AI Shift: AMD Gains, Microsoft Falters, EU Regulates
Key Points
- AMD’s latest earnings beat expectations by $0.5 billion in its GPU division, driven by strong demand for chips used in large‑language‑model training, prompting an upbeat outlook and Wall Street optimism.
- Microsoft’s earnings missed the mark as cloud revenue slowed slightly while capital expenditures jumped 60 % for AI‑related datacenter build‑outs, leading investors to doubt a timely revenue payoff and causing a stock dip.
- On August 1, the EU’s AI Act officially takes effect, introducing broad, phased‑in restrictions on a wide range of AI models and creating regulatory uncertainty even for major tech firms.
- The simultaneous market reactions to AMD’s success, Microsoft’s spending‑revenue mismatch, and the looming EU AI regulation illustrate the current tension in the AI cycle between rapid technology adoption, investor expectations, and emerging legal constraints.
Full Transcript
# AI Shift: AMD Gains, Microsoft Falters, EU Regulates **Source:** [https://www.youtube.com/watch?v=jNGQMUoGMfQ](https://www.youtube.com/watch?v=jNGQMUoGMfQ) **Duration:** 00:17:31 ## Summary - AMD’s latest earnings beat expectations by $0.5 billion in its GPU division, driven by strong demand for chips used in large‑language‑model training, prompting an upbeat outlook and Wall Street optimism. - Microsoft’s earnings missed the mark as cloud revenue slowed slightly while capital expenditures jumped 60 % for AI‑related datacenter build‑outs, leading investors to doubt a timely revenue payoff and causing a stock dip. - On August 1, the EU’s AI Act officially takes effect, introducing broad, phased‑in restrictions on a wide range of AI models and creating regulatory uncertainty even for major tech firms. - The simultaneous market reactions to AMD’s success, Microsoft’s spending‑revenue mismatch, and the looming EU AI regulation illustrate the current tension in the AI cycle between rapid technology adoption, investor expectations, and emerging legal constraints. ## Sections - [00:00:00](https://www.youtube.com/watch?v=jNGQMUoGMfQ&t=0s) **Connecting AMD GPU Surge, Microsoft Earnings, AI Cycle** - The speaker links AMD’s surprisingly strong GPU earnings, Microsoft’s modest cloud‑revenue slowdown, and what these recent developments reveal about the current stage of the AI boom. ## Full Transcript
sometimes the news cycle happens so fast
it's hard to keep track and certainly
hard to put the pieces together this
post is focused specifically on putting
the pieces together around three big
pieces of news that broke in the last
couple of days at the very end of July
in 2024 I'm going to list the pieces of
news very quickly first and then we're
going to talk about the relationship
between them and what it says about this
point in the AI
cycle number one AMD did very very well
in their latest earnings release their
revenue for their GPU chips division
which is where people go to get chips if
they're buying from AMD for machine
learning if they're buying from AMD in
particular for large language model
training which is a lot of where this
chip purchase demand comes from that's
surprised to the upside by half a
billion dollars they made half a billion
dollars more than they expected to make
from their GPU division Wall Street was
super happy had a great quarter in fact
they increased their guidance for the
future because they think this is going
to keep happening all right that's the
first piece of news second piece of news
Microsoft got punished in the markets
yesterday when they also reported
earnings not because the earnings were
particularly out of line with
expectations there was a tiny bit of
deceleration in their Cloud Revenue by
uh one or two percentage points which
does matter at their scale and in
particular there was a significant
reported annual increase in capital
expenditure which Microsoft has been
telegraphing all year long by having
giant conferences and saying we're
spending a lot on training data centers
and hey we're going to build a nuclear
power plant and all all kinds of things
but at the end of the day it does come
down to Dollars and cents for Wall
Street and Wall Street looked at the 60%
increase in capital expenditure and then
looked over at the cloud revenue and
really didn't see a corresponding
acceleration in Revenue that could be
driven by that capital expenditure in Ai
and Wall Street got worried and so
Microsoft took a bit of a hit
that's the second piece of news the
third piece of news is around Ai and
regulation it's from across the pond so
August 1st tomorrow is when the eui ACT
goes into effect it's designed to
prohibit or control or regulate many
many different kinds of AI models and it
goes in a phase roll out over time the
language appears to be relatively Broad
and I am not the only one who doesn't
know exactly what it means uh in fact
very large companies with lotss of very
well-paid lawyers also don't know enough
about what it means to be confident in
their AI approach in the European Union
for example meta has decided not to roll
out their AI capabilities for search you
know that little blue circle if you live
in the US where you type it in and it
will sort of do AI search we kid about
it because it's not always useful in the
US but the EU doesn't get to kid about
it because they don't get it at all
because meta decided to depos it because
of the the uncertainty around the AI act
all right those are the three pieces of
news what does it mean how do these go
together so the first takeaway I have is
that this remains a year when the
companies that are selling the first
part of the value chain are the ones
that are doing well Nvidia has reported
really strong earnings this year AMD is
now reporting really strong earnings
this is what people are uh calling the
picks and shovels part of the Gold Rush
where you are just selling the equipment
that enables everybody else to do what
they do with AI and you were doing very
well on Revenue as a result the capital
expenditure that Microsoft was reporting
shows up on chipmakers balance sheets in
their case I think they're using
primarily Nvidia but the point stands
right fundamentally the money that is an
expense for a software company building
out cloud is a revenue for the chip
makers and so all of those chip makers
are having great years they're all
surprising to the upside they all have
strong guidance like there's all of this
sort of momentum in the chip
space and that is exactly what we would
expect in the first part of a larger
value chain that's unlocking for AI
That's not a surprise it shouldn't be a
surprise if you've been watching
AI what's interesting to ask yourself is
when that part of the acceleration curve
levels off and the signals we're getting
both from the manufacturers of chips
themselves and from their customers are
that this is not leveling off in the
next couple of years and that's why
everyone's so excited about chip
manufacturers it looks like they have a
great Market they've stumbled into
basically making making chips for the
greatest technological project that
Humanity May ever work on with this
whole artificial general intelligence
thing that's not the topic of this video
but that's what people are thinking
about right and that's the kind of
investment Horizon they have what if we
invent into a general intelligence we'd
pay anything for it that's what's
driving the appetite
Etc
so they're doing well the chip
manufacturers and it's interesting
because they're doing well even though
the software companies that buy their
chips know that they are in for a
beating in the markets for spending that
much on AI and they just don't care
because they think it's so important so
Google meta were actually leaked to have
leaking memos uh back and forth a couple
week weeks ago basically saying they
know they are overspending by billions
of dollars on AI and from a game theory
perspective they would rather be all in
on AI versus risk a catastrophic loss of
business if they miss the boat on this
AI Revolution so they're overspending on
purpose and that includes overspending
potentially on chips which is great if
you're selling chips which is what AMD
is doing so this kind of brings us to
Microsoft they're also in that same sort
of game Game Theory position where they
need to establish a strong Cloud product
that is AI driven they have this startup
that is the darling of the world and
open AI that they're also helping to
prop up and support and they need to
spend like crazy to make sure that they
maintain their position uh across their
Azure division across open Ai and sort
of the model development
curve and that's what they've been
signaling they're going to spend spend
spend spend span there was a picture on
a slide that the CTO of Microsoft gave
that basically was like these are the
the size of our models and it was
different animals right and it was like
you know hey look here's you know a a
shark and here's an orca whale and
here's this freaking big gray whale over
here and they're basically saying these
models are going to get bigger which
means bigger and bigger and bigger chips
uh or bigger and bigger chips uh sizes
in in data centers right more and more
count of chips in data
centers and that all translates into
exponentially growing revenue for the
for the chip
side the problem for the markets is that
the founders the CEOs of these companies
are thinking in much much longer time
Horizons than street so Wall Street is
like is is liking you know 12 18mth Roi
they feel like they've given these
companies plenty of leash to go ahead
and produce results from Ai and look at
this they haven't really seen a gain in
Revenue well if you work in Tech you
know actually rolling out AI
applications takes a ton of work
actually getting AI Enterprise is really
hard and in many cases it's going to
take a long time for a full incremental
Revenue shift to actually translate into
the bottom line or even the top line for
a company like
Microsoft lots of companies are doing AI
experimentation AI demand which the CFO
actually called out on the call is
exceeding their capacity despite their
60% SP uh increase in
capex so they are seeing tremendous this
demand they say which I believe for
their AI services but that demand isn't
necessarily at the scaled application
stage and what it suggests to me is
they're investing a huge amount just to
get to that sort of initial
experimentation phase just to get to the
idea of rolling out these applications
over time and I have two thoughts there
one the investment Horizon needs to be
longer if you're actually looking at AI
value and that something that the CEOs
in these large companies seem to
recognize and Wall Street doesn't it's
going to be 36 it's going to be 48
months it's going to be much longer than
the street typically tolerates and so I
would expect to see continued
frustration from Wall Street towards
some of these Executives as they see the
larger opportunity long term five seven
years out and the street doesn't look at
it that way they look at the quarters
this happened famously with Amazon where
Jeff Bezos when he started the company
basically told Wall Street don't look
for profits anytime soon I'm just not
going to produce them I'm going to work
on scaling the
business and that's sort of the position
that we're
in now what's interesting is that the
software side of the business is the
next one we would expect to catch up
after the chips are starting to sort of
fully built
out you know after the revenue is done
from AMD from Nvidia and I I don't want
to say done in the sense that they're
not going to continue to spend on chips
all the signs are that they are for the
next you know or two years here at least
but when you have purchased a chip you
then have to put it into a
model at a data center get the model to
train across a data set using that chip
and then at that point you can finally
use it but then you have to have a
separate runtime application environment
and a separate runtime Cloud Center
where you can run it at scale and so
there's a whole another piece of capital
expenditure that goes into sort of
running the model enabling retrieval all
of
that so it's a whole of upfront expense
and then when you're running it it's not
clear whether the monetization that
you're
offering is going to be truly
incremental across an Enterprise's Cloud
spend or not is it something where a big
company like Walmart is going to say my
total Cloud spend is X I'm going to pay
X+ y because of
AI and I'm willing to tolerate a growing
expense in my cloud footprint even if
every anything else in my environment is
really margin competitive or are they
going to say you know what this is not
really worth it I want to make sure that
my Revenue overall or that my spend
overall on cloud stays flat that's
Microsoft's revenue and I will put AI in
but I'll negotiate out negotiate out a
bunch of other things like is AI
effectively replacing things on the
balance sheet now what's interesting is
that we won't necessarily know that for
a few Quarters here my hunch is that
eventually it's going to be incremental
and you're going to still see Cloud
revenues rise and I think the question
for Microsoft is and for for Amazon and
for Google these all run huge clouds is
whether the rise that you see is going
to be enough to re accelerate their
enormous Cloud businesses so their Cloud
businesses famously tend to run at
double- digigit growth per year 20% 25
30 whatever uh and the point is are they
going to have enough incremental growth
to re accelerate those businesses so
instead of it being you know 25 this
year and 24 the next year and 23 the
year after that is it 25 27 29 32 right
like are you seeing this acceleration
curve in Revenue growth because there's
so much demand for AI that is certainly
what Wall Street is hoping for they
always look for that sort of
accelerating Revenue picture it is
increasingly hard to deliver as your
business gets larger and
larger I will be curious to see how it
plays out it is not at all clear to me
even if Revenue grows net net
whether revenue is going to grow faster
and faster for these large Cloud
companies based on AI or whether it's
going to essentially prop up and enable
continued scale up of existing Cloud
resources existing Cloud divisions at
these
companies we will have to see stepping
back what you're seeing is the chip side
selling the Pix and shovels in the Gold
Rush doing very well today you're seeing
immense capital expenditure from these
cloud providers which is sort of the
next piece in the value chain that
corresponds to revenue for the chips
people and then from there what you
really need to see is that there is a
corresponding Revenue guidance increase
in the End customer we're a long way
from that but if they're going to spend
more using AI on these Cloud providers
ultimately the only reason they will do
that is if they believe their revenue
and their customer experience and
everything else about their Core
Business is going to substantially
improve it's going to improve enough to
be worth the AI spent it is not at all
clear yet when that will happen and so I
see this in three phases right we're in
solidly in phase one where we see this
huge chips Revenue increase phase two is
a question mark we see the spend side on
data centers for companies like
Microsoft we do not yet see the revenue
side come through and then phase three
is end companies companies like Walmart
other companies that are consumer facing
or business facing that are consumers of
the cloud they need to start to come
through and say AI is going to deliver
this kind of value in our business and
so we're spending more on it that hasn't
really happened yet either and I'm
keeping an eye out for it we should
start to see it in the next year or two
here if we're going to start to see this
value chain for AI play out the way
people are
hoping last but not least let's talk
about the EU
piece at the end of the day this is an
environment where we are seeing some of
the fastest technological change in
history even fast by Silicon Valley
standards it is so fast it is hard to
deploy on a lot of AI tool chains right
now it's just really tricky because they
keep updating them all the
time so in that
environment you need to have a
regulatory space that enables The
Innovation to
happen and that is the piece that the EU
is saying we're going to take a stance
on early and that's very typical of the
EU they tend to sort of regulate first
and think about permissiveness after and
I'm not saying that in a pajora of sense
that's just sort of how they've
approached things that they how they
approached it with gdpr we have worries
about privacy so we're going to make
sure that we catch that early and we're
not going to let it happen to our people
and then figure out the regulations
later
so the US has a different approach the
us if it regulates at all in these
spaces tends to regulate after the fact
and it tends to regulate in in broadly
speaking in a more corporate friendly
way and so what you're seeing is that
when the technological pace of changes
really fast when the future is uncertain
when you have to put big bets down and
you're intentionally playing
effectively um a game theory arms race
with other major tech companies you want
to be sure that you're not multiplying
your uncertainty you want to make sure
that you are limiting your uncertainty
to the already immense risks associated
with the actual AI capex expenditure
you're making so it is irrational in
that environment if you were a major
Company CEO to go out and say I'm going
to multiply my Risk by rolling out this
very expensive model in a place where my
lawyers can't be exactly sure what is
going to happen over the next two or
three years in the AI space in that
regulatory environment and that is the
reason why meta decided to not roll out
their AI in the EU but what that means
for EU businesses is that they are going
to lose some of the AI cont context that
they need to be competitive on a global
level so if you're running a business in
the EU and your American counterpart
would have access to AI modeling and uh
AI Services it is not at all clear that
you will have the same kinds of access
because of the regulatory environment
and that access is changing so fast
because we're innovating so fast that
there is an increased risk that you as
an EU business owner using AI may get
get more left behind now that's not a
guarantee you do have a major AI player
in mrr that is based in France so there
is definitely a AI startup activity in
the EU I don't want to take that away at
all but the uncertainty is making
American companies more cautious about
moving into the EU and that is going to
affect the long-term competitive picture
for tech companies in Europe in ways
that we don't fully understand and all
of it is stemming from some sort of
uncertainty around how these regulations
will roll out and take effect in the
face of this overwhelmingly fast uh
technological Revolution that we're all
living through okay that's a lot all
from three key pieces of news that came
out in the last day or two around AMD
around Microsoft around EU I hope the
breakdown was helpful to you I hope you
feel like you understand a little bit
better some of the underlying drivers
that are shaping the world that we are
all living in with they are let me know
what you think in the comments h