi. Good afternoon, Susan and Chad. Susan, you’re giving us a very clear
message about how OpEx growth is going to inflect next year. It looks like 23%
growth at the midpoint this year. Just any help, color, thoughts on the
magnitude of the acceleration in OpEx growth into next year?
And just one quick one if I can. It looks like ad impression growth decelerated in
the U.S. and Canada. That’s divergent from the rest of the geography. Just any
color on why that would be. Thank you.
Susan Li: Tom, thank you for the question. On the first part, we are still in the middle of
our budgeting process for ‘26, so I just want to be clear that this is all still in
flight. But the factors that we really think are driving year-over-year expense
growth, the predominant one is infrastructure costs. That will be the single
largest contributor to 2026 expense growth.
And that is driven by a significant step up in both the growth of infrastructure
operating expenses, which includes not only the traditional OpEx that scales
with our data center footprint, things like energy, leases, OpEx, but also a
significant step up in cloud costs as we have been trying to accelerate near-
term capacity, per Mark’s comments on the call.

-- 6 of 10 --

7
We do also expect a significant acceleration in depreciation growth driven by us
recognizing incremental depreciation from the infrastructure that we’ve been
adding throughout 2025 and will add in 2026. So, that’s really the biggest
driver. After infrastructure, employee comp is the next largest driver of growth
in ‘26. That’s primarily driven by us recognizing a full year of compensation
expenses for the talent that we hired this year. Particularly, AI hires, as well as
the other investments we’ve been making in technical talent throughout this
year and next.
Chad Heaton: And on your question about North America ad impressions, so they grew 8%
year-over-year, which was driven by growth in engagement and users, with
engagement benefiting from ranking improvements on feed and video.
The deceleration, you mentioned there was a slight one-point deceleration
from Q2. That was driven by lower ad load on Facebook due to a growing
proportion of video time on the app, which, as you know, has less ads per unit
of time.
Operator: Your next question comes from the line of Rich Greenfield with LightShed
Partners. Please go ahead.
Rich Greenfield: Hi, thanks for taking the question. I guess for Susan, just following up on that
last question. If you think about the guardrails for how you think about the
expense growth relative to the revenue, you’re talking about strong revenue
growth, but obviously also significant cost growth. And how related are they in
terms of how far the expense growth goes relative to what you are seeing as
you move into next year, as you budget for next year on the revenue side? I’d
be curious how you think about the interplay between those two.
And then just a quick separate question. The advertising or e-comm advertising
in video games has started to really explode. AppLovin’s been talking a lot
about it. And obviously, they’ve been certainly targeting some of your
customers. I’m curious how you think about advertising and attacking
advertising inside of video games, and leveraging your advertiser base, and
where that is as a priority for Meta.
Susan Li: Thank you for the question. So, on your first question, we have shared in the
past that our primary focus from a profitability perspective is driving
consolidated operating profit growth over time. We expect that path is not
going to be linear. In many of the past periods, we’ve delivered above-average
profit growth. We expect that in periods where we’re making big investments,
that could impact our near-term profitability.
And right now, we’re in a very dynamic planning environment. We think there
are these big strategic benefits to adding infrastructure capacity sooner than
later, given how R&D are progressing today.
So, right now, the highest priority for the company is investing our resources to
position ourselves as a leader in AI. That means that I think for the immediate
period of time ahead of us, we could see some financial pressure during which

-- 7 of 10 --

8
our operating profit could be lumpy. But over time, we certainly believe that
that framework of consolidated operating profit time will not only be true, but I
mean obviously it would be disappointing if we didn’t do better than that over
the long term.
Operator: Your next question comes from the line of Barton Crockett with Rosenblatt.
Please go ahead.
Barton Crockett: Okay, thank you for taking it.
Chad Heaton: There was a second part of the question, sorry, that we didn’t quite get to. So,
you mentioned a question about AppLovin specifically. Obviously, we’re not
going to speak to any specific competitors with respect to budget allocations
between our platforms and others. But we just say generally, it remains a very
competitive environment.
And obviously, over time, advertisers are going to allocate their budgets where
they’re getting strong ROI. So, our focus is just going to be continuing to
deliver the most competitive ad products that we can. And we feel really good
about the ad performance gains that we continue to deliver for advertisers.
And we see opportunities to drive ongoing ad performance gains in the future.
You touched on the gaming vertical specifically. So, year-over-year growth in
gaming was strong in Q3. That benefited from spend by the U.S. and Asia-
based advertisers. Growth was also accelerated notably in Q3 as we lapped a
period of weaker gaming spend from Asia-based advertisers.
Kenneth Dorell: Krista, we can go to Barton now. Thank you.
Operator: Barton, please go ahead.
Barton Crockett: Okay, great. Thanks for taking the question. So, I was wondering about the loss
profile in the Reality segment. As you’re focusing on investing so much more
on AI across the core Meta apps, and as you’re going through this transition
from Quest to the augmented reality glasses, is there any thing you can tell us
about how we should think about those losses in Reality and the investments
there? Is there an argument maybe perhaps that they scale back as you go
through this transitional period?
Susan Li: Yes, it’s a great question. We’re not sharing an outlook for Reality Labs
operating losses in 2026 in part, again, because we are still working through
the budget details. What I can say from the process so far is we’re really trying
to shift momentum towards AI glasses.
And that’s been one of the biggest priorities looking at the Reality Labs’ budget
and roadmap for 2026. And that’s in part because we have seen that there is a
lot of demand. I would say we were stocked out multiple times over the course
of this year, and we want to get ahead of that.

-- 8 of 10 --

9
Both because there is product market fit, and also because it’s a great and very
natural platform or form factor for AI experiences that can be built on top of
those and carried with you into the world. So, that’s really what we are looking
at as our highest priority when we think about what the 2026 Reality Labs
roadmap looks like.
Operator: Your next question comes from the line of Aaron Kessler with Seaport. Please
go ahead.
Aaron Kessler: Great. A couple of questions. Maybe just on -- I know you’ve been asked this
before, but on measuring the ROI for non-core investments, I mean, should we
think about this over a multi-year period in terms of that measurement? And
then just any feedback thus far on AI videos and Reels, just I think consumer
engagement with those. Do they like them? Do they not like them? Thank you.
Susan Li: On the first question, certainly, I would say the non-core investments have a
multi-year time horizon. I think Mark alluded to this in his script earlier in terms
of there’s some debate about in exactly what time frame will we get super
intelligence. I think he laid it out as two to three years or five or seven. But
those are single-digit year time horizons that he has laid out there.
So, I would say that while it’s definitely -- I think in 2026, we don’t expect that
we are going to generate a lot of revenue from the MSL or gen AI efforts. I do
think this is not a decades-long time horizon by which we would be evaluating
the ROI of the spend.
And I also think, a point that I just want to make sure really is clear from Mark’s
comments, is he really wants to be prepared for -- to not err on the side of
being under-invested when that paradigm shift happens. And so, we think for
those reasons, and including looking at the roadmap of our researchers and the
progress that is going on in TBD Labs, that it makes sense for us to more
aggressively front-load building that capacity. But that does mean that it puts
you again on this multi-year time horizon in terms of evaluating what the return
profile will look like.
Chad Heaton: In terms of the AI videos and engagement, look, it’s very early. So, we’re just
looking at very early signals here. But I think it’s clear to us that AI-assisted and
generated media is going to be a big deal based on some of these early signals
that we’re seeing. So, even with the limitations of the models today, we’re
seeing some pretty strong growth.
People have created over 20 billion images using our products. And since
launching Vibes within Meta AI in September, we’ve seen media generation in
the app more than tenfold. So, again, super early, but the signals seem to
suggest that this is going to be a pretty big thing.
Operator: Your last question comes from the line of Dan Salmon with New Street
Research. Please go ahead.

-- 9 of 10 --

10
Dan Salmon: Good evening, everyone. Susan, can you just clarify the $60 billion run rate for
the Advantage+ suite? Is that for the entire suite of products, or is that for just
the three main campaign modules, thinking of sales, app, and lead generation?
Thanks.
Kenneth Dorell: Hey, Dan, this is Ken. Yes, so it is specifically the run rate for the end-to-end
automated solution that we have in market. So, that includes sales objectives,
it includes lead objectives, it includes app objectives.
This is the solution that basically automates and optimizes several aspects of
the campaign set-up process at once, including selecting the audience, where
to show the ad, how the campaign budget gets distributed across the ad sets
to drive the most efficient outcome. So, it is specifically focusing on that end-
to-end automated solution.
In addition to that, we have single-step automated solutions like Advantage+
Audiences, but the $60 billion stat is specifically referring to that full end-to-
end automated solution for the objectives that I mentioned.
I think we are at time, so we are going to wrap it here. Thank you guys so much
for taking the time. Really appreciate everyone dialing in, and we look forward
to speaking with you again soon.
Operator: This concludes today’s conference call. Thank you for your participation, and
you may now disconnect.

-- 10 of 10 --