FidelityConnects: Future of advice – AI possibilities with Mark Schmehl and Annie Wang
Join Mark Schmehl, Portfolio Manager, and Annie Wang, Director of Private Investments, for a discussion on AI's past, present and future, as well as investment possibilities.
Transcript
00:06.439 --> 00:09.476
Without further ado I'm delighted to be joined in studio by
00:09.476 --> 00:13.780
portfolio manager, Mark Schmehl and Fidelity Director of Private Equity,
00:13.780 --> 00:17.350
Annie Wang. Mark, Annie, wonderful to see you both here.
00:17.350 --> 00:20.587
Now, Mark you're here from San Francisco via Montreal.
00:20.587 --> 00:22.856
You were at a client event.
00:22.856 --> 00:26.059
Annie, you are ... and actually, I just wanted to mention before we get to you,
00:26.059 --> 00:29.729
Annie, you and I had a great discussion in Palm Beach recently to thousands of
00:29.729 --> 00:32.332
advisors through that forum on your funds.
00:32.332 --> 00:35.835
We're not going to talk about that today. We're gonna talk about thematic AI.
00:35.835 --> 00:37.670
I'm really looking forward to that.
00:37.670 --> 00:41.775
Now, everybody knows who you are and Annie, you're going to be famous after
00:41.775 --> 00:45.912
this video. Could you please tell everybody what your role is and what your
00:45.912 --> 00:47.781
responsibility is at Fidelity?
00:47.781 --> 00:52.485
Absolutely. I'm based in the San Francisco office along with Mark.
00:52.485 --> 00:56.523
I'm on the private equity team and, really, my focus is AI
00:56.523 --> 00:59.926
startups.
00:59.926 --> 01:03.363
Mark, I'd like to start with you then from a key question.
01:03.363 --> 01:08.301
Let's clear some stuff up. There's a lot of talk today about an AI bubble.
01:08.301 --> 01:12.439
Now, Bill Gates would say that we are in a bubble, so would Michael Burry, and
01:12.439 --> 01:14.507
then Jay Powell would say, we're not.
01:14.507 --> 01:17.210
Can we start there before we get into the whole theme of AI?
01:17.210 --> 01:21.448
Sure. It's very topical, the market is wrestling with that as we speak.
01:21.448 --> 01:25.985
I think that the easiest way for me to address that is there
01:25.985 --> 01:30.857
are two camps. There is a very bullish
01:30.857 --> 01:35.095
engineering camp who sees the application and they're
01:35.095 --> 01:38.898
so excited to be working on it and they are saying look at all the things I can
01:38.898 --> 01:41.101
do with AI that I could never do before.
01:41.101 --> 01:42.969
Oh my goodness, I'm so excited.
01:42.969 --> 01:47.107
Then there's the finance camp which is
01:47.107 --> 01:50.844
very skeptical and saying this has gone too far too fast and I don't think you
01:50.844 --> 01:54.681
engineers know what you're talking about, and it's clearly a bubble because my
01:54.681 --> 01:56.983
experience tells me it's a bubble.
01:56.983 --> 02:01.354
I think that what you have to answer for yourself as an investor is who
02:01.354 --> 02:05.425
do I want to believe, the finance geeks in Wall Street or the
02:05.425 --> 02:08.461
engineering geeks in Silicon Valley?
02:08.461 --> 02:12.499
For me, the engineers are seeing this more clearly than the finance guys are.
02:12.499 --> 02:15.468
They see the applications, they're in it, they're doing it.
02:15.468 --> 02:17.537
I'm going to go with engineers on this one.
02:17.537 --> 02:21.541
My view is it does seem a little overheated but I do not
02:21.541 --> 02:24.144
believe it's a bubble as we speak now.
02:24.144 --> 02:28.181
If we think of the chart room, which you're a user of, there's always
02:28.181 --> 02:30.116
little retracements within the market.
02:30.116 --> 02:34.053
Overall, the theme is there, it's powerful, maybe some things are overheated,
02:34.053 --> 02:36.589
whatever, but that's what portfolio management's all about.
02:36.589 --> 02:38.791
Nothing has changed. Fundamentals haven't changed.
02:38.791 --> 02:41.694
If anything, they're getting better so I don't know.
02:41.694 --> 02:44.831
Annie, this is what the analyst side is all about as well.
02:44.831 --> 02:48.568
As director of private equities talk to us about the importance of the private
02:48.568 --> 02:51.204
space within this realm.
02:51.204 --> 02:55.175
Absolutely. This is where a lot of innovation is happening.
02:55.175 --> 03:00.180
We're talking to the engineers and understanding where the big ideas are.
03:00.180 --> 03:04.317
It's really my job and my team's job to be on the ground, meeting founders,
03:04.317 --> 03:08.288
talking to people that are building the space and almost trying
03:08.288 --> 03:12.292
to serve as a bridge between that community with the portfolio
03:12.292 --> 03:16.629
managers and people who spend most of the time in the public markets to
03:16.629 --> 03:19.199
understand what could happen in the future.
03:19.199 --> 03:22.936
You're based in San Francisco, what size team are you part of?
03:22.936 --> 03:24.737
Our team is actually quite small.
03:24.737 --> 03:28.942
The private equity team is about 10 people but we're very
03:28.942 --> 03:32.979
fortunate in that we're able to work with portfolio managers like Mark and lots
03:32.979 --> 03:37.617
of peers on the public side who are constantly thinking about AI as well to
03:37.617 --> 03:38.451
collaborate together.
03:38.451 --> 03:41.688
And then of course you have the whole universe of Fidelity to be working with
03:41.688 --> 03:45.725
and drawing on and contributing to as well, not just in San Francisco with your
03:45.725 --> 03:49.128
key team. Mark, how do you work with the private equity team?
03:49.128 --> 03:51.798
I talk to them every day. Annie sits right across from me.
03:51.798 --> 03:54.667
The rest of her team is dispersed around the world.
03:54.667 --> 03:57.704
We have Vishal in San Francisco as well.
03:57.704 --> 04:01.641
I think that they have just the greatest window on the future
04:01.641 --> 04:04.677
and they see what's happening before everyone else does.
04:04.677 --> 04:08.081
The themes, the smartest people are in the private world.
04:08.081 --> 04:11.217
They're working on things we've never even heard of before.
04:11.217 --> 04:15.488
It's just a lens on what's coming at us.
04:15.488 --> 04:19.125
In the public markets you're always dealing with the future, in the private
04:19.125 --> 04:22.128
markets you're living in the future, so to speak.
04:22.128 --> 04:24.931
From that standpoint it's a great information source.
04:24.931 --> 04:28.868
It's really great to talk to these innovative founders.
04:28.868 --> 04:31.237
Annie will come to me every day and go, have you heard of this company?
04:31.237 --> 04:33.473
She does this, what was the name of the company? I can't remember.
04:33.473 --> 04:35.441
Have you ever heard of his company? No.
04:35.441 --> 04:37.810
Do you know anyone who uses it? No.
04:37.810 --> 04:40.346
This is the kind of back and forth we have all the time.
04:40.346 --> 04:43.950
Then we try and figure out does anyone like this company, do they use it?
04:43.950 --> 04:47.987
It is incredibly valuable and being based in San Francisco is a
04:47.987 --> 04:50.690
huge, huge edge for all things private.
04:50.690 --> 04:52.625
Annie, you're discovering things every day.
04:52.625 --> 04:53.760
Yeah, absolutely.
04:53.760 --> 04:58.398
I just want to continue on the background for AI before we get into use cases.
04:58.398 --> 05:02.602
You told me in 2022 somebody on your team saw a
05:02.602 --> 05:05.905
posting on a billboard or something on a lamp post that said there was an AI
05:05.905 --> 05:08.408
conference and that got your interest.
05:08.408 --> 05:09.509
That was you, Annie.
05:09.509 --> 05:15.448
Yeah, it was this small warehouse in the middle of San Francisco,
05:15.448 --> 05:19.452
Cerebral Valley, a neighbourhood had been renamed to Cerebral Valley,
05:19.452 --> 05:23.189
and it was incredible. We had no idea, it was just something we heard about.
05:23.189 --> 05:26.492
You saw into the future.
05:26.492 --> 05:28.795
Well, yeah.
05:28.795 --> 05:31.564
I mean, it was pretty remarkable, I think.
05:31.564 --> 05:35.635
When I joined Fidelity three years ago ChatGPT was not
05:35.635 --> 05:38.604
a thing. It hadn't even been created then.
05:38.604 --> 05:41.607
No one had heard of OpenAI.
05:41.607 --> 05:43.910
Being in San Francisco is so valuable.
05:43.910 --> 05:47.780
I had actually gone to business school in the Bay Area and the first time I
05:47.780 --> 05:52.151
learned about OpenAI was actually a few months before I joined Fidelity,
05:52.151 --> 05:55.088
and it was my friend's Instagram.
05:55.088 --> 06:00.059
She's a founder asking for early invite coach, GPT-3,
06:00.059 --> 06:03.663
which is pretty remarkable. I think it's just so much of the benefits of being
06:03.663 --> 06:07.700
in San Francisco and just hearing about what founders and
06:07.700 --> 06:09.736
engineers are excited about.
06:09.736 --> 06:11.738
It's been a pretty cool journey.
06:11.738 --> 06:15.575
You have to be open to every medium, if you will, as far as information that's
06:15.575 --> 06:18.911
coming at you and thinking how could this be additive to a portfolio manager,
06:18.911 --> 06:21.414
how could it be additive for our advisors and their clients?
06:21.414 --> 06:23.916
It's pretty fascinating.
06:23.916 --> 06:27.487
Everybody knows the Mag Seven but we're really talking about the incubation of
06:27.487 --> 06:31.557
the next wave. How do you weed out those companies that
06:31.557 --> 06:35.061
aren't going to survive, or that you estimate aren't going to survive?
06:35.061 --> 06:38.064
That's a fantastic question and it's something that we think a lot about.
06:38.064 --> 06:42.201
I think about
06:42.201 --> 06:45.838
where I've spent time and the team has spent time in terms of these AI
06:45.838 --> 06:50.009
companies. There's so many of them. I think if I were to go back
06:50.009 --> 06:53.980
to 2022, that's when the ChatGPT moment happened, and I
06:53.980 --> 06:56.649
think we all kind of said to ourselves, huh, this is something that we should
06:56.649 --> 07:00.653
pay attention to. It was really hard to figure out who the winners and losers
07:00.653 --> 07:06.993
were at that point but I think we knew that this AI thing was going to be big.
07:06.993 --> 07:10.062
I really spent a lot of time at the infrastructure level looking at the picks
07:10.062 --> 07:12.231
and shovels.
07:12.231 --> 07:16.235
We've invested in names like CoreWeave which is a next gen AI cloud or
07:16.235 --> 07:20.506
VAST Data which is focused on compute — sorry, storage
07:20.506 --> 07:23.009
for next gen AI workloads.
07:23.009 --> 07:28.181
If we were to fast forward today there's just a lot of focus on the
07:28.181 --> 07:31.918
applications and the use cases of AI.
07:31.918 --> 07:35.755
I really think of that as kind of our journey of kind of moving up the stack as
07:35.755 --> 07:37.590
we learn more and about AI.
07:37.590 --> 07:40.593
You go as broad also to look at inputs to a data centre.
07:40.593 --> 07:44.730
For example, I hear that these things are very droney and they put them in
07:44.730 --> 07:48.734
backyards of peoples' neighbourhoods and you need insulation for the
07:48.734 --> 07:53.372
cooling stacks. There's lots of companies that can be inputs to
07:53.372 --> 07:57.743
a structure like that, that provides a lot of opportunity for you.
07:57.743 --> 08:01.280
Absolutely. In terms of the AI supply chain that's definitely something that
08:01.280 --> 08:05.218
we're paying attention to. We also have someone who really specializes in
08:05.218 --> 08:09.222
the semi space, Adam Benjamin, who spends a lot of time there too.
08:09.222 --> 08:13.159
Now, what are use cases of AI in your view, from what you do
08:13.159 --> 08:15.661
but also what our clients can take away?
08:15.661 --> 08:18.598
Absolutely. think there's a few that really come up.
08:18.598 --> 08:22.768
On the consumer side
08:22.768 --> 08:26.506
ChatGPT as a personal assistant and people just use it for so many different
08:26.506 --> 08:31.310
things from finding Christmas gifts
08:31.310 --> 08:34.380
to improving my writing. I think on the consumer's side you've really seen a
08:34.380 --> 08:38.351
lot of traction there. On the enterprise side coding is something
08:38.351 --> 08:42.388
that has emerged as a killer use case this year.
08:42.388 --> 08:46.125
I also see a lot of traction with customer success.
08:46.125 --> 08:50.263
There's just a lot of startups that are building the
08:50.263 --> 08:55.067
space to build more specialized agents for certain domains.
08:55.067 --> 08:58.304
I think it's still early but definitely a lot of smart people working in the
08:58.304 --> 08:58.971
space.
08:58.971 --> 09:02.341
Early is a key point because there's a lot still to be learned, and the way
09:02.341 --> 09:05.344
that it's learned is by trial and error.
09:05.344 --> 09:09.315
When EVs came out, autonomous driving, there was a lot of trial and error.
09:09.315 --> 09:12.585
We won't get into that. But that's true with AI in general, isn't it?
09:12.585 --> 09:16.656
There's a lot that needs to be out there in the world for the
09:16.656 --> 09:20.526
companies to learn more from so a lot still to go.
09:20.526 --> 09:20.826
Yes, definitely.
09:20.826 --> 09:24.830
It's in its infancy. Why don't we go to the journey of a
09:24.830 --> 09:26.732
private company as well?
09:26.732 --> 09:29.835
As far as the development of a private company to then public, is that
09:29.835 --> 09:34.206
something that you want to talk to the customers about?
09:34.206 --> 09:38.311
Absolutely. The way we think about it is we want to meet
09:38.311 --> 09:42.481
the big, important public companies of tomorrow as soon as we can
09:42.481 --> 09:46.886
and invest as soon as we gain conviction that
09:46.886 --> 09:48.354
they'll be an important company.
09:48.354 --> 09:53.793
Thank you. Mark, how do we co-exist with super intelligence?
09:53.793 --> 09:54.727
Good question.
09:54.727 --> 09:56.896
Just a brief answer is fine.
09:56.896 --> 10:00.032
I don't know that we're going to see super intelligence.
10:00.032 --> 10:03.002
I'm just going to be happy with a better tool, generally speaking.
10:03.002 --> 10:05.037
I think that that's how I think about AI.
10:05.037 --> 10:08.140
It's a better tool for a lot of different use cases.
10:08.140 --> 10:10.776
As Annie was saying, we don't really know where they are.
10:10.776 --> 10:14.914
They're planting seeds everywhere. They're taking their super intelligence and
10:14.914 --> 10:16.649
throwing it at all these different problems.
10:16.649 --> 10:18.517
There's other ones that I
10:18.517 --> 10:20.920
can't even remember all the names anymore that are attacking coding.
10:20.920 --> 10:24.790
We just don't know where it's going to really work.
10:24.790 --> 10:28.928
I think the one way to think about AI is this is
10:28.928 --> 10:31.097
the worst it's ever going to be.
10:31.097 --> 10:33.032
It's just going to keep getting better.
10:33.032 --> 10:36.902
We know from having watched it over the last three or four years that the
10:36.902 --> 10:41.007
capabilities of ChatGPT or Anthropic gets amazingly
10:41.007 --> 10:42.708
different over two years.
10:42.708 --> 10:46.345
As we throw more compute at it and the algorithms get better it just feels as
10:46.345 --> 10:48.147
though this is going to continue to get better.
10:48.147 --> 10:51.851
I don't know that we get the super intelligence but what about super tool?
10:51.851 --> 10:55.821
It's going to be a super tool and it's going to allow us to do a lot of things.
10:55.821 --> 10:59.625
That's less ominous if it's thought of as a super tool versus super
10:59.625 --> 11:02.361
intelligence because we want to work alongside.
11:02.361 --> 11:05.064
Do you see the day where you have a robot walking around in your house washing
11:05.064 --> 11:07.767
your dishes?
11:07.767 --> 11:11.604
Yes. Yes. We are currently looking at robotic companies in the Bay Area.
11:11.604 --> 11:13.539
It's funny, they're all working on folding.
11:13.539 --> 11:17.076
This is really funny, there's ten different companies, they're all working on
11:17.076 --> 11:18.778
folding clothes with the arms of the robot.
11:18.778 --> 11:22.815
I think that robotics is not a hardware
11:22.815 --> 11:24.650
problem, I think it's a software problem.
11:24.650 --> 11:28.888
The brain of the robot has to be good enough so we need the
11:28.888 --> 11:33.659
intelligence to get there. You can sort of see the runway to getting there.
11:33.659 --> 11:37.630
I don't think it's outrageous to think that we may have robots in our
11:37.630 --> 11:42.468
homes in the next 10 years or 15, depending on supply chain and whatnot.
11:42.468 --> 11:45.171
You can see what they're doing and you can sort of see the path.
11:45.171 --> 11:49.008
It's sort of like it took, I don't know how long it took Waymo to get going, 15
11:49.008 --> 11:53.245
years from beginning to where it is now, maybe 15
11:53.245 --> 11:56.816
years for robotics, maybe not, I don't know but you can see the path there.
11:56.816 --> 12:00.252
You're a big fan of Waymo, I'm sure you are as well, and I am as well when I go
12:00.252 --> 12:03.856
down to Scottsdale. I'm sure all of our clients are wondering what is that, or
12:03.856 --> 12:05.558
perhaps they've tried.
12:05.558 --> 12:09.128
Annie, when you're meeting with companies do you hear a lot about jobs as far
12:09.128 --> 12:13.432
as people saying is this going to be a cost for my job, am I going to lose
12:13.432 --> 12:17.470
a job? Is it taking away jobs or is additive, or just replacing
12:17.470 --> 12:19.705
the way people do things?
12:19.705 --> 12:23.776
I think that's something that definitely comes up. These questions
12:23.776 --> 12:27.646
are being asked of the companies that are building in this space as well.
12:27.646 --> 12:31.717
I really think that people are very thoughtful in terms of
12:31.717 --> 12:35.654
how to build these tools and
12:35.654 --> 12:39.692
virtual collaborators. Just based on
12:39.692 --> 12:43.629
my own experience I would not trust AI to completely do something
12:43.629 --> 12:48.734
autonomously but it does help me be more succinct
12:48.734 --> 12:52.838
or help me with better decision making or do grunt work that I don't really
12:52.838 --> 12:56.475
want to do myself and frees up my time to do more interesting work.
12:56.475 --> 13:00.713
I think this human-in-the-loop model really resonates
13:00.713 --> 13:04.683
both with people that are using these tools and also the companies that are
13:04.683 --> 13:05.684
building the space.
13:05.684 --> 13:09.955
Excellent. Mark, do you see jobs as something as
13:09.955 --> 13:13.926
an evolution, that it's really about you're not going to be out of
13:13.926 --> 13:16.862
work, but it's really a shift in what it is that you do?
13:16.862 --> 13:20.699
The best example I can give on this is my son is a software engineer and
13:20.699 --> 13:25.237
they're the pointy end of the spear in terms of being replaced by AI.
13:25.237 --> 13:28.841
When we talk he just gets more productive.
13:28.841 --> 13:31.076
Who doesn't want a more productive employee?
13:31.076 --> 13:35.147
We are all in the business of competing with our competitors so if my employees
13:35.147 --> 13:39.151
get 10x more competitive and you don't I'm going to beat you
13:39.151 --> 13:43.122
at the business game. For me, I do feel as though we're
13:43.122 --> 13:45.825
going through a transition where companies are trying to figure out their human
13:45.825 --> 13:50.629
resource needs but I think that the smart companies are saying AI is
13:50.629 --> 13:54.800
a tool for productivity improvement, not a tool cost
13:54.800 --> 13:58.871
reduction, therefore, I'm going to invest as much
13:58.871 --> 14:03.209
as possible in people who know how to use AI to make my productivity go up.
14:03.209 --> 14:05.377
I think not everybody's there yet.
14:05.377 --> 14:07.913
There's still a lot of companies that are in the I'm going to use AI and fire
14:07.913 --> 14:11.116
all these people. I think the smart companies are like, I'm going to go hire a
14:11.116 --> 14:14.486
bunch of these people and make them way more productive and then I'm going to
14:14.486 --> 14:16.488
out-compete my competition.
14:16.488 --> 14:20.092
We're in this moment and I think you have two different strategies and I do
14:20.092 --> 14:24.363
believe that the productivity is the key part of the strategy.
14:24.363 --> 14:28.801
Ultimately, the most value in the market is ascribed to productivity,
14:28.801 --> 14:32.872
not to cost savings. If we paid for cost savings we'd all own like a
14:32.872 --> 14:38.544
bank. But no, we pay for productivity so we would all rather own Google.
14:38.544 --> 14:42.248
That's what's more valuable to society, is productivity enhancement not cost
14:42.248 --> 14:44.216
reduction. I think that's where we're going.
14:44.216 --> 14:45.684
Your son is set up well now.
14:45.684 --> 14:46.118
Yeah, yeah.
14:46.118 --> 14:50.222
That's good, good to know. Productivity enhancement, from tools that
14:50.222 --> 14:54.627
you come across where our advisors may be inspired,
14:54.627 --> 14:58.097
really there's tools that help them become more efficient in what they do, they
14:58.097 --> 15:00.332
can take on more business, more clients.
15:00.332 --> 15:03.135
Can you talk about any of those types of tools that come across your desk?
15:03.135 --> 15:06.405
I don't use any of them. I don't know, Annie, have you seen any of the finance
15:06.405 --> 15:07.907
tools that are out there?
15:07.907 --> 15:10.709
We use our internal tools, they've been quite helpful.
15:10.709 --> 15:13.946
But thematically it's really about efficiency and productivity within the
15:13.946 --> 15:14.513
business.
15:14.513 --> 15:18.918
Imagine if you're getting your emails and you have an AI bot
15:18.918 --> 15:22.955
scanning through your emails and saying client XYZ just sent me this
15:22.955 --> 15:26.725
request for whatever it was, and it's some random paper trail I have to spend
15:26.725 --> 15:28.928
five days chasing out.
15:28.928 --> 15:32.932
The AI should be able to take that request, know what it is, find the paper, do
15:32.932 --> 15:36.335
the work and send it back, and you don't have to do it.
15:36.335 --> 15:39.204
Theoretically, that is what you're going to see in AI.
15:39.204 --> 15:43.142
A lot of this paper chasing, many people have paper chasing aspects of their
15:43.142 --> 15:47.079
job. Those should be solvable from a workflow standpoint by
15:47.079 --> 15:51.016
AI. I think we're going see that and I think we're going to see that faster
15:51.016 --> 15:55.020
than we expect. I also think we are not going to
15:55.020 --> 15:58.424
necessarily buy a software platform to do that.
15:58.424 --> 16:02.461
It's going to be the AI will just do it on its own and you won't need to go get
16:02.461 --> 16:06.598
a separate AI. You can just do it through Chat GPT or Claude or whatever.
16:06.598 --> 16:09.802
I see.
16:09.802 --> 16:14.039
Annie, New York Times recently had an article on AI versus
16:14.039 --> 16:17.643
AI, and what it had to do with was what's real anymore?
16:17.643 --> 16:21.046
Do you have any thoughts, and I'd love to ask you this as well, Mark, just on
16:21.046 --> 16:24.817
how should advisors work with their clients on what to believe, what not to
16:24.817 --> 16:28.687
believe. For instance, if I saw you on TV during the World Series with a Blue
16:28.687 --> 16:32.124
Jays jersey, I'd know it's you, but if you had a Dodgers jersey on I'd think
16:32.124 --> 16:34.126
maybe that was constructed by AI.
16:34.126 --> 16:38.097
Annie, do you have thoughts from
16:38.097 --> 16:42.167
a cybercrime standpoint, what's believable and what's not, and maybe messages
16:42.167 --> 16:44.603
that advisors could pass to their clients.
16:44.603 --> 16:48.607
I think it's such a fantastic question. I really think you need everyone around
16:48.607 --> 16:51.210
the ecosystem to be working on this.
16:51.210 --> 16:55.481
I think the model companies are actually doing a lot of work on
16:55.481 --> 16:59.685
this front. For example, OpenAI restructured last week
16:59.685 --> 17:04.123
and their nonprofit made a $25 billion commitment to two areas.
17:04.123 --> 17:08.060
One is for health and curing diseases and the second area is for
17:08.060 --> 17:13.098
AI resilience, which is kind of like this next gen ecosystem
17:13.098 --> 17:18.070
to think about some of the issues that we will face.
17:18.070 --> 17:21.774
I think it's evolving but there's definitely a lot of thought intentionality on
17:21.774 --> 17:23.075
this.
17:23.075 --> 17:25.377
So there is policing, do you sense?
17:25.377 --> 17:26.912
Mark, what are your thoughts on that?
17:26.912 --> 17:31.250
I think it's a fast-moving area that nobody really has good answers to.
17:31.250 --> 17:36.021
I think that different companies have different approaches, and
17:36.021 --> 17:39.291
it's contentious sometimes. Anthropic has one approach, OpenAI has one
17:39.291 --> 17:42.461
approach, Google has one approach. If you talk to all three of them they all
17:42.461 --> 17:44.596
think they have the right approach to how to do it.
17:44.596 --> 17:48.367
I think that it's one of those areas that is important and needs to be talked
17:48.367 --> 17:52.137
about, and maybe it's not being addressed holistically by everybody at the
17:52.137 --> 17:56.375
moment but I do think it's an issue that everybody is aware of and
17:56.375 --> 17:59.912
I just don't think we have a coherent solution just yet That's not saying we
17:59.912 --> 18:03.115
won't. I mean, we eventually figured out how to do rules for driving.
18:03.115 --> 18:07.786
We've eventually figured it out how to do a lot of different complex things.
18:07.786 --> 18:12.124
We've built laws and jurisprudence around all of this stuff so we
18:12.124 --> 18:13.926
will get there but it's going to take some time.
18:13.926 --> 18:17.930
It took a long time — remember when Google came out and we had the
18:17.930 --> 18:21.900
keywords and nobody understood what keywords were, we're like what's a keyword?
18:21.900 --> 18:24.903
I took us all about 10 years to figure out online advertising.
18:24.903 --> 18:27.606
I think that AI will be a similar sort of progression.
18:27.606 --> 18:29.174
It's going to take a while.
18:29.174 --> 18:32.377
And what's the future looking like to you, Annie, at this point as far as where
18:32.377 --> 18:35.948
you'd like to see the industry go and what you're seeing so far, and maybe
18:35.948 --> 18:39.518
catalysts that you're looking for?
18:39.518 --> 18:42.754
I'm really excited about the future.
18:42.754 --> 18:46.758
Mark and I talk about this all the time. When you have smart people working
18:46.758 --> 18:49.695
on interesting problems that they're really passionate about really good things
18:49.695 --> 18:52.931
happen.
18:52.931 --> 18:57.302
It's really about use cases applications and making it
18:57.302 --> 19:02.875
really useful for more niche audiences.
19:02.875 --> 19:06.044
I think we'll see a lot of progress there.
19:06.044 --> 19:09.948
Mark, your thoughts on the future for AI and things that you'd like to see,
19:09.948 --> 19:11.216
maybe a direction.
19:11.216 --> 19:15.254
I would love to see AI solve my compliance problems for me
19:15.254 --> 19:17.856
I'd love to have an AI that can do my taxes.
19:17.856 --> 19:20.759
You know there's a lot of advisors nodding, not about you but about their own
19:20.759 --> 19:23.629
practice just to make sure that things are monitored correctly.
19:23.629 --> 19:24.763
Where are we with that?
19:24.763 --> 19:28.433
I don't think we're anywhere yet but these are the types of pain points that AI
19:28.433 --> 19:30.636
is perfect for, doing your taxes for you.
19:30.636 --> 19:34.740
All these painful things that you have to do in your life, it's usually
19:34.740 --> 19:38.677
grunt work and it's paperwork and it's process driven and it's really hard and
19:38.677 --> 19:42.514
it takes up a ton of time. There's so many of those tasks that AI could really
19:42.514 --> 19:45.350
solve and that would make all of our lives better so we could go play
19:45.350 --> 19:49.321
pickleball or do something else. I don't know, I think that there's a
19:49.321 --> 19:51.657
lot of upside to the AI.
19:51.657 --> 19:56.028
Really, it's about creating balance in people's practices and peoples' lives.
19:56.028 --> 19:58.764
Can you talk, Mark, from a trading standpoint?
19:58.764 --> 20:00.966
Not exactly what you're trading, what you're buying and selling, we wouldn't go
20:00.966 --> 20:02.668
there but how do you ...
20:02.668 --> 20:06.805
with the craziness in the market, the volatility that we see,
20:06.805 --> 20:10.676
how do you work with your traders? This is a high octane area of the
20:10.676 --> 20:14.279
marketplace. How do you make sure you're getting the best fills, if you will,
20:14.279 --> 20:16.148
but also efficiency for your portfolios?
20:16.148 --> 20:18.150
That is a question you need to ask my trading desk.
20:18.150 --> 20:22.187
They have a myriad of tools that allow
20:22.187 --> 20:24.590
them to optimize for all of this stuff.
20:24.590 --> 20:29.194
Again, AI is a great application for all things trading
20:29.194 --> 20:31.063
and I know there are companies using it.
20:31.063 --> 20:34.733
There are companies using quantum for trading.
20:34.733 --> 20:39.304
There are many ways to use this technology
20:39.304 --> 20:41.840
and I think that we just haven't seen them all yet.
20:41.840 --> 20:45.310
It's interesting, when you talk to the big companies they're like, demand is
20:45.310 --> 20:48.814
insatiable for AI. Of course, everyone else is like, well, it's a bubble.
20:48.814 --> 20:51.917
Then you talk like Microsoft and they're like it's insatiable demand, we can't
20:51.917 --> 20:55.754
meet at all. I think because everyone is experimenting with all these different
20:55.754 --> 20:59.992
applications and they are having success but they're not necessarily
20:59.992 --> 21:03.795
telling anyone. They'll go to Annie and they'll say, hey, I got this use case
21:03.795 --> 21:06.531
and I got 300 million in revenue and you've never heard of me before.
21:06.531 --> 21:09.067
Annie's like, what?
21:09.067 --> 21:13.338
I think there are a lot of those companies and a lot those applications
21:13.338 --> 21:17.743
that we will see over the next three years evolve
21:17.743 --> 21:21.713
and then we will all look back and go, oh yeah, yeah, it was used for
21:21.713 --> 21:24.082
all these things and we didn't know.
21:24.082 --> 21:27.052
Annie, volume of information that comes to you.
21:27.052 --> 21:31.089
I know that AI helps but companies that are private must realize Fidelity
21:31.089 --> 21:35.060
is additive to their business, how do you make
21:35.060 --> 21:38.797
that efficient? We talked earlier about weeding out the companies who are less
21:38.797 --> 21:42.834
desirable but when all this information's coming to you there must be a line of
21:42.834 --> 21:45.971
questions, you probably can't tell us everything but there must be an efficient
21:45.971 --> 21:47.773
way that you assess these companies.
21:47.773 --> 21:50.075
Can you talk to us a bit about that?
21:50.075 --> 21:54.546
In terms of big picture, one, it's like what's
21:54.546 --> 21:58.817
the customer pain point that you're solving.
21:58.817 --> 22:02.721
A real example is ChatGPT, 800 million users, clearly, they're solving a
22:02.721 --> 22:05.023
problem for someone.
22:05.023 --> 22:09.194
I think the second part is just around the team.
22:09.194 --> 22:12.898
Are they executing, do they have some differentiated insight?
22:12.898 --> 22:15.400
I know we've talked about Anthropic a little bit.
22:15.400 --> 22:19.504
Early on they realized how important alignment and safety is to their
22:19.504 --> 22:23.742
enterprise customers and so they invested very heavily in those efforts.
22:23.742 --> 22:27.913
If you're going to have AI help you with decision making for HR or legal
22:27.913 --> 22:31.883
you want to make sure that there's guardrails that are
22:31.883 --> 22:33.919
built in place.
22:33.919 --> 22:40.592
I think the third one is just around
22:40.592 --> 22:43.428
do they have some type of differentiated insight?
22:43.428 --> 22:47.699
What is your right to win in this very, very competitive space?
22:47.699 --> 22:50.702
Those are some of the things that I think about as we meet with these really
22:50.702 --> 22:52.804
exciting companies.
22:52.804 --> 22:54.473
Excellent. Let's talk about those data centres that I was talking about
22:54.473 --> 22:58.143
earlier. I've seen they're a million to two million square feet.
22:58.143 --> 23:00.345
They're in backyards of homes.
23:00.345 --> 23:03.248
They're probably causing brownouts on electrical grids.
23:03.248 --> 23:07.219
This is a huge power demand, or is it something that's
23:07.219 --> 23:10.055
being met okay? What are your thoughts on that, Mark, as far as power
23:10.055 --> 23:12.891
consumption and criticism about that?
23:12.891 --> 23:15.160
I have a long answer for this.
23:15.160 --> 23:18.964
I'll try and shorten it. I think in the near term it is an issue.
23:18.964 --> 23:23.201
There is a power crunch. Power is an engineering problem that is
23:23.201 --> 23:27.339
solvable. There are a myriad of private companies currently working
23:27.339 --> 23:29.708
on lower power options.
23:29.708 --> 23:33.678
If you think about Google TPU versus Nvidia GPU, the TPU is
23:33.678 --> 23:35.614
much less power hungry.
23:35.614 --> 23:39.317
Power is a variable that every engineer right now is trying to solve.
23:39.317 --> 23:41.953
That's the biggest cost for every single data centre.
23:41.953 --> 23:46.024
I think that what folks should think about is 5 to
23:46.024 --> 23:48.560
10 years from now they will have engineered better power.
23:48.560 --> 23:52.731
The GPUs will be more efficient on a per watt basis.
23:52.731 --> 23:55.300
That's a silver lining from all of this, really.
23:55.300 --> 23:57.736
100% it's the biggest cost.
23:57.736 --> 24:01.840
It's like airlines trying to figure out how to get the fuel load
24:01.840 --> 24:03.675
down. The same with AI.
24:03.675 --> 24:07.345
For me, I don't think power is a long term structural problem.
24:07.345 --> 24:11.316
I think it is a short term bottleneck that's going to be around for two years
24:11.316 --> 24:15.120
or so and they're going to engineer solutions to get around this power
24:15.120 --> 24:20.292
bottleneck with better chips, better design, better everything.
24:20.292 --> 24:22.694
Yes, there's a lot of data centres and they have some issues.
24:22.694 --> 24:26.731
I don't know that we're going to need millions of power plants
24:26.731 --> 24:29.100
to feed these data centres. I think they're just going to get more power
24:29.100 --> 24:31.336
efficient, but we'll see. It depends, right?
24:31.336 --> 24:36.208
If the applications for AI turn out to be incredibly numerous
24:36.208 --> 24:39.044
maybe we do need more data centres. This is the debate that's currently going
24:39.044 --> 24:42.881
on but I don't think power is the problem.
24:42.881 --> 24:47.152
Some who are old enough like me who remember the buildup in the late '90s,
24:47.152 --> 24:48.487
remember fibre optic cables--
24:48.487 --> 24:49.221
I remember.
24:49.221 --> 24:52.424
--telecommunications. It was all about the number of terabits and whatever the
24:52.424 --> 24:54.159
other jargon was that they would carry.
24:54.159 --> 24:57.796
Then the determination was made we don't need that much capacity.
24:57.796 --> 25:01.333
That, obviously, was part of the reason for the demise of the market for that
25:01.333 --> 25:05.437
period. These data centres, there'd be some who would think these
25:05.437 --> 25:07.239
are too big, there's too many of them.
25:07.239 --> 25:12.577
Annie,
do you think that that's a case from a capacity standpoint or are they being utilized quite well?
25:12.577 --> 25:16.581
Well, every company that we've talked to for the last two years has been
25:16.581 --> 25:19.584
telling us we need more compute, we need more compute, we need more compute.
25:19.584 --> 25:22.354
This is across the board. I don't think we've met a single company that hasn't
25:22.354 --> 25:24.723
said that.
25:24.723 --> 25:27.893
Those are the data points that I have for today.
25:27.893 --> 25:32.197
Quantum computing, a question's come in, how does that integrate with AI?
25:32.197 --> 25:36.101
That's a good question. I have no idea. I don't think that they are related.
25:36.101 --> 25:39.004
I think that's a complementary technology that people are getting excited
25:39.004 --> 25:42.941
about. Just to go back to your previous question a little bit,
25:42.941 --> 25:47.012
I think AI is one part
25:47.012 --> 25:50.649
of the story for data centres but the other part of the story, and Jensen at
25:50.649 --> 25:54.719
Nvidia will tell you this, the laws of physics have made the CPU
25:54.719 --> 25:58.790
irrelevant. We've hit the max, we can't get any smaller.
25:58.790 --> 26:03.161
In order to have more compute for even just the basics of what we do today
26:03.161 --> 26:07.299
we need to move to these GPUs and this parallel processing and
26:07.299 --> 26:09.267
all these new things.
26:09.267 --> 26:13.238
The data centre requirement is not just AI, it's just
26:13.238 --> 26:17.208
general compute has to increase for all the things we
26:17.208 --> 26:20.045
do already. You've got a couple of things going on.
26:20.045 --> 26:24.583
You have AI as an accelerator of demand but you also have the CPU
26:24.583 --> 26:28.687
sort of dying out as a workhorse for overall data
26:28.687 --> 26:30.889
centre stuff.
26:30.889 --> 26:33.525
You've got more than one thing going on and I think that's why we're probably
26:33.525 --> 26:37.195
going to see data centre buildouts at a pretty large scale for quite a long
26:37.195 --> 26:38.263
time.
26:38.263 --> 26:41.466
Hello, investors. We'll be back to the show in just a moment.
26:41.466 --> 26:45.036
I wanted to share that here at Fidelity, we value your opinion.
26:45.036 --> 26:47.339
Please take a few minutes to help us shape the future of Fidelity Connects
26:47.339 --> 26:52.444
podcasts. Complete our listener survey by visiting fidelity.ca/survey,
26:52.444 --> 26:54.412
and you could win one of our branded tumblers.
26:54.412 --> 26:57.482
Periodic draws ending by March 30th, 2026.
26:57.482 --> 27:01.086
And don't forget to listen to Fidelity Connects, the Upside, and French
27:01.086 --> 27:05.123
DialoguesFidelity podcasts available on Apple, Spotify, YouTube, or wherever
27:05.123 --> 27:09.327
else you get your podcasts. Now back to today's show.
27:09.327 --> 27:13.565
You've both made it very clear that AI is here to stay and it's got a
27:13.565 --> 27:17.569
huge future, lots to learn, lots of ebbs and flows all over the place.
27:17.569 --> 27:21.640
Mark, top three things that you would say are on your radar right now from
27:21.640 --> 27:23.074
an AI standpoint.
27:23.074 --> 27:25.377
I'm always looking for applications, and so is Annie.
27:25.377 --> 27:27.679
We're always trying to figure out who's the next application.
27:27.679 --> 27:31.616
I think that the magic moment for me this year was when we saw Claude for code
27:31.616 --> 27:35.654
really accelerate and the engineers embrace it and use it, and the tokens
27:35.654 --> 27:37.455
account just went crazy.
27:37.455 --> 27:40.558
We are actively looking for the next big applications.
27:40.558 --> 27:44.496
We may not find it, we keep looking, but
27:44.496 --> 27:48.233
it could be that it's just a whole bunch of little applications, it could be, I
27:48.233 --> 27:52.470
don't know. That is what I spend most of my day thinking about, who's
27:52.470 --> 27:56.975
using AI, how are they using it, who wins and who loses from that,
27:56.975 --> 28:00.945
and then trying to build a portfolio such that I can take advantage of
28:00.945 --> 28:02.847
that. Very hard to do.
28:02.847 --> 28:06.951
We don't have that much visibility but if we don't any visibility good luck
28:06.951 --> 28:10.555
to anyone else. We're in the heart of it all and we're seeing all of the
28:10.555 --> 28:12.424
companies doing it.
28:12.424 --> 28:13.825
That's what we're spending our time on.
28:13.825 --> 28:16.995
Annie, what would you add to that list of key things that are on your radar as
28:16.995 --> 28:18.596
far as AI?
28:18.596 --> 28:22.534
I think it's use cases and also just smart people continuing to be excited
28:22.534 --> 28:24.869
and building in the space.
28:24.869 --> 28:28.239
We talk to engineers, founders, so many different people.
28:28.239 --> 28:32.510
They can see things in advance of what I can see
28:32.510 --> 28:36.381
in terms of the roadmap. There's just so much excitement.
28:36.381 --> 28:38.316
I'm excited about the year ahead.
28:38.316 --> 28:42.353
What's exciting too is how you source information and by seeing that
28:42.353 --> 28:47.192
piece of paper on a lamppost, or what have you, in 2022, you're
28:47.192 --> 28:50.729
still looking at the nooks and crannies, aren't you, for what other aspects are
28:50.729 --> 28:54.666
complementary or new and interesting things to talk about
28:54.666 --> 28:58.703
through social media, through something as simple as posting on
28:58.703 --> 29:01.306
a billboard or what have you.
29:01.306 --> 29:05.310
I was listening to a podcast yesterday. It's a next gen AI company
29:05.310 --> 29:09.447
and they want to reach a younger audience so they actually
29:09.447 --> 29:11.850
post a lot of stuff through TikTok.
29:11.850 --> 29:14.119
I have friends sending me stuff on Twitter all the time.
29:14.119 --> 29:18.123
We have other investors that are making introductions
29:18.123 --> 29:20.058
through the company that Mark was talking about.
29:20.058 --> 29:22.861
He had actually texted his son this morning asking have you heard of this
29:22.861 --> 29:26.798
coding review company, the Series A company that had
29:26.798 --> 29:30.535
been referred to us by one of their customers and investors.
29:30.535 --> 29:34.539
We've been really lucky, one of the best parts of working
29:34.539 --> 29:38.877
at Fidelity is you get to talk to almost anyone.
29:38.877 --> 29:40.612
We're really lucky to have good access.
29:40.612 --> 29:42.380
Can I give you a fun story?
29:42.380 --> 29:43.648
I love fun stories.
29:43.648 --> 29:48.052
I was watching the Seattle Mariners game at my local bar,
29:48.052 --> 29:49.387
Harry's.
29:49.387 --> 29:50.388
Of course it's named Harry's.
29:50.388 --> 29:54.292
It's called Harry's Bar. I was sitting there with my wife watching the Jays
29:54.292 --> 29:58.429
play the Mariners. At the table next to me was a robotics company
29:58.429 --> 30:02.667
from Chicago talking to a VC who was wearing his Seattle Mariners hat.
30:02.667 --> 30:04.936
They were doing the meeting while the game was on.
30:04.936 --> 30:07.071
My wife was like this and she nudged me, she's like, hey, Mark, robotics,
30:07.071 --> 30:12.143
they're talking robotics.
30:12.143 --> 30:16.548
About the sixth inning, we weren't ahead yet, I
30:16.548 --> 30:19.517
said, you guys in robotics?
30:19.517 --> 30:23.454
I work at Fidelity, private, so we talked for 20 minutes about their little
30:23.454 --> 30:25.924
robotics company in Chicago.
30:25.924 --> 30:29.027
That's the kind of thing that happens in the Bay Area.
30:29.027 --> 30:33.164
They flew in just to meet VCs and they're like, this is amazing, we're
30:33.164 --> 30:36.534
at the bar and we're talking to Fidelity. I can't remember who the other VC
30:36.534 --> 30:39.637
was. That is an example.
30:39.637 --> 30:43.441
There's some real logic to that and what it takes me back to is One Up On Wall
30:43.441 --> 30:47.412
Street, Peter Lynch talking about just absorbing from
30:47.412 --> 30:50.081
what's around you, family members, friends and so on.
30:50.081 --> 30:53.952
That was in an analog way, now it's in a digital way or simply at Harry's Bar.
30:53.952 --> 30:55.753
Harry's Bar watching the baseball game.
30:55.753 --> 30:56.855
It's crazy.
30:56.855 --> 30:59.290
Annie, as we wrap are there key thoughts that you want to leave with our
30:59.290 --> 31:03.094
audience today on what it is that you're doing as director of private equities
31:03.094 --> 31:05.296
but also for the future that you see?
31:05.296 --> 31:09.801
Absolutely. I think I have just one of the best jobs at Fidelity.
31:09.801 --> 31:12.237
I feel so lucky to be where I am and.
31:12.237 --> 31:14.873
There's just so much excitement in this space.
31:14.873 --> 31:17.108
We're just talking to the people that are building in the space.
31:17.108 --> 31:20.778
I think Mark described it really well, there's the finance folks and then there
31:20.778 --> 31:24.816
are the people who are building the space and they're
31:24.816 --> 31:28.519
just really passionate about what's ahead so that makes me excited as well.
31:28.519 --> 31:31.422
You talk about it with a big smile and that shows that there's a lot of
31:31.422 --> 31:33.625
interest and a lot of excitement and a lot of future as well.
31:33.625 --> 31:36.194
Mark, some key takeaways for our audience today.
31:36.194 --> 31:40.164
I think that AI is here to stay and it's going to revolutionize everything
31:40.164 --> 31:44.035
you do. If you're a planner it's going to take away all those terrible, awful
31:44.035 --> 31:47.705
compliance and paperwork and email chains that you have.
31:47.705 --> 31:49.874
I know I've seen some of them.
31:49.874 --> 31:53.845
I think this is going to make your life better. I'm not scared
31:53.845 --> 31:56.014
of it, I'm really not.
31:56.014 --> 31:59.651
Science fiction makes us scared of AI but I think the use cases are just going
31:59.651 --> 32:02.921
to get rid of all the grunt work, just the stuff that we hate to do.
32:02.921 --> 32:05.690
I think it's going to happen quickly.
32:05.690 --> 32:09.093
What I would say too is, and this is what I tell a lot of the youngsters,
32:09.093 --> 32:11.696
embrace it, it's not going away.
32:11.696 --> 32:15.900
Figure out how to use it before everyone else does and just
32:15.900 --> 32:19.103
love it because it's going to get better and you're going to need to use it to
32:19.103 --> 32:20.805
compete so embrace it.
32:20.805 --> 32:24.075
That's a great takeaway from both of you. I love the embrace it side of it
32:24.075 --> 32:28.112
because if it's not going away figure out for me, for you, for you it's
32:28.112 --> 32:30.581
going to be used differently but it's going to be additive and it's going to
32:30.581 --> 32:33.985
allow us to play pickleball or whatever else you want to do.
32:33.985 --> 32:37.422
Excellent. Annie and Mark, thank you so much for joining us today.
32:37.422 --> 32:39.724
It was very interesting and I hope you enjoyed it as well.
32:39.724 --> 32:40.358
Thank you so much.
32:40.358 --> 32:41.359
Thank you so much.
32:41.359 --> 32:45.296
Thanks for watching or listening to the Fidelity Connects
32:45.296 --> 32:49.434
podcast. Now if you haven't done so already, please subscribe to Fidelity
32:49.434 --> 32:52.236
Connects on your podcast platform of choice.
32:52.236 --> 32:55.073
And if you like what you're hearing, please leave a review or a five-star
32:55.073 --> 32:59.043
rating. Fidelity Mutual Funds and ETFs are available by working with
32:59.043 --> 33:02.413
a financial advisor or through an online brokerage account.
33:02.413 --> 33:06.117
Visit fidelity.ca/howtobuy for more information.
33:06.117 --> 33:09.954
While on Fidelity.ca, you can also find more information on future live
33:09.954 --> 33:14.092
webcasts. And don't forget to follow Fidelity Canada on YouTube, LinkedIn,
33:14.092 --> 33:16.094
and Instagram.
33:16.094 --> 33:18.963
We'll end today's show with a short disclaimer.
33:18.963 --> 33:22.800
The views and opinions expressed on this podcast are those of the participants,
33:22.800 --> 33:26.738
and do not necessarily reflect those of Fidelity Investments Canada ULC or
33:26.738 --> 33:30.742
its affiliates. This podcast is for informational purposes only, and should not
33:30.742 --> 33:33.277
be construed as investment, tax, or legal advice.
33:33.277 --> 33:35.580
It is not an offer to sell or buy.
33:35.580 --> 33:39.917
Or an endorsement, recommendation, or sponsorship of any entity or securities
33:39.917 --> 33:44.722
cited. Read a fund's prospectus before investing, funds are not guaranteed.
33:44.722 --> 33:48.292
Their values change frequently, and past performance may not be repeated.
33:48.292 --> 33:52.130
Fees, expenses, and commissions are all associated with fund investments.
33:52.130 --> 33:54.432
Thanks again. We'll see you next time.
