Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:01):
The value of investments and the incomefrom them may go down as well as up,
and investors may not get backthe amounts originally invested.
Past performance is not aguide to future performance.
The information is not an offer,solicitation or recommendation of any
funds, services or products,or to adopt any investment strategy.
This pod is marketing material issued bySchroder Investment Management Limited,
(00:25):
registered number 189-3222 England.
Authorized and regulated by the FinancialConduct Authority for
informational purposes only.
Please contact your financial adviserbefore making any investment decisions.
Welcome to the Investor Download, the
(00:49):
podcast about the themes driving markets
and the economy now and in the future.
I'm your host, David Brett.
For investors in the theme of disruption,the start of 2025 has
been one for the ages.
(01:09):
There are growing fears of a major globaltrade war tonight after President Trump
announced sweeping tariffs on imports fromCanada, Mexico, and
China starting tomorrow.
He added that very substantial tariffswould follow for the European Union.
(01:30):
Yet despite the column inches taken up bythe new administration in the US, topping
the list of disruptive events was theemergence of a formerly little
known Chinese AI startup, Deep Seek.
Technology shares on Wall Street havefallen sharply in response to the
emergence of a low-cost chatbot built bya Chinese artificial intelligence firm.
The chatbot launched by the Chinese firmDeep Seek has become the most downloaded
(01:55):
free app in the United States sinceit was launched earlier this month.
This has been in the makings for a while.
While it hit the public conscience, almostwith surprise, in the world of AI
practitioners, the people who are makingthese models and testing and running them,
(02:16):
our developer world.
Deep Seek was a known entity.
That's Ankur Dubey,an investment director in private markets.
While Deep Seek's app wasn't a surprise tothose in the know, the
advanced nature of the technology was.
They essentially invented a new way ofmaking the large language models
(02:36):
that underpin things like ChatGPT.
It's the way they did that that is sointeresting because they did that in a
very efficient and cost-efficient wayversus the incumbent methodologies.
That's Paddy Flood, an investorin the theme of disruption.
This led to quite a few questions aboutwhether if you can now build something
like ChatGPT in a far more efficient way,what does it mean for spending on AI
(03:01):
infrastructure, which has been a reallybig driver of the markets over the last
couple of years, as we've seenwith areas like semiconductors.
Competition for Silicon Valley, costefficiencies, and the potential for a new
world order in techand AI, roiled markets.
We've got a bit of a tech sell-off thismorning, and it's being caused by
(03:23):
Earth-shatteringdevelopments in the AI space.
Let's take a look at this.
You see the Dow down morethan 140 points right now.
The S&P is solidly lower, but the realaction is over here in the Nasdaq, 600
points lower, nearly 3% on track for oneof its worst days in the past two years.
Here's why.
There's a Chinese startup that few peoplehad ever heard of until the past few days,
(03:46):
and it has emerged as a realplayer in the AI arms race.
It's called Deep Seek.
A month on, and with the dust beginning tosettle, Paddy and Ankur tell us from a
public and private market's perspective,what's changed and what can investors
learn from the experienceabout investing in AI?
On Apple Podcasts,Spotify, or wherever you get your
(04:09):
podcasts, you're listeningto the Investor Download.
When DeepSeek launched in January,it took the world by storm.
Shortly after, it took over rival OpenAI'scoveted spot as the most downloaded
free app in the US on Apple's App Store.
Rave reviews for its performance and thefact that it was open source so users
could see how the app workedproved too hard for consumers to ignore.
(04:32):
I think it was quite evident early onthat they'd done some quite interesting
innovationsaround how you actually go about
creating these large language models.
Because it's open source, thatis apparent quite quickly.
I think it's clear that generally in theindustry, costs to deliver AI have
(04:53):
been coming down through time.
Deep Seek's innovations, which everyonecan now use or should be able to use
will accelerate that cost-down curve.
It's likely to mean that the cost ofproviding these services, people
using Open AI, for example,will continue to fall.
Deep Seek claims its engineers neededaround $6 million in raw computing
power to build their app.
(05:15):
Roughly one-tenth of what Meta spentin building its latest AI technology.
I think it leads down to Jevons paradox,which is a popular term used in the space.
Any time as there are efficiency gains forany resource, the
natural instinct is to think that thatresource will not be as
(05:36):
much in demand anymore.
But what ends up happening is that becauseof the marginal cost of
using that resource decreases, more andmore people start to use that resource.
If Deep Seek's claims are true, it couldopen the door to smaller competitors to
participate in a market that's to datebeen dominated by the Mag 7 in the US.
(05:58):
In this case, compute or computing powerall the way from using from the Nvidia or
any cloud computing capabilities.
They have been a big hurdle for smallcompanies, small founders
to build AI applicationsbecause they're super expensive.
(06:21):
Now, as efficienciesimprove, as demonstrated by Deep Seek,
and much of that is out of necessitybecause of the restrictions
imposed on China from US.
As those efficiency gains happen, more andmore AI application
(06:42):
producers or founders and companies thatwill start building on AI will
start growing.
Because of that, the overall volume ofdemand for those compute resources and
infrastructure resourcesare expected to go up.
So think of itin the era of mobile, when the data prices
(07:05):
went down, bandwidth increases,there was almost an explosion in number
of mobile applications out there.
We have experienced a similarphenomenon in the Internet era.
As bandwidth With increase from dial-up tohigh speed to fiber and beyond, the number
of internet websites andinternet use cases exploded.
In AI, it would likely lead to a continuedpush and race for all these companies to
(07:28):
make better and cheaperlarge language models.
In this case, the number ofapplications are likely to go up.
It will become moredemocratised in that sense.
Then because of which, while the price perunit of compute will go down, the expected
overall volume is likely to go up.
(07:49):
Therefore, for the AI infrastructurespace, overall, the addressable
market is expected to increase.
But as cost fall, it poses a question forinvestors in AI about the risks
associated with those investments.
That's coming up in part two of the show.
Get in touch with us by email atschroderspodcasts@schroders.
(08:12):
com or visit our website, schroders.com/theinvestordownload.
Today, most investors, certainly retailinvestors, will have looked to the
so-called Magnificent Seven in the USto gain exposure to the theme of AI.
Magnificent Seven have been a publicmarket proxy for exposure to AI because
those are the ones who are doing all sortsof innovations, right from
(08:34):
compute to applications.
So big had the Mag7 become, that by theend of 2024, that accounted for more than
a third of the S&P 500 and nearlya quarter of the MSCI world.
But that would ignore the many othercompanies that operate in the AI space,
in computing, foundation,infrastructure, or applications.
(08:54):
And most of that investment is happeningaway from the public space
and in private markets.
But in the private market space,there's a lot more opportunities to invest
because a lot of innovation is happeningat every layer, right from compute.
There are competitors to NVIDIA,competitors to
(09:15):
Azure and Microsoft to all the way downright up there where OpenAI and Anthropic
sit in the large language model space anda bunch of applications
building on top of that.
What this does specifically, and bythis, I mean the deep seek innovation,
is it brings into question thehigh capital intensive nature of
(09:40):
the large language model itself.
Therehave been questions raised over the last
year and a half or two that this model ofCapEx-intensive development of large
language models, where is it going to?
Where AI is going remains a big questionthat needs answering, especially given
(10:02):
the amount of money that's being invested.
The longer that question goes unanswered,the more the risk becomes
attached to those investments.
I think it's justreally important to contextualize that AI
still has risksto investing in that theme.
But I don't think deep seekhas changed those risks.
So deep seek has made it more efficient,and I think that will drive adoption.
(10:24):
But what is the big unknownwith AI is adoption.
It's whether this technology will be usedby businesses and people all over the
world to do things that help them improvetheir productivity or whatever it may be.
That was a question before Deep Seek, andthat is still the question after
Deep Seek, and one that I think willdetermine the success or
otherwise investing in AI.
Because the end use case is still inquestion, adoption is still in question of
(10:47):
AI, how long can you keep spending onCapEx for large language models to figure
out whether this technologyhas an adoption case or not?
With this, there isat least, I mean, it's too early to say in
private markets because typically, privatemarkets lag public by six to eight
(11:10):
months in terms of showing traction.
But the discussions and ramblings havebegun on being more efficient with the
dollar in terms of how you want to developthis technology, how do you want to
develop the models, andwhat it it's pushing into is the cost of
(11:34):
experimentation withthe cost of experimentation is going down,
but you still want to be more careful withwhere you want to deploy those dollars.
So how long can investors keep plowing intheir money before they demand a payoff?
Well, I thinkfrom an investor perspective,
you need to see a payoffin terms of people using this technology
(12:00):
relatively soon.
There has been a lot of investment goinginto the AI infrastructure, and
the companieswho are trying to monetise it haven't
yet done so in a significant way.
I think we just need to see over the next12 to 18 months, more of those use cases
coming out that will justify the spend,because otherwise, there will be a
(12:22):
mismatch as we get big investmentsand not much return in terms of use yet.
That always creates uncertainty forinvestors.
Given the risks, how shouldinvestors approach investing in AI?
That's coming up in thefinal part of the show.
Deep Seek has shaken up the market interms of the costs associated with
(12:45):
building large language modelsand associated AI tools.
It's also brought into focus the need forconsumers and investors to see a payoff in
terms of how this technology will bedeployed and how it will be used in a
way to benefit productivity and society.
And with that, long-termvalue in their investments.
With that in mind, how shouldinvestors approach investing in AI?
(13:07):
Because the market in general still hasn'tseen a host
of products that have been adopted enmasse by enterprise use cases
or by consumer use cases.
The faith inthe return that will be delivered by each
dollar spent on AIdevelopment is very low.
(13:31):
So anyshock news that the market sees, the
reaction is going to be extreme,which was the case here.
However, as investors, it'skey to have a long term bigger
picture and be thematically correct.
(13:53):
So these short term volatilitiesare not of meaningful impact.
Impact.
But that being said, thisdisruption is always meaningful
from an innovation perspective.
(14:13):
Another thing I'll also add there isto Patty's point that innovation
is happening all over the world.
It's not centered into one regionspecific, which people believe that a lot
of innovation is happeningin one or two centers.
Butanother layer it's added to the people's
(14:37):
imagination is scarcity alsodrives a lot of innovation.
In this case, much of the innovation deepsea has done is born out of a lack
of availability of the resources.
Purely because of the restrictions imposedby US on China in terms of availability of
(14:57):
chips, they were able to extractmuch more out of those chips.
The questions are, there's also a debatehappening whether
you need to spend that money todrive that innovation, or if sometimes
restricting that capital is also good tomaybe improve efficiencies.
(15:20):
So don't fear disruption.It can be a good thing.
But do be aware of the short termmarket ructions disruptors can cause.
And the AI theme is not one that isdominated by just the US and China.
This is a global race.
As Deep Seek has shown, even the smallerinnovators can produce game-changing
products despite thehurdles put in their way.
(15:41):
I think it's always good totake a step back and think of the
businesses that you think have very highdefendable barriers to entry in
this space that are really driving valueto it and have long growth trajectories
that aren't priced in by markets.
It's a careful balance of assessing thefundamentals of the business
(16:01):
versus the valuation to make sure thatyou're not paying exuberant prices for a
company's destiny that might not be withinits control or might be a destiny that
is going to be increasingly competitive.
That's how we tend to think about it.
If you enjoyed the pod, you can find myfull unabridged chat with Ankur and Paddy
in its full technicolor gloryon Schroders' YouTube page.
(16:24):
There will also be an article to accompanythis pod on Schroders'
Insights pages at schroders.
com.
That was the show.
We very much hope you enjoyed it.
You can subscribe to the investordownload wherever you get your podcast.
(16:45):
And if you want to get in touch withus, it's schroderspodcasts@schroders.
com.
And you can find out much,much more at schroders.
com/insights.
New shows drop every other Thursdayat 05: 00 PM UK in the meantime.
In the meantime, keep safe and go well.
The value of investments and the incomefrom them may go down as well as up, and
(17:08):
investors may not get back theamounts originally invested.
Past performance is not aguide to future performance.
The information is not an offer,solicitation or recommendation of any
funds, services or products, orto adopt any investment strategy.
This pod is marketing material issued bySchroder Investment Management Limited.
Registered number, 18932220, England.
(17:32):
Authorized and regulated by the FinancialConduct Authority for
informational purposes only.
Please contact your financial adviserbefore making any investment decisions.