January 27, 2025 • 38 mins
Chuck Zodda and Mike Armstrong discuss the buzz DeepSeek is generating and why US tech stocks are falling today because of it. How does the release of DeepSeek impact other sectors outside of tech? Earnings season gets real as the Magnificent Seven begin reporting this week. Americans are carrying bigger credit card balances. Airlines are charging higher fares and are confident you will pay up.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
The Financial Exchange is produced by Money Matters Radio and
is hosted by employees of the Armstrong Advisory Group, a
registered investment advisor. All opinions expressed are solely those of
the hosts. Do not reflect the opinions of Armstrong Advisory
or anyone else. Investments can lose money. This program does
not offer any specific financial or investment advice. Please consult
your own financial, tax, and estate planning advisors before making
(00:20):
any investment decisions. Armstrong Advisory and the advertisers heard on
this program do not endorse each other or their services.
Armstrong and Money Matters Radio do not compensate each other
for referrals and are not affiliated. This is the Financial
Exchange with Chuck Zada and Mike Armstrong, Your exclusive look
at business and financial news affecting your day, your city,
(00:42):
your world. Stay informed and up to date about economic
and market trends plus breaking business news every day. The
Financial Exchange is a proud partner of the Disabled American
Veterans Department of Massachusetts.
Speaker 2 (00:56):
Help us support our.
Speaker 1 (00:57):
Great American heroes by visiting dav five K Boston and
making a donation today. This is the Financial Exchange with
Chuck Zada, and Mike Armstrong.
Speaker 2 (01:08):
Chuck, Mike and Tucker with you here. And we've got
a sell off happening in markets today, but really the
sell off is concentrated primarily in big tech. And the
reason why, if you have not heard of the AI
platform deep Seek, you're going to hear about it right now.
If you have heard about it, then you might know
(01:29):
some of what we're talking about. But quite honestly, we
think that we are very well versed on this topic,
and so you're going to learn some new stuff from us,
you better believe it.
Speaker 3 (01:37):
And if you had heard about deep Seek prior to
five days ago, I'd be particularly interested because because nobody.
Speaker 2 (01:45):
Because you hadn't. Let's be honest. So what is deep Seek.
It is an AI platform that is built by a
Chinese company by the name of high Flyer. High Flyer
is a hedge fund. And why did you Flyer go
and build deep Seek? Now? The reason why They had
been building internal AI processes for their investment platform, trying to,
(02:10):
you know, find better ways to beat the market investing,
and as part of that, they just had a natural
interest in kind of the latest and greatest AI technology,
and so they started building AI models a couple of
years ago. The one that they released on January twentieth
is called deep seek are one and the premise of
this is that it competes directly with open AI's flagship
(02:32):
open ai Chat GPT oh one product that came out
I want to say, in like September or October, and
now Chat GPT's been around for two years.
Speaker 3 (02:42):
Yeah, a little more than two years and it's public
facing will but this latest iteration of their large language
model is you said, one oh one.
Speaker 2 (02:50):
And the big thing that one claimed to be able
to do and that it can do is it can
handle more logical reasoning things. So you ask it, you know,
logic based quests, and it can actually kind of think
through them in certain respects. And this platform, the deep
seek are one platform, does the exact same thing. So
why does this matter? A couple reasons. The first is
(03:15):
deep Seak claims that they built this using about five
point six million dollars of what they call AI training,
basically in putting all the data into the model and
getting it to understand what it is. Sounds like a
lot of money, Chuck, is it no, Because most AI
models that are produced in the US right now cost
anywhere from about five hundred million to a billion dollars
(03:35):
to do the training aspect.
Speaker 3 (03:36):
Now keep in mind, this is the training part where
they're feeding at internet articles and the entire history of
the New York Times and Facebook and Twitter and all
these different things. This is just purely teaching the thing
how to read and think and do everything correct the
AI models can do.
Speaker 2 (03:54):
And basically the training costs is a function of two things.
How many computer chips do you need to train on
and how long do you need to do that for?
Those are the two things that matter, because you're you're
basically paying for the compute time. What you are then
doing on the the other end of that is after
the training is completed, hey, you still need to pay
for the processors to go and actually execute the queries
(04:17):
that are asked of the model. So that's why if
you're using open AI's enterprise platforms or they're you know,
the their high end stuff for one and again, these
numbers are going to be mean nothing to you. But
don't worry about it. I'll make it make sense, I promise.
If you need inputs from open ai, it costs fifteen
dollars per million inputs for outputs, it costs sixty per
(04:38):
million deep seek Remember open as fifteen per million inputs.
Deep seek is fourteen cents per million inputs one percent
of the cost. On the other side of things, outputs
open AI is sixty per million outputs. Deep seek is
two dollars and nineteen cents.
Speaker 3 (04:53):
Okay, So we've basically now understood from this company's own
I wouldn't even say disclosures, but yeah, information that they've
put out about themselves is that training and getting output
out of this large language model costs one percent on
the training side and three percent of chat GPT's own
(05:15):
disclosed information on the output side of things. I would say,
Chuck that this is one area that we probably have
the least degree of certainty about this new model. Right, Like,
it's a company held out of China, So could there
be some accounting creativity outright lies a whole bunch of
subsidies from the Chinese government. All of those things are
(05:38):
entirely possible. But what they are claiming is that, yeah,
the cost to both run and train this thing, we're
one to three percent of what we've been experiencing in
models here in the United States, and getting we'll get
to the results side.
Speaker 2 (05:53):
So the training side of things is something we're for this.
They trained on their cluster of a cluster of computers
for about two months. So what I can tell you
is that there are a ton of AI startups and
even like big AI companies like open Ai or Microsoft,
you better believe they're gonna be having all hands on
(06:14):
deck today saying hey, can we do something similar. Let's
try to get it out in the next couple months,
And so we're gonna know in the next probably I
would say by May or June, whether you can actually
replicate the process for anywhere close to the cost that
they've done, Because this is a cost thing. If you
are a big you know enterpriser, if you're any company
(06:35):
that's considering using you know, AI technology, and someone comes
and says, hey, I can give you like ninety nine
percent of the performance for one percent of the cost,
you're like, where where do I sign up? Sign me
up for it now?
Speaker 3 (06:46):
For maybe the average person we're listening right now, especially
if you're not paying for chat GPT, this might not
feel all that relevant to you correct the enterprise.
Speaker 2 (06:55):
But if you're not paying, they don't care because they're
making no money off you if if you want to
go and subscribe right now.
Speaker 3 (07:00):
Let's say you're a small business and you want to
buy an enterprise license of chat GPT and their latest models.
I think you're looking at between twenty five to thirty
dollars per month per user, and they're not really making
any money on that per se. But just to give
somebody a framework here for if you're actually paying for
this thing, what does it cost? And that's where the
(07:22):
pricing is coming in right now for like an enterprise
version of chat GPT.
Speaker 2 (07:26):
That's not even really where the money is though. It's
really in Hey, like if you're someone who's paying in
these you know, million input tokens and things like that.
Again I'm using too much, I know, jargon and everything.
The real money is, Hey, I want to pay you
just to send a bunch of data away and have
you send it back to me, not for the off
the shelf license. So what we do know is bunch
(07:51):
of companies offer benchmarks for AI models. How strightly can
you do this? How well? How accurately can you do that?
Speaker 3 (07:58):
So maview on thousand math equations and how often do
they get amound? Yes?
Speaker 2 (08:03):
Here is the LSAT for last year, what's your score
and how long does it take you to do it? Basically,
except for computers, what we have here is on the
six biggest benchmarks that are out there. This is basically
within one to two percent of the one model that's
chat GPTs.
Speaker 3 (08:20):
Chat GPT's kind of best in class, right correct.
Speaker 2 (08:24):
And the other thing that's interesting about this deepseek they
made it an open source model, so unlike chat GPT
where everything is held proprietary, like you can't see how
the black box works. This they published the code. They said,
this is how it works. Here it is. And the
fascinating thing so a couple things that they do differently
based on what the actual code shows and how it
(08:45):
actually works. First, traditional AI model generative AI. To this point,
I say traditional. They've been around for two freaking years.
What's traditional about that? They work? Each word is a
different input the pause, cat, pause, saw pause, the pause
dog open, Aye says okay, we're gonna read each of
(09:06):
these in order to figure out what's going on this.
They figured out a way to do it in one
hole chunk the cat, saw the dog, and in doing so,
instead of having to query your database, five times. You
can do it once there. The other thing that they
did is chat GPT is an example. It's got like
six hundred and eighty billion parameters and it basically inputs
that are there where if you throw something at it,
(09:28):
it runs through six hundred and eighty billion things to
figure out what to say back to you. Deepseek built
what's called an expert system where if you ask a
question about math, it says, Okay, here are the terms
we've categorized as math. We're gonna pull from that subset,
not from everything, because we don't need to access the
whole database, and so it's more efficient.
Speaker 3 (09:46):
I don't need to access eighteenth century poetry if I'm
trying to answer a physics.
Speaker 2 (09:50):
Correct Who cares about Dante? If you want to know
why equals mc squared, he didn't know either. So ultimately,
what this gets at is while this is running in
the cloud on and you can pull up like the
deep Seep deep Seek app on your phone, if you
go and download the actual model, and like you got
to do a little bit of work. You can't just
(10:12):
like download this for a Windows computer and be fine.
You can run this on a high end gaming computer.
Speaker 3 (10:19):
And but what makes that special, because that's confusing, Yes,
because I've got chat GPT on my phone.
Speaker 2 (10:25):
If what makes it special is right now, if you
put something into chat GPT or into the deep Seak app,
it goes to those servers and it gets stored on them,
and it goes out to the company they have what
you are asking it. If you run it just on
your computer, it doesn't go anywhere. It's private. It's held
(10:48):
entirely internally. And you can see this because the code
it's not sending anything back to deep seek you know,
home base and everything.
Speaker 3 (10:56):
So let's say you work for a heavily regulated industry
like the defense industry or the securities industry, where there
are frequent questions asked about your cybersecurity, about what sort
of tools you use, what the output is, where it
gets stored. When you're using chat GPT, there may be
a whole bunch of restrictions because you can't define to
(11:18):
your regulator or to your company. Hey, this is how
I know the information was processed, this is where it's stored,
and I have certainty that chat GPT won't get hacked
in my data won't go anywhere. Sure, you don't need
otherwise you have to build this whole thing in your
own mainframe and store it at raytheon fours.
Speaker 2 (11:36):
And you can't do that because you don't, like, you're
not going to spend like ten billion dollars doing that.
It doesn't make sense for you to the this you
can run on a high end gaming computer. You could
spend ten grand on it, and you can run it internally. Now,
the DoD probably still gonna be like, don't use deepseek
because it still as a Chinese you know, developed.
Speaker 3 (11:55):
So like, let's talk about the technology, yes, and the
technology seems to be that I have a model that
can be run with a fraction of the computing costs
and therefore no need to have this massive processing power.
Speaker 2 (12:09):
And it can. By the way, there are people out
there that have done this in the last week and
are like, oh, bleep, this changes everything. And these are
like smart people that actually know what they're talking about,
unlike me, who's repeating what I've read from smart people.
So why does this actually matter? Quick Break will fill
you in when we return.
Speaker 1 (12:29):
Breaking business and financial news first throughout the day only
here on the Financial Exchange Radio Network. Miss any of
the show, you can catch up at your convenience by
visiting Financial Exchange Show dot com and clicking the on
demand icon, where you'll find all of our interviews and
full shows. This is your home for the latest business
and financial news in New England and around the country.
(12:51):
This is the Financial Exchange Radio network.
Speaker 4 (12:55):
The Financial Exchange is a proud partner of the Disabled
American Veterans to I'm going to Massachusetts. By making a donation,
you can support the Veteran Advancement Program, which offers permanent
and transitional affordable housing to veterans and their dependence. The
DAV of Massachusetts is the only DAV department in the
country to lead a veterans housing initiative for single vets
(13:17):
in those with families. Visit DAV fivek dot Boston to
make a donation today and make NOE for this year's
five K, to be held on Saturday, November eighth. That's
DAV fivek dot Boston.
Speaker 2 (13:29):
All right. So during the first segment, we were talking about deepseek,
which is an AI platform that just released its most
recent iteration about a week ago and posted some really
impressive results and claiming to do so at very minimal cost.
Why does this matter and why are tech stocks getting
(13:51):
bludgeoned today, that's right, bludgeon Yeah. The answer is this,
The entire AI industry in the United States is based
on one idea in one idea alone. If we build
more powerful computers, we will get better models, and it's
willing to pay exponentially more in order to do so. Meaning, hey,
(14:13):
in order to get a model that's like ten percent
better right now, you need to spend fifty percent more.
Speaker 3 (14:18):
So, for example, a company like Nvidia has benefited greatly
from the fact that people want to build out the
latest and greatest models and until today believe that they
need the latest and greatest chips in order to do so.
Speaker 2 (14:32):
Soatnianadell at Microsoft says, hey, you know what, I got
to increase my AI capex from fifty five billion to
eighty billion this year in order to do so. Now,
if it gets proven, hey you don't need to do this,
and there is a different path to building capable, powerful AI,
which again I'm gonna say, the jury is still out
on simply because this is like Daisier. What we have
(14:54):
right now is the product that is legit, whether the costs,
it is legit as one that we don't know yet
just because when it comes to the actual pricing and
cost of an AI platform built out of China, you
just have to look at it and say, am I
really seeing everything? It's a black box, you don't know.
But here's why the entire basically AI ecosystem is being
(15:19):
hit today. And by AI ecosystem, I don't just mean
in Vidia. If you look at this, it's in Vidia,
Microsoft Broadcom, who makes basically a whole bunch of memory
and other stuff that goes in there. Oracle who is
going to build out the stargate platform, Amazon, Google, utility stocks. Okay,
you got utility stocks getting absolutely crushed today. Why utility stocks, chuck? Well,
(15:40):
if you don't need to build giant or dinner centers,
Denner Center, I do like the word gianter. I'm gonna
start using that. If you don't need to build Gianter
data centers nuclear power plants to power them, then Constellation
Energy doesn't need to turn three mile island back on.
And so Constellation is down nineteen percent today. This, this
is the entire ecosystem that is getting absolutely whacked as
(16:04):
a result of this. So I think my view here is, Look,
I've wondered for a while Hey, is this AI boom
just a CAPEX boom in a different rapper is like,
is it going to have the same features? And the
answer today is now, for the first time in the
last two years, the answer is maybe instead of yes.
(16:24):
And so that's what's being priced in today is kind
of the first iteration of maybe if over the next
four to six months, we have other companies in other
parts of the world, the US, Europe, wherever that come
out and say yeah, we've been able to do something similar. Guys,
I got everything on its head. Then I got to
tell you the entire data center buildout model.
Speaker 3 (16:48):
All that stuff we've been talking about about needing to
modernize and revolutionize America's electricity generation. How you know, I
was quite optimistic that, hey, yeah, this might result in
higher prices for electricity in the short term. Long term,
you can have massive capacity built out and all these
other features gone. Now that's the other thing's no longer
relevant if this is true. Other things to pay attention to.
Back in twenty fourteen through twenty sixteen, Russia decided, hey,
(17:11):
we don't want us fracking companies making a whole bunch
of money selling oil at ninety a barrel.
Speaker 2 (17:15):
So what did Russia do? They said, Hey, we're gonna
turn on the spigots and just throw all the oil
at the problem. Oil prices went to thirty a bunch
of US frackers went out of business, and Russia said,
ha ha, gotcha. One question that I do have on
this again, the cost piece is the one that really matters.
Is China just throwing a bunch of money away saying, look,
we'll lose money on AI forever, Like, we don't care
(17:39):
if we make money on this, but we're gonna try
to torpedo USAI development through this very much possibility. And
if that's the case, then I think very quickly this
is gonna become apparent and it'll be okay, fine, We're
just not gonna pay for you know that that model.
And here's the really interesting thing. Remember the way that
most people access deep seek is not through setting up
their own you know, serve is running it, but they
(18:00):
downloaded in the app store. Mike, is there a law
on the books about national security as it relates to
Chinese based phone apps? Have you ever heard of the
application TikTok I have? And so I do wonder does
the US at some point try to ban deep Seek.
(18:21):
It like one of the first things that came to
It's like it follows most of the same framework. Right.
Speaker 3 (18:27):
Do you have the ability to influence American's opinions on
large subjects? Do you have the ability to track information
that gets downloaded onto your phone when you use the
deep seak applet? All of these same things that we
were concerned about with TikTok would apply to deep Seak hypothetically.
Speaker 2 (18:41):
One and so is it only a matter of time
before deep seat gets banned? The question that remains then,
if it does, what does that do to AI development
in the US? Do you still see the bigger is
better framework for how we have to develop? Or do
people still try to replicate what deep seek has done here?
(19:03):
I mean even without deep seek existing in the US.
They are real questions on this. So, ultimately, things that
I think we can take away from this because we
still don't know a ton. First person through the wall
not always the winner. Number two. In the tech industry,
generally hardware tends to get commoditized over time. That might
(19:25):
be the case here in the third one. Remember China
hasn't been able to buy legally a lot of these
chips for a while. Scarcity might force innovation. We still
don't know, because we don't really know how they developed it.
But we'll see quick break. Wall Street watches next.
Speaker 1 (19:41):
Like us on Facebook and follow us on Twitter at
TFE show. Breaking business news is always first right here
on the Financial Exchange Radio Network. Time now for Wall Street.
Watch a complete look and what's moving markets so far
today right here on the Financial Exchange Network.
Speaker 4 (20:02):
Well, markets are selling off as a result of a
significant drop in the tech sector. Wall Street's also readying
for the ramp up of earning season this week, a
FED meeting on Wednesday and a GDP reading on Thursday.
Right now, the Dow is off by only twenty three points,
SMP five hundred selling off by one and a half percent,
or eighty nine points, and the tech heavy NASDAC down
(20:23):
nearly two and a half percent, or four hundred and
seventy nine points of Russell two thousand is off by
about a quarter percent. Ten year treasure yield is down
by six basis points at four point five six percent,
and crude oiled down by two thirds of a percent,
trading just above seventy four dollars a barrel. Today's tech
sell off is driven by Deep Seek, a Chinese AI
(20:43):
company that has high performance ratings and is less expensive
than well known AI in tech names with high valuation.
So with that, several tech names are selling off today.
Nvidia down by thirteen percent, on track for its its
worst day since March of twenty twenty, AMD down by
four percent, Microsoft three percent lower, Amazon one percent lower,
(21:04):
Broadcom down by thirteen percent, and energy sector taking a
little hit as well. Constallation Energy down by twenty percent
at the moment.
Speaker 2 (21:13):
Outside of tech.
Speaker 4 (21:14):
Shares in US Steal down by one percent following news
that an activist investor is preparing to wage a proxy
battle seeking to oust the steelmaker's chief executive and drop
litigation to salvage emerge merger with Japan's Nippon Steel. Meanwhile,
AT and T shares up by six percent after the
telecom giant posted better than expected quarterly earnings and revenue
(21:36):
while also adding more subscribers than forecasted, and shares in
SOFI down by nine percent after the financial service platform
beat earnings and revenue estimates.
Speaker 2 (21:47):
For the fourth quarter.
Speaker 4 (21:48):
However, the company offered disappointing outlook for twenty twenty five
I'm Tucker Silvan. That's Wall Street Watch Mike.
Speaker 2 (21:56):
I think we've done enough AI talk for today. We
might get back to it later, but yeah, there was
a lot, you know. Yeah, we embraced our inner nerd
for the better part of the last half hour, and
now it's time to talk about something not so nerdy,
Earning season. Oh wait, that's adequately aren't nerdy? Yeah? Okay, anyways,
(22:17):
Earning season really kicks into high gear this week with
some big big names reporting, including Tuesday is Defense Day
is what I've dubbed it, raytheon Boeing, Lockheed, Martin and
Royal Caribbean Cruises Striker. I was making a Royal Caribbean jo.
Starbucks two I appreciate, Yeah, Starbucks two Wednesdays where you
(22:38):
have the real heavy hitters, Microsoft, Meta, and Tesla all
in a day, so you get six trillion dollars in
market cap there. You also have T Mobile, IBM, Progressive.
Like again, there's a ton of companies reporting Thursday. You
got Apple and Visa, so it's another four trillion dollars
in market cap. Master Card as well, Yep, Caterpillar, Blackstone Ups.
(22:58):
So like a lot of names coming through there and
then we finished the week with some oil companies Exon, Chevron,
Phillip sixty six all reporting on Friday.
Speaker 3 (23:07):
So all told, one hundred and four s and p
five hundred companies reporting this week. It's the second busiest
earnings week of the season.
Speaker 2 (23:14):
By next week.
Speaker 3 (23:15):
And shockingly, I I hesitate to say this, but let's
imagine for a moment here that Microsoft absolutely knocks the
cover off the ball in this earnings report. Yeah, I'm
not sure investors care, given the deep seaknowledge.
Speaker 2 (23:31):
This is the question on This is back in the
bursting of the tech bubble. The first signs that it
was bursting in two thousand was you had companies, you know, Yahoo,
come out, hey we beat earnings by you know, a
couple pennies and the stock sold off five percent like that.
That was kind of the sign that you got. So
(23:52):
I think that just as much as the actual earnings
from the companies are going to be important, this is
one of those cases where the price action afterwards, to
tell you what investors are thinking about them, is going
to be even more important than normal. I know, everyone
obviously buys stocks for you know, they want them to
go up, So the price action matters, But I think
(24:13):
there's some real signal in the price action here depending
on how these stocks move after they report earnings. So again,
we're gonna know a lot more after the next two weeks,
just because basically most of the S and P five
hundred is done reporting. Then the only major company that's
left afterwards is in the video, which is always a
few weeks later. But this is, uh, it's gonna be
(24:36):
a meati week as far as earnings. And by the way,
have either of you looked at the economic data that
we're getting this week.
Speaker 3 (24:43):
Well, we have a FED meeting on Wednesday, meeting on GDP,
GDP Thursday, PCE Friday, So I mean this has a
crazy week. Before we knew it about a Chinese AI company,
you could.
Speaker 2 (24:54):
Live for like three years and not half a week
like this.
Speaker 3 (24:59):
Yeah, So look, there is going to be plenty of
information to digest. But I think the question especially for
Microsoft and Meta as well as Apple, but to a
more limited extent because quite frankly, they just haven't done
anything themselves on artificial intelligence. But the question for these
two is what does this news mean to you? Right
(25:23):
beside all the earnings, Like, what does this deepseek news
mean to you and how are you going to need
to change your business models to adapt to it.
Speaker 2 (25:32):
I'm sorry, I was just distracted by something on Bloomberg.
The New York City Taxi Commission just revised who's able
to provide insurance for taxis and they no longer require
the insured to prove that they're solvent. It's a good idea.
What's the worst that can happen? Which speaks to one
(25:52):
of two things. Either A, everyone solvent in the insurance
industry and you shouldn't have to worry about it, or
B you had real trouble attracting insurance money into the
New York City taxi insurance market and you're trying desperately
to get someone to ensure it, even if you don't
really trust their financials.
Speaker 3 (26:10):
Yeah, I mean, I have to believe that in the
state of New York, you cannot issue insurance if you're insolvent.
Like maybe it's just duplicative, but I don't think Bloomberg.
Speaker 2 (26:19):
Would be covering if it were just Anne. Why would
the requirement be there? Then? I have no idea wild Anyways, Yeah,
it's going to be a busy couple weeks, but yeah,
the earnings that we are going to be seeing this week. Remember, ultimately,
earnings are what drives stocks in the long run and
the short term. It can be a whole bunch of
other things, but ultimately it's earnings that drive them. If
(26:40):
you will get the long term performance of US stocks,
the vast majority of it, about seventy seventy five percent
of it is driven by the fact that earnings for
the last eighty years have grown at a compound annual
growth rate of almost seven percent, And so that is
what drives markets in the long run and the short term.
It can be I'm afraid about US, and so I'm
(27:00):
selling my stocks. I you know, China came out with
a new AI and so I'm selling my stocks the
shrimp er bat. So whatever it is can move it
in the short term. The long run, it's all about earnings.
And so I think on these calls, the tech ones
in particular, no one's really gonna care what Microsoft reports
(27:21):
for last quarter earnings. Tell us what you're seeing now,
what kind of demand are you seeing for AI? And
specifically what you're going to get is these are going
to be legitimate calls. Have you noticed any change in
the last ten days since deepseek was announced, right, and
since it seeks new version was announced, And like, how
(27:44):
quickly are you able to pivot?
Speaker 3 (27:46):
You've made a massive investment in partnership with the leading
company open ai, the creator of chat GPT. You've already
done some pivoting Microsoft, that is over the last you know,
few months here in terms of more development in house.
So does that put you out of a competitive advantage
to some others out there to be able to shift
your thinking on how you are developing products with artificial intelligence? Again,
(28:11):
like to your point, what happened in the last quarter,
I'm not gonna say it doesn't matter, but what everyone
doesn't it's not all that important in comparison to Hey,
if this thing's real, how are.
Speaker 2 (28:24):
You going to know that it's real? First? Microsoft? And
what are you going to do about it? Because the
big thing is, Remember, the whole AI race is not
just you know, can we build this? But ultimately it
has to make you money at some point. You can't
just spend it forever. Right, And we've called I.
Speaker 3 (28:39):
At least have called Microsoft the low hanging fruit in
terms of who should be able to use artificial intelligence
to drive up prices for their consumers and deliver a
usable product. Right, there's so many stupid, silly things that
we do in all of those products that are annoyances
that AI should be able to do for you, that
companies would pay a lot of money for it to
work well.
Speaker 1 (28:59):
And so.
Speaker 3 (29:02):
Like, let me tell that's been the low hanging fruit
in my mind is Microsoft Develop and Excel program that
allows me to do what I do now, but at
a fraction of the time.
Speaker 2 (29:12):
I'm someone I've had a Microsoft Copilot license for a
few months now. Yeah, it's been hot garbage for what
I understand. What's really interesting to me is that if
I have a summary of something, it can't summarize the summary.
Let's say that I have a bulleted list, like an
outline of topics and I say put this into PowerPoint.
(29:36):
It doesn't really do it well. It actually takes like
longer and it doesn't obey my instructions the way I want.
As an example, I had an eight to ten page
outline and I said, make me ten PowerPoint slides that discuss,
you know, the key topics that are in these this
word document. It made me thirty five I wanted ten.
(30:01):
I say again, no, do this again, but only with
ten slides. It made me thirty five. It's it's not
up there right now, and that's four hundred dollars a year.
So I'm very likely right now probably gonna be like
by Felicia, like I'm not, you know, gonna pay for
something that's not working. Who's Felicia? Just move on by
(30:22):
Felicia ice Cube? Got it? No? Sorry, okay? Is that Friday?
Speaker 4 (30:27):
I don't I hate that saying.
Speaker 2 (30:29):
Oh sorry, well, I won't say it anymore. I thought
this was pretty I was trying to be like all
hip and Edgy and Harry. I'm just alienating the It's
not rap, it's a movie. Oh okay, sorry, and I
think it's like two thousand and one. Thanks anyways, gosh
uh yes. So we have a whole bunch of earnings
(30:49):
that are going to be coming out this week, and
with the big tech ones, I do think it's gonna
be much more forward looking than just hey did you
beat on your earnings numbers? And you better believe a
month from now when in video reports, they're gonna have
some real questions. What kind of demand have you seen
in the last month? You know, are you seeing anyone
(31:11):
canceling orders? Are you seeing people downsizing orders? Yeah, that's
gonna be where the real signal is in terms of
you know, how activity might actually be shifting. Just take
a quick break. When we come back, let's see do
we want to do fed talker credit cards, draft credit cards?
Let's talk credit cards.
Speaker 1 (31:31):
Right after this, find daily interviews and full shows of
the Financial Exchange on our YouTube page. Like us on
YouTube and get caught up on anything and everything you
might have missed. This is the Financial Exchange Radio Network.
Miss any of the show. Catch up in your convenience
by visiting Financial Exchange show dot com and clicking the
on demand icon, where you'll find all of our interviews
(31:54):
and full shows. This is your home for the latest
business and financial news in New England and around the country.
This is the Financial Exchange Radio Network.
Speaker 2 (32:12):
Wall Street Journal's got a piece of day title. Americans
are carrying bigger credit card balances and they are now.
Part of this is because of inflation, So on an
inflation adjusted basis, the difference is not that big. What
is really interesting to me is and there's data here
that this had some of it. But I was actually
(32:32):
looking at this last week so thanks Wall Street Journal
for looking at the same report I did. So if
you go back to the early part of twenty twenty,
just before the pandemic Q one to twenty twenty, about
just under ten percent of credit card holders were making
only the minimum payment. During the pandemic, it dropped down
(32:53):
to about eight percent because of, Hey, here's a bunch
of stimulus money. I go forth and do whatever. And
today you're closing in on about eleven percent of credit
card users are only making the minimum payment.
Speaker 3 (33:08):
That's the more interesting piece to me, just because if
I'm doing the math here, So as of Q three
of twenty four, which is when this data was put together,
outstanding balances revolving credit card balances six hundred and forty
five billion dollars, which is up massively. That's up huge
from the first quarter of twenty twenty five hundred and
fifty three billion. That's a sixteen point seven percent increase.
(33:32):
Sure inflation's twenty right, So the balance piece is not relevant. No,
what is relevant is the fact that interest rates are
substantially higher now. And to your point, that other data
point of who is not paying off those balances that's growing.
Speaker 2 (33:47):
I have one other data point that's not talked about
in this piece though. Do you know what percentage of
Americans in twenty twenty Q one, twenty twenty, right before
the pandemic, we're paying the full balance off each month? No,
it's about thirty one percent. Okay, today it's about thirty five.
(34:09):
K shaped recovery is back. Now. Remember K shape. Everyone
was talking about what shapes the recovery gonna be? Here
we go. The K shaped recovery says this. The people
at the bottom end of the wage spectrum doing worse,
evidenced by more a greater percentage of people making only
the minimum payments on their credit cards, but the high
(34:30):
end of the wage spectrum cranking, especially those that might
have investment accounts and things like that that have gone
up and you know, maybe they're making their they're retired
in their you know, spending is based on, hey my
portfolio is up. Yeah, they are paying off their credit
card balances and full. The remaining twenty eight percent of
people okay, they're paying more than a minimum but less
(34:52):
than full or whatever the percentages. I did the math badly,
but you get what I'm saying.
Speaker 3 (34:57):
There's a lot of stuff that comes out of Dave
Ramsey's that I don't always agree with. The more and
more I've studied the credit card issue, the more I've
come to his way of thinking. His way of thinking,
he's never used any sort of data, you know, pay
off all sorts of debt other than your home. And look,
I have personally never carried a credit card balance, and
so credit cards are great for me. But his point is, look,
(35:19):
that's not the average experience. And even when people say
they're going to pay off the credit card balance, they
usually don't. And the fact of the matter is that
he's right on this. So over half of credit card
holders over the last twelve months had a point in
time where they couldn't pay off the full balance.
Speaker 2 (35:34):
So could the Mike Armstrong corollary be, if you do
not have the money in a bank account to pay
off your credit card balance, then you shouldn't put it
on the credit card. Frankly, even then, I'm not sure
it's adequate.
Speaker 3 (35:47):
Interesting, yeah, because how many times have we heard of
you know, I had the money in the bank, but
then my car broke down and I needed a three
thousand dollars a pair in there goes the money that
I had in my bank account, and now I'm sitting
on the credit card. What about credit balance? I don't
know that there's a thresh. I don't know that there's
a number. Because it comes back to human behavior, and
(36:08):
what we seem to know about human behavior is that
most of them run a credit card balance month after
month and eat away at frankly any of the benefits
that come with the credit card, which is disappointing because
there are a lot of benefits. Yeah, you've got a
bunch of cash back rewards, you've got a bunch of
protection from fraud, but those benefits seem to be outweighed
(36:30):
by the average credit card user experience.
Speaker 2 (36:32):
So can we then amend Benjamin Franklin's famous saying, here
we go to be nothing is certain except debts and taxes. Yeah?
I like that? Is that where you'd like to go?
Speaker 4 (36:42):
I appreciate that the calendars turned and the cold is
here to stay. So now's the perfect time to plan
that getaway to the US Virgin Islands. Act now and
take advantage of a special promotion to Saint Croix, experience
a vibe like no other, and receive a two hundred
and fifty dollars per person airfare credit and a free
hotel night when you book a five night minimum stay
with any participating a hotel before June thirtieth. San Croix
(37:04):
is known for its incredible history and spectacular architecture, along
with world class cuisine, incredible golf courses, and a wide
variety of water sports. From the moment you arrive, you'll
fall naturally in rhythm with the heartbeat of the islands.
There's no money to exchange, and travel from New England
could not be easier. America's Caribbean paradise is waiting for you.
Experience a vibe like no other on the island of
(37:24):
San Croix. Go to visit USVII dot com, use the
code Vibe twenty twenty five for more information and book
your trip today. That's visit USVII dot com and code
Vibe twenty twenty five.
Speaker 2 (37:36):
Airlines are charging higher affairs and are confident that you'll
pay up. That's because there's minimal competition in the space,
and as such, if you want to fly on probably
like eighty percent of routes throughout the United States, you
have one or two options, and generally they come as
close as you can to price fixing without going over
that line.
Speaker 3 (37:51):
And one of the famous low cost providers out there
is in the midst of bankruptcy, so they have a
fair bit of evidence that this model works. Yeah, monopoly
tends to, you know, like go, let's be honest here,
pipoly Chuck all right, triopoly, triopoly, quadopoly, there's there's four
big ones Southwest, American United Delta.
Speaker 2 (38:10):
Right, monopoly. It's studied in economic text textbooks. Let's take
a quick break here, and when we return, we got
our two coming up.
The Financial Exchange is produced by Money Matters Radio and
is hosted by employees of the Armstrong Advisory Group, a
registered investment advisor. All opinions expressed are solely those of
the hosts. Do not reflect the opinions of Armstrong Advisory
or anyone else. Investments can lose money. This program does
not offer any specific financial or investment advice. Please consult
your own financial, tax, and estate planning advisors before making
(00:20):
any investment decisions. Armstrong Advisory and the advertisers heard on
this program do not endorse each other or their services.
Armstrong and Money Matters Radio do not compensate each other
for referrals and are not affiliated. This is the Financial
Exchange with Chuck Zada and Mike Armstrong, Your exclusive look
at business and financial news affecting your day, your city,
(00:42):
your world. Stay informed and up to date about economic
and market trends plus breaking business news every day. The
Financial Exchange is a proud partner of the Disabled American
Veterans Department of Massachusetts.
Speaker 2 (00:56):
Help us support our.
Speaker 1 (00:57):
Great American heroes by visiting dav five K Boston and
making a donation today. This is the Financial Exchange with
Chuck Zada, and Mike Armstrong.
Speaker 2 (01:08):
Chuck, Mike and Tucker with you here. And we've got
a sell off happening in markets today, but really the
sell off is concentrated primarily in big tech. And the
reason why, if you have not heard of the AI
platform deep Seek, you're going to hear about it right now.
If you have heard about it, then you might know
(01:29):
some of what we're talking about. But quite honestly, we
think that we are very well versed on this topic,
and so you're going to learn some new stuff from us,
you better believe it.
Speaker 3 (01:37):
And if you had heard about deep Seek prior to
five days ago, I'd be particularly interested because because nobody.
Speaker 2 (01:45):
Because you hadn't. Let's be honest. So what is deep Seek.
It is an AI platform that is built by a
Chinese company by the name of high Flyer. High Flyer
is a hedge fund. And why did you Flyer go
and build deep Seek? Now? The reason why They had
been building internal AI processes for their investment platform, trying to,
(02:10):
you know, find better ways to beat the market investing,
and as part of that, they just had a natural
interest in kind of the latest and greatest AI technology,
and so they started building AI models a couple of
years ago. The one that they released on January twentieth
is called deep seek are one and the premise of
this is that it competes directly with open AI's flagship
(02:32):
open ai Chat GPT oh one product that came out
I want to say, in like September or October, and
now Chat GPT's been around for two years.
Speaker 3 (02:42):
Yeah, a little more than two years and it's public
facing will but this latest iteration of their large language
model is you said, one oh one.
Speaker 2 (02:50):
And the big thing that one claimed to be able
to do and that it can do is it can
handle more logical reasoning things. So you ask it, you know,
logic based quests, and it can actually kind of think
through them in certain respects. And this platform, the deep
seek are one platform, does the exact same thing. So
why does this matter? A couple reasons. The first is
(03:15):
deep Seak claims that they built this using about five
point six million dollars of what they call AI training,
basically in putting all the data into the model and
getting it to understand what it is. Sounds like a
lot of money, Chuck, is it no, Because most AI
models that are produced in the US right now cost
anywhere from about five hundred million to a billion dollars
(03:35):
to do the training aspect.
Speaker 3 (03:36):
Now keep in mind, this is the training part where
they're feeding at internet articles and the entire history of
the New York Times and Facebook and Twitter and all
these different things. This is just purely teaching the thing
how to read and think and do everything correct the
AI models can do.
Speaker 2 (03:54):
And basically the training costs is a function of two things.
How many computer chips do you need to train on
and how long do you need to do that for?
Those are the two things that matter, because you're you're
basically paying for the compute time. What you are then
doing on the the other end of that is after
the training is completed, hey, you still need to pay
for the processors to go and actually execute the queries
(04:17):
that are asked of the model. So that's why if
you're using open AI's enterprise platforms or they're you know,
the their high end stuff for one and again, these
numbers are going to be mean nothing to you. But
don't worry about it. I'll make it make sense, I promise.
If you need inputs from open ai, it costs fifteen
dollars per million inputs for outputs, it costs sixty per
(04:38):
million deep seek Remember open as fifteen per million inputs.
Deep seek is fourteen cents per million inputs one percent
of the cost. On the other side of things, outputs
open AI is sixty per million outputs. Deep seek is
two dollars and nineteen cents.
Speaker 3 (04:53):
Okay, So we've basically now understood from this company's own
I wouldn't even say disclosures, but yeah, information that they've
put out about themselves is that training and getting output
out of this large language model costs one percent on
the training side and three percent of chat GPT's own
(05:15):
disclosed information on the output side of things. I would say,
Chuck that this is one area that we probably have
the least degree of certainty about this new model. Right, Like,
it's a company held out of China, So could there
be some accounting creativity outright lies a whole bunch of
subsidies from the Chinese government. All of those things are
(05:38):
entirely possible. But what they are claiming is that, yeah,
the cost to both run and train this thing, we're
one to three percent of what we've been experiencing in
models here in the United States, and getting we'll get
to the results side.
Speaker 2 (05:53):
So the training side of things is something we're for this.
They trained on their cluster of a cluster of computers
for about two months. So what I can tell you
is that there are a ton of AI startups and
even like big AI companies like open Ai or Microsoft,
you better believe they're gonna be having all hands on
(06:14):
deck today saying hey, can we do something similar. Let's
try to get it out in the next couple months,
And so we're gonna know in the next probably I
would say by May or June, whether you can actually
replicate the process for anywhere close to the cost that
they've done, Because this is a cost thing. If you
are a big you know enterpriser, if you're any company
(06:35):
that's considering using you know, AI technology, and someone comes
and says, hey, I can give you like ninety nine
percent of the performance for one percent of the cost,
you're like, where where do I sign up? Sign me
up for it now?
Speaker 3 (06:46):
For maybe the average person we're listening right now, especially
if you're not paying for chat GPT, this might not
feel all that relevant to you correct the enterprise.
Speaker 2 (06:55):
But if you're not paying, they don't care because they're
making no money off you if if you want to
go and subscribe right now.
Speaker 3 (07:00):
Let's say you're a small business and you want to
buy an enterprise license of chat GPT and their latest models.
I think you're looking at between twenty five to thirty
dollars per month per user, and they're not really making
any money on that per se. But just to give
somebody a framework here for if you're actually paying for
this thing, what does it cost? And that's where the
(07:22):
pricing is coming in right now for like an enterprise
version of chat GPT.
Speaker 2 (07:26):
That's not even really where the money is though. It's
really in Hey, like if you're someone who's paying in
these you know, million input tokens and things like that.
Again I'm using too much, I know, jargon and everything.
The real money is, Hey, I want to pay you
just to send a bunch of data away and have
you send it back to me, not for the off
the shelf license. So what we do know is bunch
(07:51):
of companies offer benchmarks for AI models. How strightly can
you do this? How well? How accurately can you do that?
Speaker 3 (07:58):
So maview on thousand math equations and how often do
they get amound? Yes?
Speaker 2 (08:03):
Here is the LSAT for last year, what's your score
and how long does it take you to do it? Basically,
except for computers, what we have here is on the
six biggest benchmarks that are out there. This is basically
within one to two percent of the one model that's
chat GPTs.
Speaker 3 (08:20):
Chat GPT's kind of best in class, right correct.
Speaker 2 (08:24):
And the other thing that's interesting about this deepseek they
made it an open source model, so unlike chat GPT
where everything is held proprietary, like you can't see how
the black box works. This they published the code. They said,
this is how it works. Here it is. And the
fascinating thing so a couple things that they do differently
based on what the actual code shows and how it
(08:45):
actually works. First, traditional AI model generative AI. To this point,
I say traditional. They've been around for two freaking years.
What's traditional about that? They work? Each word is a
different input the pause, cat, pause, saw pause, the pause
dog open, Aye says okay, we're gonna read each of
(09:06):
these in order to figure out what's going on this.
They figured out a way to do it in one
hole chunk the cat, saw the dog, and in doing so,
instead of having to query your database, five times. You
can do it once there. The other thing that they
did is chat GPT is an example. It's got like
six hundred and eighty billion parameters and it basically inputs
that are there where if you throw something at it,
(09:28):
it runs through six hundred and eighty billion things to
figure out what to say back to you. Deepseek built
what's called an expert system where if you ask a
question about math, it says, Okay, here are the terms
we've categorized as math. We're gonna pull from that subset,
not from everything, because we don't need to access the
whole database, and so it's more efficient.
Speaker 3 (09:46):
I don't need to access eighteenth century poetry if I'm
trying to answer a physics.
Speaker 2 (09:50):
Correct Who cares about Dante? If you want to know
why equals mc squared, he didn't know either. So ultimately,
what this gets at is while this is running in
the cloud on and you can pull up like the
deep Seep deep Seek app on your phone, if you
go and download the actual model, and like you got
to do a little bit of work. You can't just
(10:12):
like download this for a Windows computer and be fine.
You can run this on a high end gaming computer.
Speaker 3 (10:19):
And but what makes that special, because that's confusing, Yes,
because I've got chat GPT on my phone.
Speaker 2 (10:25):
If what makes it special is right now, if you
put something into chat GPT or into the deep Seak app,
it goes to those servers and it gets stored on them,
and it goes out to the company they have what
you are asking it. If you run it just on
your computer, it doesn't go anywhere. It's private. It's held
(10:48):
entirely internally. And you can see this because the code
it's not sending anything back to deep seek you know,
home base and everything.
Speaker 3 (10:56):
So let's say you work for a heavily regulated industry
like the defense industry or the securities industry, where there
are frequent questions asked about your cybersecurity, about what sort
of tools you use, what the output is, where it
gets stored. When you're using chat GPT, there may be
a whole bunch of restrictions because you can't define to
(11:18):
your regulator or to your company. Hey, this is how
I know the information was processed, this is where it's stored,
and I have certainty that chat GPT won't get hacked
in my data won't go anywhere. Sure, you don't need
otherwise you have to build this whole thing in your
own mainframe and store it at raytheon fours.
Speaker 2 (11:36):
And you can't do that because you don't, like, you're
not going to spend like ten billion dollars doing that.
It doesn't make sense for you to the this you
can run on a high end gaming computer. You could
spend ten grand on it, and you can run it internally. Now,
the DoD probably still gonna be like, don't use deepseek
because it still as a Chinese you know, developed.
Speaker 3 (11:55):
So like, let's talk about the technology, yes, and the
technology seems to be that I have a model that
can be run with a fraction of the computing costs
and therefore no need to have this massive processing power.
Speaker 2 (12:09):
And it can. By the way, there are people out
there that have done this in the last week and
are like, oh, bleep, this changes everything. And these are
like smart people that actually know what they're talking about,
unlike me, who's repeating what I've read from smart people.
So why does this actually matter? Quick Break will fill
you in when we return.
Speaker 1 (12:29):
Breaking business and financial news first throughout the day only
here on the Financial Exchange Radio Network. Miss any of
the show, you can catch up at your convenience by
visiting Financial Exchange Show dot com and clicking the on
demand icon, where you'll find all of our interviews and
full shows. This is your home for the latest business
and financial news in New England and around the country.
(12:51):
This is the Financial Exchange Radio network.
Speaker 4 (12:55):
The Financial Exchange is a proud partner of the Disabled
American Veterans to I'm going to Massachusetts. By making a donation,
you can support the Veteran Advancement Program, which offers permanent
and transitional affordable housing to veterans and their dependence. The
DAV of Massachusetts is the only DAV department in the
country to lead a veterans housing initiative for single vets
(13:17):
in those with families. Visit DAV fivek dot Boston to
make a donation today and make NOE for this year's
five K, to be held on Saturday, November eighth. That's
DAV fivek dot Boston.
Speaker 2 (13:29):
All right. So during the first segment, we were talking about deepseek,
which is an AI platform that just released its most
recent iteration about a week ago and posted some really
impressive results and claiming to do so at very minimal cost.
Why does this matter and why are tech stocks getting
(13:51):
bludgeoned today, that's right, bludgeon Yeah. The answer is this,
The entire AI industry in the United States is based
on one idea in one idea alone. If we build
more powerful computers, we will get better models, and it's
willing to pay exponentially more in order to do so. Meaning, hey,
(14:13):
in order to get a model that's like ten percent
better right now, you need to spend fifty percent more.
Speaker 3 (14:18):
So, for example, a company like Nvidia has benefited greatly
from the fact that people want to build out the
latest and greatest models and until today believe that they
need the latest and greatest chips in order to do so.
Speaker 2 (14:32):
Soatnianadell at Microsoft says, hey, you know what, I got
to increase my AI capex from fifty five billion to
eighty billion this year in order to do so. Now,
if it gets proven, hey you don't need to do this,
and there is a different path to building capable, powerful AI,
which again I'm gonna say, the jury is still out
on simply because this is like Daisier. What we have
(14:54):
right now is the product that is legit, whether the costs,
it is legit as one that we don't know yet
just because when it comes to the actual pricing and
cost of an AI platform built out of China, you
just have to look at it and say, am I
really seeing everything? It's a black box, you don't know.
But here's why the entire basically AI ecosystem is being
(15:19):
hit today. And by AI ecosystem, I don't just mean
in Vidia. If you look at this, it's in Vidia,
Microsoft Broadcom, who makes basically a whole bunch of memory
and other stuff that goes in there. Oracle who is
going to build out the stargate platform, Amazon, Google, utility stocks. Okay,
you got utility stocks getting absolutely crushed today. Why utility stocks, chuck? Well,
(15:40):
if you don't need to build giant or dinner centers,
Denner Center, I do like the word gianter. I'm gonna
start using that. If you don't need to build Gianter
data centers nuclear power plants to power them, then Constellation
Energy doesn't need to turn three mile island back on.
And so Constellation is down nineteen percent today. This, this
is the entire ecosystem that is getting absolutely whacked as
(16:04):
a result of this. So I think my view here is, Look,
I've wondered for a while Hey, is this AI boom
just a CAPEX boom in a different rapper is like,
is it going to have the same features? And the
answer today is now, for the first time in the
last two years, the answer is maybe instead of yes.
(16:24):
And so that's what's being priced in today is kind
of the first iteration of maybe if over the next
four to six months, we have other companies in other
parts of the world, the US, Europe, wherever that come
out and say yeah, we've been able to do something similar. Guys,
I got everything on its head. Then I got to
tell you the entire data center buildout model.
Speaker 3 (16:48):
All that stuff we've been talking about about needing to
modernize and revolutionize America's electricity generation. How you know, I
was quite optimistic that, hey, yeah, this might result in
higher prices for electricity in the short term. Long term,
you can have massive capacity built out and all these
other features gone. Now that's the other thing's no longer
relevant if this is true. Other things to pay attention to.
Back in twenty fourteen through twenty sixteen, Russia decided, hey,
(17:11):
we don't want us fracking companies making a whole bunch
of money selling oil at ninety a barrel.
Speaker 2 (17:15):
So what did Russia do? They said, Hey, we're gonna
turn on the spigots and just throw all the oil
at the problem. Oil prices went to thirty a bunch
of US frackers went out of business, and Russia said,
ha ha, gotcha. One question that I do have on
this again, the cost piece is the one that really matters.
Is China just throwing a bunch of money away saying, look,
we'll lose money on AI forever, Like, we don't care
(17:39):
if we make money on this, but we're gonna try
to torpedo USAI development through this very much possibility. And
if that's the case, then I think very quickly this
is gonna become apparent and it'll be okay, fine, We're
just not gonna pay for you know that that model.
And here's the really interesting thing. Remember the way that
most people access deep seek is not through setting up
their own you know, serve is running it, but they
(18:00):
downloaded in the app store. Mike, is there a law
on the books about national security as it relates to
Chinese based phone apps? Have you ever heard of the
application TikTok I have? And so I do wonder does
the US at some point try to ban deep Seek.
(18:21):
It like one of the first things that came to
It's like it follows most of the same framework. Right.
Speaker 3 (18:27):
Do you have the ability to influence American's opinions on
large subjects? Do you have the ability to track information
that gets downloaded onto your phone when you use the
deep seak applet? All of these same things that we
were concerned about with TikTok would apply to deep Seak hypothetically.
Speaker 2 (18:41):
One and so is it only a matter of time
before deep seat gets banned? The question that remains then,
if it does, what does that do to AI development
in the US? Do you still see the bigger is
better framework for how we have to develop? Or do
people still try to replicate what deep seek has done here?
(19:03):
I mean even without deep seek existing in the US.
They are real questions on this. So, ultimately, things that
I think we can take away from this because we
still don't know a ton. First person through the wall
not always the winner. Number two. In the tech industry,
generally hardware tends to get commoditized over time. That might
(19:25):
be the case here in the third one. Remember China
hasn't been able to buy legally a lot of these
chips for a while. Scarcity might force innovation. We still
don't know, because we don't really know how they developed it.
But we'll see quick break. Wall Street watches next.
Speaker 1 (19:41):
Like us on Facebook and follow us on Twitter at
TFE show. Breaking business news is always first right here
on the Financial Exchange Radio Network. Time now for Wall Street.
Watch a complete look and what's moving markets so far
today right here on the Financial Exchange Network.
Speaker 4 (20:02):
Well, markets are selling off as a result of a
significant drop in the tech sector. Wall Street's also readying
for the ramp up of earning season this week, a
FED meeting on Wednesday and a GDP reading on Thursday.
Right now, the Dow is off by only twenty three points,
SMP five hundred selling off by one and a half percent,
or eighty nine points, and the tech heavy NASDAC down
(20:23):
nearly two and a half percent, or four hundred and
seventy nine points of Russell two thousand is off by
about a quarter percent. Ten year treasure yield is down
by six basis points at four point five six percent,
and crude oiled down by two thirds of a percent,
trading just above seventy four dollars a barrel. Today's tech
sell off is driven by Deep Seek, a Chinese AI
(20:43):
company that has high performance ratings and is less expensive
than well known AI in tech names with high valuation.
So with that, several tech names are selling off today.
Nvidia down by thirteen percent, on track for its its
worst day since March of twenty twenty, AMD down by
four percent, Microsoft three percent lower, Amazon one percent lower,
(21:04):
Broadcom down by thirteen percent, and energy sector taking a
little hit as well. Constallation Energy down by twenty percent
at the moment.
Speaker 2 (21:13):
Outside of tech.
Speaker 4 (21:14):
Shares in US Steal down by one percent following news
that an activist investor is preparing to wage a proxy
battle seeking to oust the steelmaker's chief executive and drop
litigation to salvage emerge merger with Japan's Nippon Steel. Meanwhile,
AT and T shares up by six percent after the
telecom giant posted better than expected quarterly earnings and revenue
(21:36):
while also adding more subscribers than forecasted, and shares in
SOFI down by nine percent after the financial service platform
beat earnings and revenue estimates.
Speaker 2 (21:47):
For the fourth quarter.
Speaker 4 (21:48):
However, the company offered disappointing outlook for twenty twenty five
I'm Tucker Silvan. That's Wall Street Watch Mike.
Speaker 2 (21:56):
I think we've done enough AI talk for today. We
might get back to it later, but yeah, there was
a lot, you know. Yeah, we embraced our inner nerd
for the better part of the last half hour, and
now it's time to talk about something not so nerdy,
Earning season. Oh wait, that's adequately aren't nerdy? Yeah? Okay, anyways,
(22:17):
Earning season really kicks into high gear this week with
some big big names reporting, including Tuesday is Defense Day
is what I've dubbed it, raytheon Boeing, Lockheed, Martin and
Royal Caribbean Cruises Striker. I was making a Royal Caribbean jo.
Starbucks two I appreciate, Yeah, Starbucks two Wednesdays where you
(22:38):
have the real heavy hitters, Microsoft, Meta, and Tesla all
in a day, so you get six trillion dollars in
market cap there. You also have T Mobile, IBM, Progressive.
Like again, there's a ton of companies reporting Thursday. You
got Apple and Visa, so it's another four trillion dollars
in market cap. Master Card as well, Yep, Caterpillar, Blackstone Ups.
(22:58):
So like a lot of names coming through there and
then we finished the week with some oil companies Exon, Chevron,
Phillip sixty six all reporting on Friday.
Speaker 3 (23:07):
So all told, one hundred and four s and p
five hundred companies reporting this week. It's the second busiest
earnings week of the season.
Speaker 2 (23:14):
By next week.
Speaker 3 (23:15):
And shockingly, I I hesitate to say this, but let's
imagine for a moment here that Microsoft absolutely knocks the
cover off the ball in this earnings report. Yeah, I'm
not sure investors care, given the deep seaknowledge.
Speaker 2 (23:31):
This is the question on This is back in the
bursting of the tech bubble. The first signs that it
was bursting in two thousand was you had companies, you know, Yahoo,
come out, hey we beat earnings by you know, a
couple pennies and the stock sold off five percent like that.
That was kind of the sign that you got. So
(23:52):
I think that just as much as the actual earnings
from the companies are going to be important, this is
one of those cases where the price action afterwards, to
tell you what investors are thinking about them, is going
to be even more important than normal. I know, everyone
obviously buys stocks for you know, they want them to
go up, So the price action matters, But I think
(24:13):
there's some real signal in the price action here depending
on how these stocks move after they report earnings. So again,
we're gonna know a lot more after the next two weeks,
just because basically most of the S and P five
hundred is done reporting. Then the only major company that's
left afterwards is in the video, which is always a
few weeks later. But this is, uh, it's gonna be
(24:36):
a meati week as far as earnings. And by the way,
have either of you looked at the economic data that
we're getting this week.
Speaker 3 (24:43):
Well, we have a FED meeting on Wednesday, meeting on GDP,
GDP Thursday, PCE Friday, So I mean this has a
crazy week. Before we knew it about a Chinese AI company,
you could.
Speaker 2 (24:54):
Live for like three years and not half a week
like this.
Speaker 3 (24:59):
Yeah, So look, there is going to be plenty of
information to digest. But I think the question especially for
Microsoft and Meta as well as Apple, but to a
more limited extent because quite frankly, they just haven't done
anything themselves on artificial intelligence. But the question for these
two is what does this news mean to you? Right
(25:23):
beside all the earnings, Like, what does this deepseek news
mean to you and how are you going to need
to change your business models to adapt to it.
Speaker 2 (25:32):
I'm sorry, I was just distracted by something on Bloomberg.
The New York City Taxi Commission just revised who's able
to provide insurance for taxis and they no longer require
the insured to prove that they're solvent. It's a good idea.
What's the worst that can happen? Which speaks to one
(25:52):
of two things. Either A, everyone solvent in the insurance
industry and you shouldn't have to worry about it, or
B you had real trouble attracting insurance money into the
New York City taxi insurance market and you're trying desperately
to get someone to ensure it, even if you don't
really trust their financials.
Speaker 3 (26:10):
Yeah, I mean, I have to believe that in the
state of New York, you cannot issue insurance if you're insolvent.
Like maybe it's just duplicative, but I don't think Bloomberg.
Speaker 2 (26:19):
Would be covering if it were just Anne. Why would
the requirement be there? Then? I have no idea wild Anyways, Yeah,
it's going to be a busy couple weeks, but yeah,
the earnings that we are going to be seeing this week. Remember, ultimately,
earnings are what drives stocks in the long run and
the short term. It can be a whole bunch of
other things, but ultimately it's earnings that drive them. If
(26:40):
you will get the long term performance of US stocks,
the vast majority of it, about seventy seventy five percent
of it is driven by the fact that earnings for
the last eighty years have grown at a compound annual
growth rate of almost seven percent, And so that is
what drives markets in the long run and the short term.
It can be I'm afraid about US, and so I'm
(27:00):
selling my stocks. I you know, China came out with
a new AI and so I'm selling my stocks the
shrimp er bat. So whatever it is can move it
in the short term. The long run, it's all about earnings.
And so I think on these calls, the tech ones
in particular, no one's really gonna care what Microsoft reports
(27:21):
for last quarter earnings. Tell us what you're seeing now,
what kind of demand are you seeing for AI? And
specifically what you're going to get is these are going
to be legitimate calls. Have you noticed any change in
the last ten days since deepseek was announced, right, and
since it seeks new version was announced, And like, how
(27:44):
quickly are you able to pivot?
Speaker 3 (27:46):
You've made a massive investment in partnership with the leading
company open ai, the creator of chat GPT. You've already
done some pivoting Microsoft, that is over the last you know,
few months here in terms of more development in house.
So does that put you out of a competitive advantage
to some others out there to be able to shift
your thinking on how you are developing products with artificial intelligence? Again,
(28:11):
like to your point, what happened in the last quarter,
I'm not gonna say it doesn't matter, but what everyone
doesn't it's not all that important in comparison to Hey,
if this thing's real, how are.
Speaker 2 (28:24):
You going to know that it's real? First? Microsoft? And
what are you going to do about it? Because the
big thing is, Remember, the whole AI race is not
just you know, can we build this? But ultimately it
has to make you money at some point. You can't
just spend it forever. Right, And we've called I.
Speaker 3 (28:39):
At least have called Microsoft the low hanging fruit in
terms of who should be able to use artificial intelligence
to drive up prices for their consumers and deliver a
usable product. Right, there's so many stupid, silly things that
we do in all of those products that are annoyances
that AI should be able to do for you, that
companies would pay a lot of money for it to
work well.
Speaker 1 (28:59):
And so.
Speaker 3 (29:02):
Like, let me tell that's been the low hanging fruit
in my mind is Microsoft Develop and Excel program that
allows me to do what I do now, but at
a fraction of the time.
Speaker 2 (29:12):
I'm someone I've had a Microsoft Copilot license for a
few months now. Yeah, it's been hot garbage for what
I understand. What's really interesting to me is that if
I have a summary of something, it can't summarize the summary.
Let's say that I have a bulleted list, like an
outline of topics and I say put this into PowerPoint.
(29:36):
It doesn't really do it well. It actually takes like
longer and it doesn't obey my instructions the way I want.
As an example, I had an eight to ten page
outline and I said, make me ten PowerPoint slides that discuss,
you know, the key topics that are in these this
word document. It made me thirty five I wanted ten.
(30:01):
I say again, no, do this again, but only with
ten slides. It made me thirty five. It's it's not
up there right now, and that's four hundred dollars a year.
So I'm very likely right now probably gonna be like
by Felicia, like I'm not, you know, gonna pay for
something that's not working. Who's Felicia? Just move on by
(30:22):
Felicia ice Cube? Got it? No? Sorry, okay? Is that Friday?
Speaker 4 (30:27):
I don't I hate that saying.
Speaker 2 (30:29):
Oh sorry, well, I won't say it anymore. I thought
this was pretty I was trying to be like all
hip and Edgy and Harry. I'm just alienating the It's
not rap, it's a movie. Oh okay, sorry, and I
think it's like two thousand and one. Thanks anyways, gosh
uh yes. So we have a whole bunch of earnings
(30:49):
that are going to be coming out this week, and
with the big tech ones, I do think it's gonna
be much more forward looking than just hey did you
beat on your earnings numbers? And you better believe a
month from now when in video reports, they're gonna have
some real questions. What kind of demand have you seen
in the last month? You know, are you seeing anyone
(31:11):
canceling orders? Are you seeing people downsizing orders? Yeah, that's
gonna be where the real signal is in terms of
you know, how activity might actually be shifting. Just take
a quick break. When we come back, let's see do
we want to do fed talker credit cards, draft credit cards?
Let's talk credit cards.
Speaker 1 (31:31):
Right after this, find daily interviews and full shows of
the Financial Exchange on our YouTube page. Like us on
YouTube and get caught up on anything and everything you
might have missed. This is the Financial Exchange Radio Network.
Miss any of the show. Catch up in your convenience
by visiting Financial Exchange show dot com and clicking the
on demand icon, where you'll find all of our interviews
(31:54):
and full shows. This is your home for the latest
business and financial news in New England and around the country.
This is the Financial Exchange Radio Network.
Speaker 2 (32:12):
Wall Street Journal's got a piece of day title. Americans
are carrying bigger credit card balances and they are now.
Part of this is because of inflation, So on an
inflation adjusted basis, the difference is not that big. What
is really interesting to me is and there's data here
that this had some of it. But I was actually
(32:32):
looking at this last week so thanks Wall Street Journal
for looking at the same report I did. So if
you go back to the early part of twenty twenty,
just before the pandemic Q one to twenty twenty, about
just under ten percent of credit card holders were making
only the minimum payment. During the pandemic, it dropped down
(32:53):
to about eight percent because of, Hey, here's a bunch
of stimulus money. I go forth and do whatever. And
today you're closing in on about eleven percent of credit
card users are only making the minimum payment.
Speaker 3 (33:08):
That's the more interesting piece to me, just because if
I'm doing the math here, So as of Q three
of twenty four, which is when this data was put together,
outstanding balances revolving credit card balances six hundred and forty
five billion dollars, which is up massively. That's up huge
from the first quarter of twenty twenty five hundred and
fifty three billion. That's a sixteen point seven percent increase.
(33:32):
Sure inflation's twenty right, So the balance piece is not relevant. No,
what is relevant is the fact that interest rates are
substantially higher now. And to your point, that other data
point of who is not paying off those balances that's growing.
Speaker 2 (33:47):
I have one other data point that's not talked about
in this piece though. Do you know what percentage of
Americans in twenty twenty Q one, twenty twenty, right before
the pandemic, we're paying the full balance off each month? No,
it's about thirty one percent. Okay, today it's about thirty five.
(34:09):
K shaped recovery is back. Now. Remember K shape. Everyone
was talking about what shapes the recovery gonna be? Here
we go. The K shaped recovery says this. The people
at the bottom end of the wage spectrum doing worse,
evidenced by more a greater percentage of people making only
the minimum payments on their credit cards, but the high
(34:30):
end of the wage spectrum cranking, especially those that might
have investment accounts and things like that that have gone
up and you know, maybe they're making their they're retired
in their you know, spending is based on, hey my
portfolio is up. Yeah, they are paying off their credit
card balances and full. The remaining twenty eight percent of
people okay, they're paying more than a minimum but less
(34:52):
than full or whatever the percentages. I did the math badly,
but you get what I'm saying.
Speaker 3 (34:57):
There's a lot of stuff that comes out of Dave
Ramsey's that I don't always agree with. The more and
more I've studied the credit card issue, the more I've
come to his way of thinking. His way of thinking,
he's never used any sort of data, you know, pay
off all sorts of debt other than your home. And look,
I have personally never carried a credit card balance, and
so credit cards are great for me. But his point is, look,
(35:19):
that's not the average experience. And even when people say
they're going to pay off the credit card balance, they
usually don't. And the fact of the matter is that
he's right on this. So over half of credit card
holders over the last twelve months had a point in
time where they couldn't pay off the full balance.
Speaker 2 (35:34):
So could the Mike Armstrong corollary be, if you do
not have the money in a bank account to pay
off your credit card balance, then you shouldn't put it
on the credit card. Frankly, even then, I'm not sure
it's adequate.
Speaker 3 (35:47):
Interesting, yeah, because how many times have we heard of
you know, I had the money in the bank, but
then my car broke down and I needed a three
thousand dollars a pair in there goes the money that
I had in my bank account, and now I'm sitting
on the credit card. What about credit balance? I don't
know that there's a thresh. I don't know that there's
a number. Because it comes back to human behavior, and
(36:08):
what we seem to know about human behavior is that
most of them run a credit card balance month after
month and eat away at frankly any of the benefits
that come with the credit card, which is disappointing because
there are a lot of benefits. Yeah, you've got a
bunch of cash back rewards, you've got a bunch of
protection from fraud, but those benefits seem to be outweighed
(36:30):
by the average credit card user experience.
Speaker 2 (36:32):
So can we then amend Benjamin Franklin's famous saying, here
we go to be nothing is certain except debts and taxes. Yeah?
I like that? Is that where you'd like to go?
Speaker 4 (36:42):
I appreciate that the calendars turned and the cold is
here to stay. So now's the perfect time to plan
that getaway to the US Virgin Islands. Act now and
take advantage of a special promotion to Saint Croix, experience
a vibe like no other, and receive a two hundred
and fifty dollars per person airfare credit and a free
hotel night when you book a five night minimum stay
with any participating a hotel before June thirtieth. San Croix
(37:04):
is known for its incredible history and spectacular architecture, along
with world class cuisine, incredible golf courses, and a wide
variety of water sports. From the moment you arrive, you'll
fall naturally in rhythm with the heartbeat of the islands.
There's no money to exchange, and travel from New England
could not be easier. America's Caribbean paradise is waiting for you.
Experience a vibe like no other on the island of
(37:24):
San Croix. Go to visit USVII dot com, use the
code Vibe twenty twenty five for more information and book
your trip today. That's visit USVII dot com and code
Vibe twenty twenty five.
Speaker 2 (37:36):
Airlines are charging higher affairs and are confident that you'll
pay up. That's because there's minimal competition in the space,
and as such, if you want to fly on probably
like eighty percent of routes throughout the United States, you
have one or two options, and generally they come as
close as you can to price fixing without going over
that line.
Speaker 3 (37:51):
And one of the famous low cost providers out there
is in the midst of bankruptcy, so they have a
fair bit of evidence that this model works. Yeah, monopoly
tends to, you know, like go, let's be honest here,
pipoly Chuck all right, triopoly, triopoly, quadopoly, there's there's four
big ones Southwest, American United Delta.
Speaker 2 (38:10):
Right, monopoly. It's studied in economic text textbooks. Let's take
a quick break here, and when we return, we got
our two coming up.
Popular Podcasts
My Favorite Murder with Karen Kilgariff and Georgia Hardstark
My Favorite Murder is a true crime comedy podcast hosted by Karen Kilgariff and Georgia Hardstark. Each week, Karen and Georgia share compelling true crimes and hometown stories from friends and listeners. Since MFM launched in January of 2016, Karen and Georgia have shared their lifelong interest in true crime and have covered stories of infamous serial killers like the Night Stalker, mysterious cold cases, captivating cults, incredible survivor stories and important events from history like the Tulsa race massacre of 1921. My Favorite Murder is part of the Exactly Right podcast network that provides a platform for bold, creative voices to bring to life provocative, entertaining and relatable stories for audiences everywhere. The Exactly Right roster of podcasts covers a variety of topics including historic true crime, comedic interviews and news, science, pop culture and more. Podcasts on the network include Buried Bones with Kate Winkler Dawson and Paul Holes, That's Messed Up: An SVU Podcast, This Podcast Will Kill You, Bananas and more.
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com