1
00:00:00,010 --> 00:00:00,260
Mr.
2
00:00:00,260 --> 00:00:02,470
Higher for longer, for a very long time, .
3
00:00:02,470 --> 00:00:04,222
And he delivered on that promise.
4
00:00:05,181 --> 00:00:08,433
Torsten Slok is one of the most highly watched economists on Wall Street, and he's
5
00:00:08,433 --> 00:00:12,478
been saying for a while that interest rates would be higher for longer.
6
00:00:12,478 --> 00:00:15,898
And now that's certainly the case as the Fed starts a hiking cycle.
7
00:00:15,898 --> 00:00:20,610
So what does that mean for some of the most critical parts of the economy,
8
00:00:20,610 --> 00:00:22,153
particularly artificial intelligence?
9
00:00:22,153 --> 00:00:25,990
This AI thing better work out because otherwise we'll have problems both in the economy and
10
00:00:25,990 --> 00:00:27,991
also in financial markets.
11
00:00:44,213 --> 00:00:46,173
Welcome to the Bridge by iCapital.
12
00:00:46,173 --> 00:00:51,052
I'm Sonali Basak, the Chief Investment Strategist at iCapital, and today I am joined by
13
00:00:51,052 --> 00:00:54,221
Torsten Slok, the chief economist at Apollo Global Management.
14
00:00:54,221 --> 00:00:57,057
He has been focusing a lot on artificial intelligence.
15
00:00:57,099 --> 00:01:01,144
And Torsten, thank you for joining because we're in this moment where you've been talking
16
00:01:01,144 --> 00:01:04,313
about it a lot, this concentration of AI in portfolios.
17
00:01:04,313 --> 00:01:06,023
Here's my question to you.
18
00:01:06,023 --> 00:01:09,567
We are in an era where borrowing costs have been rising.
19
00:01:09,567 --> 00:01:12,904
What does that mean for the AI trade?
20
00:01:12,904 --> 00:01:16,532
Well, the risk when interest rates go up is that any business that has cash
21
00:01:16,532 --> 00:01:20,160
flows far out in the future is going to be more vulnerable because when the
22
00:01:20,160 --> 00:01:23,746
discount rate, meaning interest rate starts to go up, that means that cash flows have
23
00:01:23,746 --> 00:01:26,165
to be discounted in a way where they're worth less.
24
00:01:26,165 --> 00:01:30,335
Because now I can take my money, put it into fixed income and get higher
25
00:01:30,335 --> 00:01:30,919
interest rate.
26
00:01:30,919 --> 00:01:35,130
So that raises the bar for cash flows, and especially for equity that pays off
27
00:01:35,130 --> 00:01:36,840
far out in the future.
28
00:01:36,840 --> 00:01:40,677
And that's very important for AI because a lot of the investments in AI, especially
29
00:01:40,677 --> 00:01:44,513
AI and venture capital, are only cash flows that come far, far out in the
30
00:01:44,513 --> 00:01:44,763
future.
31
00:01:44,763 --> 00:01:49,267
So that's why when interest rates go up, it does increase the bar for all
32
00:01:49,267 --> 00:01:53,479
other investments because suddenly I can get a risk-free interest rate or an interest rate
33
00:01:53,479 --> 00:01:57,691
or yield in public credit, private credit, fixed income, which is higher, and therefore raises
34
00:01:57,691 --> 00:02:00,485
the bar for what I can get in other investments.
35
00:02:00,485 --> 00:02:04,697
So higher interest rates has been making it a bit more difficult for the AI
36
00:02:04,697 --> 00:02:04,988
trade.
37
00:02:04,988 --> 00:02:05,239
Well,
38
00:02:05,239 --> 00:02:09,284
It's so interesting what you're saying it sounds like is just credit is on average
39
00:02:09,284 --> 00:02:12,953
or on balance, rather, a little more attractive than equities in the scenario.
40
00:02:12,953 --> 00:02:16,540
Yeah, and if you think about that from a Federal Reserve perspective, this is actually
41
00:02:16,540 --> 00:02:18,541
the whole idea with raising interest rates.
42
00:02:18,541 --> 00:02:21,752
If someone from the Fed were sitting here next to you and me, they would
43
00:02:21,752 --> 00:02:25,005
say, but the reason why we are raising interest rates is that we want to
44
00:02:25,005 --> 00:02:27,507
slow the economy down to get inflation to slowly come down.
45
00:02:27,507 --> 00:02:27,757
Yeah.
46
00:02:27,757 --> 00:02:31,469
So that's why they raise in small increments of 25 basis points because they want
47
00:02:31,469 --> 00:02:33,220
to do this very, very gradually.
48
00:02:33,220 --> 00:02:37,474
And when you think about that from a 60/40 asset allocation perspective, that means
49
00:02:37,474 --> 00:02:41,727
that the Fed is really saying, please take a little bit money out of public
50
00:02:41,727 --> 00:02:45,981
equities and put it instead into fixed income, higher quality credit, things that pay you
51
00:02:45,981 --> 00:02:50,068
a higher level of yield because that is now more attractive in relative terms.
52
00:02:50,068 --> 00:02:50,443
We'll
53
00:02:50,443 --> 00:02:54,530
Talk more about the Federal Reserve because to your credit, you've been talking about higher
54
00:02:54,530 --> 00:02:58,616
for longer, for a very long time, well before the market started even baking in
55
00:02:58,616 --> 00:03:00,493
interest rate hikes for example.
56
00:03:00,493 --> 00:03:04,997
But on AI, what is at stake if the, if the trade really does slow
57
00:03:04,997 --> 00:03:05,831
down?
58
00:03:05,831 --> 00:03:06,957
That's a lot at stake.
59
00:03:06,957 --> 00:03:10,460
If the AI trade does slow down, first of all, a lot of GDP growth
60
00:03:10,460 --> 00:03:12,336
is driven by AI spending at the moment.
61
00:03:12,419 --> 00:03:16,965
We calculate that GDP growth, which is normally 2% today, roughly 1% is point of
62
00:03:16,965 --> 00:03:17,632
that namely
63
00:03:17,632 --> 00:03:19,509
half of that is because of AI.
64
00:03:19,509 --> 00:03:23,595
And it's not only the building of data centers and energy associated with data centers,
65
00:03:23,595 --> 00:03:24,972
it's also spending on tokens.
66
00:03:24,972 --> 00:03:25,389
Mm-hmm .
67
00:03:25,389 --> 00:03:27,807
It's also spending on AI tools for consumers.
68
00:03:27,807 --> 00:03:30,143
It's also spending on AI tools for businesses.
69
00:03:30,143 --> 00:03:34,271
And it's also the wealth effect that has come along with ai naming home price
70
00:03:34,271 --> 00:03:38,441
in San Francisco have been going up because there's a lot of wealth generated as
71
00:03:38,441 --> 00:03:39,567
a result of ai.
72
00:03:39,567 --> 00:03:43,695
So the first answer to your question is if AI does begin to slow down,
73
00:03:43,695 --> 00:03:45,989
it would also slow down the economy.
74
00:03:45,989 --> 00:03:50,910
And most importantly, really is also that AI is absolutely also now everywhere in people's
75
00:03:50,910 --> 00:03:55,872
asset allocation. The equity market, the 10 biggest stocks, they make up 40% of the
76
00:03:55,872 --> 00:03:56,331
index.
77
00:03:56,331 --> 00:03:59,876
And most of the returns in s and p 500 in the last several years
78
00:03:59,876 --> 00:04:01,252
has been driven by ai.
79
00:04:01,252 --> 00:04:05,630
Similarly, in fixed income, if you look at especially public credit, you also have that
80
00:04:05,630 --> 00:04:07,423
hyperscalers have been issuing more debt.
81
00:04:07,423 --> 00:04:08,758
That is also now more AI.
82
00:04:08,758 --> 00:04:13,929
And finally, if you also own venture capital, venture capital used to be pharma, biotech,
83
00:04:13,929 --> 00:04:17,891
prescription drugs, but today 87% of venture capital is also AI.
84
00:04:17,891 --> 00:04:21,936
So the answer to your question is, if AI does not work out, we will
85
00:04:21,936 --> 00:04:24,062
have some problems for the GDP outlook.
86
00:04:24,062 --> 00:04:26,272
We will also have some problems from an investing perspective.
87
00:04:26,272 --> 00:04:30,359
So the conclusion is this AI thing better work out because otherwise we'll have problems
88
00:04:30,359 --> 00:04:33,237
both in the economy and also in financial markets.
89
00:04:33,237 --> 00:04:34,780
Let's define workout, right?
90
00:04:34,780 --> 00:04:36,781
Because you're saying there's a slowdown.
91
00:04:36,781 --> 00:04:39,158
A slowdown is very different from a crash.
92
00:04:39,158 --> 00:04:42,494
I think a lot of people are very nervous out there about the worst scenarios
93
00:04:42,494 --> 00:04:44,079
that could happen in a slowdown here.
94
00:04:44,121 --> 00:04:47,582
And I think that's the art of the moment that we're in is crafting that
95
00:04:47,582 --> 00:04:49,000
landing.
96
00:04:49,000 --> 00:04:53,545
So when you think about what a slowdown really means, you know, does it mean
97
00:04:53,545 --> 00:04:57,674
that, you know, we might see some companies fall into really meaningful distress.
98
00:04:57,674 --> 00:05:02,719
Does it mean that we'll see even a probability of a recession start to rise?
99
00:05:02,719 --> 00:05:03,804
I mean, what does that actually mean to you?
100
00:05:03,804 --> 00:05:04,221
Yeah, so
101
00:05:04,221 --> 00:05:07,307
There are four different buckets of investing in AI.
102
00:05:07,307 --> 00:05:11,810
One is of course the frontier labs and the models and those of course that
103
00:05:11,810 --> 00:05:15,105
are exposed directly to the consumer and the user of AI.
104
00:05:15,105 --> 00:05:16,856
These are Open AI, Anthropic,
105
00:05:16,856 --> 00:05:17,565
For example.
106
00:05:17,565 --> 00:05:21,235
They have negative profit margins at the moment because they're developing the technology.
107
00:05:21,235 --> 00:05:22,653
And that of course is expensive.
108
00:05:22,653 --> 00:05:27,365
That's why those that are facing the consumer, which is unusual, have negative profit margins.
109
00:05:27,365 --> 00:05:29,909
The next step in the value chain is the hyperscalers.
110
00:05:29,909 --> 00:05:33,954
Those who are building the data centers, they have positive profit margins, but they of
111
00:05:33,954 --> 00:05:38,040
course are also an important part of whether AI working out or not is going
112
00:05:38,040 --> 00:05:39,333
to be impacting them.
113
00:05:39,333 --> 00:05:41,085
The next step is energy and grit.
114
00:05:41,085 --> 00:05:43,628
So that means that this provides the energy to AI.
115
00:05:43,628 --> 00:05:47,382
They are of course also at risk to the upside and downside as a function
116
00:05:47,382 --> 00:05:49,300
of whether AI is succeeding or not.
117
00:05:49,300 --> 00:05:53,345
And finally we have silicon and equipment, meaning those who built the chips and the
118
00:05:53,345 --> 00:05:57,390
semis, and they of course at the ultimate end because they're the, the ones that
119
00:05:57,390 --> 00:06:01,477
all the models are running on, whether it's closed source, open source, of course the
120
00:06:01,477 --> 00:06:04,521
chips are ultimately what is solving all the problems in AI.
121
00:06:04,521 --> 00:06:08,733
And the answer to your question is different parts of that value chain can absolutely
122
00:06:08,733 --> 00:06:12,652
get differently impacted as a result of what AI not working out exactly means.
123
00:06:12,652 --> 00:06:16,656
Let's talk about that a little more because you had mentioned that Frontier Labs typically
124
00:06:16,656 --> 00:06:19,241
have been losing money from a margin perspective, right?
125
00:06:19,241 --> 00:06:22,828
Um, hyperscalers though that that buffer is much thicker.
126
00:06:22,828 --> 00:06:26,873
Does that mean in an environment like this, there's so many people worried about the
127
00:06:26,873 --> 00:06:31,376
free cash flow impacts the declines in free cash flow, but they're relatively insulated actually.
128
00:06:31,376 --> 00:06:33,712
So are you actually really worried about them raising money?
129
00:06:33,712 --> 00:06:34,004
Yeah.
130
00:06:34,004 --> 00:06:34,170
This
131
00:06:34,170 --> 00:06:34,504
In the market?
132
00:06:34,504 --> 00:06:38,382
Yeah, this, this is a really important discussion because at the end of the day
133
00:06:38,372 --> 00:06:41,458
today, the demand for compute is really going up significantly.
134
00:06:41,468 --> 00:06:46,597
Not only because of all of us writing questions into ChatGPT and to Claude,
135
00:06:46,597 --> 00:06:50,142
but also because of agentic AI becoming much more pronounced.
136
00:06:50,142 --> 00:06:53,186
That means that demand for compute is rising very, very significantly.
137
00:06:53,186 --> 00:06:57,315
Meaning people want to run on chips in data centers, but at the same time,
138
00:06:57,315 --> 00:07:00,400
the competitive landscape is now also changing for the models.
139
00:07:00,400 --> 00:07:04,821
Namely open source models are becoming much more used and much more pronounced in the
140
00:07:04,821 --> 00:07:06,572
landscape of those who use AI.
141
00:07:06,572 --> 00:07:11,326
Whereas closed source models, of course, are now getting more competitive pressure from the open
142
00:07:11,326 --> 00:07:12,160
source models.
143
00:07:12,160 --> 00:07:15,496
This raises all kinds of questions about what is the competitive edge for closed source
144
00:07:15,496 --> 00:07:17,331
models for relative to open source models.
145
00:07:17,331 --> 00:07:21,335
But the answer to your question is this race is demand for compute is going
146
00:07:21,335 --> 00:07:25,338
up very, very rapidly, but at the same time, the price of that compute and
147
00:07:25,338 --> 00:07:29,299
the price of tokens and ultimately the competitive landscape for those who use it has
148
00:07:29,299 --> 00:07:31,718
actually been under more downward and more competitive pressure.
149
00:07:31,718 --> 00:07:35,430
And that's the race in ai that is the most critical part of understanding whether
150
00:07:35,430 --> 00:07:37,390
it's working out or not over the next several
151
00:07:37,390 --> 00:07:37,932
Years.
152
00:07:37,932 --> 00:07:41,476
So very simply, when you watch all these hyperscalers, the, you know, um, Amazons of
153
00:07:41,476 --> 00:07:44,229
the world go out and raise debt in markets, right?
154
00:07:44,229 --> 00:07:45,396
Uh, Alphabet, right?
155
00:07:45,396 --> 00:07:47,773
Do you, do you worry about it?
156
00:07:47,773 --> 00:07:51,818
Well, markets are already voting with their feet on that question because CDS spreads on,
157
00:07:51,818 --> 00:07:54,612
uh, several of these names have been widening out.
158
00:07:54,612 --> 00:07:58,407
CDS spreads of course is measuring what is the probability that these companies might not
159
00:07:58,407 --> 00:08:01,701
be able to pay back their debt in the next five years.
160
00:08:01,701 --> 00:08:05,246
And that means that that premium meaning the price of insuring yourself against the risk
161
00:08:05,246 --> 00:08:08,791
that they will not be paying that debt, has been moving a little bit higher,
162
00:08:08,791 --> 00:08:12,377
which some people are beginning to look at, including me and say, well, we need
163
00:08:12,377 --> 00:08:13,920
to understand better why that's happening.
164
00:08:13,920 --> 00:08:18,757
Similarly, we've also seen spreads on hyperscale debt also widen out, which is also saying
165
00:08:18,757 --> 00:08:20,884
that the market has not much.
166
00:08:20,884 --> 00:08:24,554
But started to worry a little bit more about what is the trajectory that we are
167
00:08:24,554 --> 00:08:25,346
on here?
168
00:08:25,346 --> 00:08:26,972
Are the revenues going to increase?
169
00:08:26,972 --> 00:08:31,601
And if you look at the consensus expectations from FactSet today, the operating cash flow
170
00:08:31,601 --> 00:08:34,312
in the five hyperscalers is about $600 billion.
171
00:08:34,312 --> 00:08:38,065
And the consensus is expecting that over the next several years will grow from
172
00:08:38,065 --> 00:08:39,566
$600 billion to about $2 trillion.
173
00:08:39,566 --> 00:08:44,445
In other words, a very dramatic increase in the revenues and the cash flows for
174
00:08:44,445 --> 00:08:45,112
the hyperscalers.
175
00:08:45,112 --> 00:08:48,574
And that is the question, if that revenue comes in faster, comes in slower, is
176
00:08:48,574 --> 00:08:49,783
the answer to the question.
177
00:08:49,783 --> 00:08:52,744
Namely, what is the scenario we have ahead of us for
178
00:08:52,744 --> 00:08:53,119
AI?
179
00:08:53,119 --> 00:08:56,580
And it's interesting because we are talking about this as it pertains to the hyperscalers,
180
00:08:56,580 --> 00:08:58,916
but the hyperscalers we just said have a bigger buffer.
181
00:08:58,916 --> 00:08:59,624
Right?
182
00:08:59,624 --> 00:09:01,376
But what about the frontier labs?
183
00:09:01,376 --> 00:09:05,254
At the end of the day, you saw something happen where you didn't really see
184
00:09:05,254 --> 00:09:09,132
this happen in prior years where pre IPO companies are actually raising fairly meaningful amounts
185
00:09:09,132 --> 00:09:10,217
of debt.
186
00:09:10,217 --> 00:09:11,384
Is that healthy?
187
00:09:11,384 --> 00:09:14,971
Well, one first observation about this, that these companies are private.
188
00:09:14,971 --> 00:09:18,265
So this tells you and me a very important conclusion that you and I have
189
00:09:18,265 --> 00:09:21,559
talked about for many years, namely, that a very significant part of the economy is
190
00:09:21,559 --> 00:09:23,102
outside the S&P 500, right?
191
00:09:23,102 --> 00:09:26,147
And there are a lot of private companies that are wildly successful.
192
00:09:26,147 --> 00:09:30,150
In fact, if you look at businesses in the U.S. that have revenue of more
193
00:09:30,150 --> 00:09:33,611
than a hundred million dollars, 90% of them are actually not public.
194
00:09:33,611 --> 00:09:36,905
So there's a different way of saying there's a lot of private companies including an
195
00:09:36,905 --> 00:09:40,158
Open AI and Anthropic that are very successful at what they do and generate a lot of
196
00:09:40,158 --> 00:09:40,784
revenue.
197
00:09:40,784 --> 00:09:44,161
So in that sense, they are not publishing the same way.
198
00:09:44,161 --> 00:09:47,915
Of course the companies in S&P 500 are doing, but if they are about to
199
00:09:47,915 --> 00:09:51,668
do IPO, of course this will give much more insight, much more light and transparency
200
00:09:51,668 --> 00:09:55,379
in terms of where are they making money and how are they making money?
201
00:09:55,379 --> 00:09:59,257
And that will give the market a better idea about this competitive landscape that we're
202
00:09:59,257 --> 00:10:01,676
seeing between open source models and closed source models.
203
00:10:01,676 --> 00:10:05,137
The reason I ask this also is because in private markets there's the venture capital
204
00:10:05,137 --> 00:10:08,640
aspect, which we've talked about the, you know, the the growth equity ven, uh, private
205
00:10:08,640 --> 00:10:10,684
equity type aspect that we were talking about.
206
00:10:10,684 --> 00:10:14,687
But even data centers to a large degree are tied to the credit ratings of
207
00:10:14,687 --> 00:10:17,689
these firms and the credit health of these firms.
208
00:10:17,689 --> 00:10:20,734
And so there's a whole ecosystem that hinges on this question it feels like.
209
00:10:20,734 --> 00:10:25,404
And and that's why ultimately in particular from a GDP perspective, data center buildout
210
00:10:25,404 --> 00:10:28,907
is absolutely critical for anyone's forecast for GDP growth.
211
00:10:28,907 --> 00:10:33,327
So normally people would say, oh, CapEx and business spending, which data center, the data
212
00:10:33,327 --> 00:10:37,122
center buildout is, is normally a smaller issue in my forecast.
213
00:10:37,122 --> 00:10:41,209
Normally it's all about the consumer, but now your entire forecast depends almost completely on
214
00:10:41,209 --> 00:10:43,961
what is your view on the data center buildout.
215
00:10:43,961 --> 00:10:47,756
And it's very clear, yeah, that the data center buildout depends on whether the
216
00:10:47,756 --> 00:10:51,551
hyperscalers are going to generate all the revenue on the back of all the compute
217
00:10:51,551 --> 00:10:52,552
that's being created.
218
00:10:52,552 --> 00:10:56,847
So yes, there's unlimited demand for compute, but the question is over time whether the
219
00:10:56,847 --> 00:11:00,850
price of that compute is at risk potentially of eventually beginning to go down.
220
00:11:00,850 --> 00:11:04,145
Because if that begins to go down, then you will have two opposite forces.
221
00:11:04,145 --> 00:11:08,023
Yes, we all want to do more at Agentic AI and use AI, but at
222
00:11:08,023 --> 00:11:11,859
the same time if the price of that usage, especially with the competition from open
223
00:11:11,859 --> 00:11:15,738
source models starts to go down, then of course there will be some questions exactly
224
00:11:15,738 --> 00:11:19,366
about what does that then mean for the overall data center build out
225
00:11:19,366 --> 00:11:21,242
Bullish or bearish?
226
00:11:21,242 --> 00:11:24,870
I am still bullish on the economic output because I still think that the most
227
00:11:24,870 --> 00:11:27,497
important impact of AI is also on the labor market.
228
00:11:27,497 --> 00:11:31,667
That yes, there will be some people that likely will lose their jobs, but I
229
00:11:31,667 --> 00:11:35,879
think that a much more important effect on the labor market is that the moment
230
00:11:35,879 --> 00:11:40,049
today in the census data, we are seeing the most businesses created in U.S. economy
231
00:11:40,049 --> 00:11:41,467
ever in U.S. history.
232
00:11:41,467 --> 00:11:46,305
In other words, the business dynamics that have appeared because of AI have just been
233
00:11:46,305 --> 00:11:46,847
incredible.
234
00:11:46,847 --> 00:11:49,474
People are leaving finance to open their own business.
235
00:11:49,474 --> 00:11:54,144
They're leaving consulting, they're leaving media, they're leading or leaving of course legal services and
236
00:11:54,144 --> 00:11:58,565
they go home to their basement and they basically start a solo founder firm.
237
00:11:58,565 --> 00:12:01,192
There are many more solo founder firms now than multi founder.
238
00:12:01,192 --> 00:12:04,695
And that is of course if some of these businesses, it's just a fraction of
239
00:12:04,695 --> 00:12:08,156
the many business that are created at the moment end up creating some success, they
240
00:12:08,156 --> 00:12:09,657
will also create more jobs.
241
00:12:09,657 --> 00:12:13,244
And I think when we put that up on the scale, that will be positive
242
00:12:13,244 --> 00:12:13,911
for employment.
243
00:12:13,911 --> 00:12:14,119
And
244
00:12:14,119 --> 00:12:14,870
You think that sticks?
245
00:12:14,870 --> 00:12:15,370
I truly, even
246
00:12:15,370 --> 00:12:19,457
The last, I truly think AI is a miracle drug that will both create higher
247
00:12:19,457 --> 00:12:20,833
productivity and also higher employment.
248
00:12:20,833 --> 00:12:23,210
And that's by the way also what the data is showing so far.
249
00:12:23,210 --> 00:12:26,671
The employment report, as you and I have talked about, continues to be strong, has
250
00:12:26,671 --> 00:12:30,133
been strong for the last six months so far at this point, there is no
251
00:12:30,133 --> 00:12:33,594
effect of the labor displacement effect, meaning the effect that people are losing their jobs
252
00:12:33,594 --> 00:12:35,012
because of AI being particularly significant.
253
00:12:35,012 --> 00:12:38,181
Okay, one more, one more question on AI before we move to the macro.
254
00:12:38,181 --> 00:12:40,600
We'll put some AI in there too because it's so related.
255
00:12:40,600 --> 00:12:45,229
But a tough question I had gotten this week from a large investor in your
256
00:12:45,229 --> 00:12:47,272
space was what happens?
257
00:12:47,272 --> 00:12:51,442
You have multiple companies now, three of them at least, that have surged past a
258
00:12:51,442 --> 00:12:55,612
trillion dollars in value in a timeframe we have never seen before reaching almost
259
00:12:55,612 --> 00:12:58,781
$2 trillion in some of those cases.
260
00:12:58,781 --> 00:13:02,243
What happens if one of those companies doesn't exist in the next five years?
261
00:13:02,243 --> 00:13:02,535
Yeah,
262
00:13:02,535 --> 00:13:03,911
This is really, really important.
263
00:13:03,911 --> 00:13:08,164
So that's also why the famous discussion at the moment is about the moat.
264
00:13:08,164 --> 00:13:12,626
What is really the moat that different businesses, especially in the tech world have?
265
00:13:12,626 --> 00:13:16,838
Because if the tech companies, the broadly speaking, the magnificent seven and the hyperscalers, if
266
00:13:16,838 --> 00:13:21,008
they do not have a moat where they can protect themselves, then of course they
267
00:13:21,008 --> 00:13:24,553
will ultimately begin to shrink and they will be more at risk.
268
00:13:24,553 --> 00:13:28,264
So the question of course to ask will all the hyperscalers survive?
269
00:13:28,264 --> 00:13:32,893
Are some of them at risk of going down because they might not have the
270
00:13:32,893 --> 00:13:37,564
right moat or are we seeing all of them surviving and growing into even much
271
00:13:37,564 --> 00:13:38,356
bigger companies?
272
00:13:38,356 --> 00:13:42,526
And that becomes hugely a function of this question around is it open source models
273
00:13:42,526 --> 00:13:43,694
that are winning?
274
00:13:43,694 --> 00:13:45,487
Is it closed source models that are winning?
275
00:13:45,487 --> 00:13:47,322
What is regulation going to look like?
276
00:13:47,322 --> 00:13:48,615
Are we going to see regulation?
277
00:13:48,615 --> 00:13:52,284
There's a lot of talk at the moment about human extinction, maybe everyone losing their
278
00:13:52,284 --> 00:13:52,910
jobs.
279
00:13:52,910 --> 00:13:56,996
I mean all these impacts of AI on the economy ultimately will be very critical
280
00:13:56,996 --> 00:14:01,041
for the answer to the question of who are the winners and who are the
281
00:14:01,041 --> 00:14:01,292
losers.
282
00:14:01,292 --> 00:14:05,420
So another way of saying this is many in Silicon Valley talk about technology as
283
00:14:05,420 --> 00:14:06,129
an S-curve.
284
00:14:06,129 --> 00:14:09,549
First we invented the technology and now we're trying to figure out how do we
285
00:14:09,549 --> 00:14:10,466
use this technology?
286
00:14:10,466 --> 00:14:11,634
How do we implement it?
287
00:14:11,634 --> 00:14:15,303
If that implementation phase takes six months, well then we're certainly off to the races
288
00:14:15,303 --> 00:14:18,931
and saying this is how the technology should be used, this is going to pay
289
00:14:18,931 --> 00:14:19,432
off.
290
00:14:19,432 --> 00:14:23,060
But if that implementation phase takes like 2, 3, 4 years, then we have a
291
00:14:23,060 --> 00:14:25,729
risk that the takeoff is only going to come later.
292
00:14:25,729 --> 00:14:29,940
And if there is an S-curve in implementation, maybe to your question, there's also an
293
00:14:29,940 --> 00:14:30,775
S-curve in valuations.
294
00:14:30,775 --> 00:14:31,192
Hmm.
295
00:14:31,192 --> 00:14:34,986
And if that's the case, maybe the companies that invented the technology are not the
296
00:14:34,986 --> 00:14:37,905
ones that ultimately will be winning once the technology is used.
297
00:14:37,905 --> 00:14:42,117
So that's a different way of saying the magnificent seven and hyperscalers did very well
298
00:14:42,117 --> 00:14:46,329
initially, but maybe what we're waiting for now is the s and p 4 93,
299
00:14:46,329 --> 00:14:50,541
meaning the rest of the economy that now needs to take off and show the
300
00:14:50,541 --> 00:14:53,627
returns of AI adoption in healthcare in airlines, in consumer products.
301
00:14:53,627 --> 00:14:57,630
Because if that takeoff takes a longer time, then of course we will not get
302
00:14:57,630 --> 00:15:01,091
the payoff from all the investments that were done to develop the technology.
303
00:15:01,091 --> 00:15:01,508
Yeah, absolutely.
304
00:15:01,508 --> 00:15:05,261
So lemme say there was a very slightly different words, let's think about the invention
305
00:15:05,261 --> 00:15:06,221
of Ozempic.
306
00:15:06,221 --> 00:15:10,432
Ozempic was invented and therefore there was several years where those who invented Ozempic were
307
00:15:10,432 --> 00:15:14,769
able to generate higher profits because this was the only GLP-1 in the market.
308
00:15:14,769 --> 00:15:17,271
Then later on other GLP-1s came to the market.
309
00:15:17,271 --> 00:15:21,859
But that long period generated a situation where you could create some pricing power to
310
00:15:21,859 --> 00:15:23,568
pay back your development cost.
311
00:15:23,568 --> 00:15:27,697
But today the difference between the closed source models and the open source models is
312
00:15:27,697 --> 00:15:29,907
only like three, four months in terms of
313
00:15:29,907 --> 00:15:30,407
capability.
314
00:15:30,407 --> 00:15:32,451
It's like generic drugs versus premium drugs.
315
00:15:32,451 --> 00:15:32,951
Exactly.
316
00:15:32,951 --> 00:15:36,871
That's why there is a risk that we don't have a long period where you
317
00:15:36,871 --> 00:15:40,749
can squeeze out high, high pricing power because they're just right behind you, the open
318
00:15:40,749 --> 00:15:41,333
source models.
319
00:15:41,333 --> 00:15:44,836
And that creates all kinds of corporate finance questions around, normally our textbook would say,
320
00:15:44,836 --> 00:15:48,381
we invent you and I a new product, therefore we have that on the market
321
00:15:48,381 --> 00:15:49,298
for a period.
322
00:15:49,298 --> 00:15:52,843
But that period becomes really, really important for what were the revenue and the profitability
323
00:15:52,843 --> 00:15:56,429
be of our business in particular if that period is at risk of being relatively
324
00:15:56,429 --> 00:15:56,804
short.
325
00:15:56,804 --> 00:15:57,096
Okay.
326
00:15:57,096 --> 00:16:00,683
So I could be completely off on this, but I do want to kind of
327
00:16:00,683 --> 00:16:04,227
give you the victory lap here on the higher for longer thing because when the
328
00:16:04,227 --> 00:16:07,772
feds, uh, summary of economic projections came out and I saw that higher neutral rate
329
00:16:07,772 --> 00:16:11,858
baked in, I just wanted to sign your name at the bottom of it, .
330
00:16:11,858 --> 00:16:16,112
And so what does this all mean at the end when the Federal Reserve unanimously
331
00:16:16,112 --> 00:16:20,366
voted for a higher interest rate in the most recent FLMC, um, they're projecting another
332
00:16:20,366 --> 00:16:23,868
rate hike for the rest of the year at least one.
333
00:16:23,868 --> 00:16:28,289
Um, and they're looking at longer term rates being higher.
334
00:16:28,289 --> 00:16:29,665
What does that amount to in the end?
335
00:16:29,665 --> 00:16:29,957
Yeah,
336
00:16:29,957 --> 00:16:33,168
This is really important because the economy is doing well because of AI spending and
337
00:16:33,168 --> 00:16:36,421
they wanna Big Beautiful Bill, but that's another part of the economy that
338
00:16:36,421 --> 00:16:37,296
is not doing well.
339
00:16:37,296 --> 00:16:40,966
And that's the part that's sensitive to interest rates and that part that's sensitive to
340
00:16:40,966 --> 00:16:43,468
interest rates is housing is not doing well.
341
00:16:43,468 --> 00:16:48,013
Mortgage rates during the pandemic where 2.7 normally mortgage rates are at a level of
342
00:16:48,013 --> 00:16:51,683
around three and four, but today mortgage rates are now seven.
343
00:16:51,683 --> 00:16:55,311
So that means we've seen a very significant increase in mortgage rates that is weighing
344
00:16:55,311 --> 00:16:57,396
dramatically on the housing market at the moment.
345
00:16:57,396 --> 00:17:01,692
Similarly, auto sector and car sales have also been dragged down by interest rates going
346
00:17:01,692 --> 00:17:05,987
up because signing a lease to buy a car, the monthly payment has just gone
347
00:17:05,987 --> 00:17:08,072
up because interest rates have gone up.
348
00:17:08,072 --> 00:17:11,992
So we have a bifurcated economy where the boom is happening in AI spending and
349
00:17:11,992 --> 00:17:14,369
in the one Big Beautiful Bill the fiscal expansion.
350
00:17:14,369 --> 00:17:17,872
But we have another part of the economy that is dragging things down, namely housing
351
00:17:17,872 --> 00:17:18,581
and autos.
352
00:17:18,581 --> 00:17:23,293
The good news from a macroeconomic perspective is that the Fed still is experiencing that
353
00:17:23,293 --> 00:17:28,047
the tailwind coming from AI spending is so strong that they still see more hikes
354
00:17:28,047 --> 00:17:30,757
coming along because the economy still does well.
355
00:17:30,757 --> 00:17:34,510
But that just gets back to the discussion around, well, if AI does not continue
356
00:17:34,510 --> 00:17:38,264
to do well, if there is any type of slowdown, either because the models are
357
00:17:38,264 --> 00:17:42,017
dangerous or because there's risk to the models in terms of security and safety or
358
00:17:42,017 --> 00:17:45,728
if that's the case, that may also mean that the tailwind to the economy from
359
00:17:45,728 --> 00:17:49,481
AI would also be weaker and that could potentially begin to, to ultimately slow the
360
00:17:49,481 --> 00:17:50,148
economy down.
361
00:17:50,148 --> 00:17:53,902
Maybe sometime in our view, sometimes in the middle of 2027.
362
00:17:53,902 --> 00:17:57,905
But at this point where we're sitting here still in 2026, we still have strong
363
00:17:57,905 --> 00:18:01,950
growth from AI coming and that is the reason why the Fed is seeing higher
364
00:18:01,950 --> 00:18:06,037
inflation combined also with higher oil prices and combined with some delayed effects of tariffs.
365
00:18:06,037 --> 00:18:09,081
But all that is arguing for still a strong economy, at least for the next
366
00:18:09,081 --> 00:18:09,915
six to nine months.
367
00:18:09,915 --> 00:18:10,332
Right.
368
00:18:10,332 --> 00:18:15,086
The, the job market seemed to have kept enough room for a rate hike to
369
00:18:15,086 --> 00:18:19,047
happen, but do you ultimately agree that a hiking cycle is warranted?
370
00:18:19,047 --> 00:18:22,217
Well, the risk of course with beginning to raise rates because markets, as you said,
371
00:18:22,217 --> 00:18:25,136
are out pricing a hike this year and then two more hikes next year.
372
00:18:25,136 --> 00:18:25,595
Yeah.
373
00:18:25,595 --> 00:18:29,348
The risk is of course that the interest rate sensitive parts of the economy begins
374
00:18:29,348 --> 00:18:30,974
to slow down even more.
375
00:18:30,974 --> 00:18:33,601
And if that's the case, then of course is it warranted.
376
00:18:33,601 --> 00:18:37,354
Well, you are going to still continue to see a significant drag downwards in housing
377
00:18:37,354 --> 00:18:40,065
and autos because those sectors are sensitive to interest rates.
378
00:18:40,065 --> 00:18:43,484
So that's why it's a very difficult balance that they FMC is striking here.
379
00:18:43,484 --> 00:18:43,943
Right.
380
00:18:43,943 --> 00:18:47,404
The main thing that they are telling us at the last meeting is that, well,
381
00:18:47,404 --> 00:18:50,866
we still have inflation today, which is in round number three and a half at
382
00:18:50,866 --> 00:18:54,327
the peak during the pandemic it was nine, now it's three and a half, but
383
00:18:54,327 --> 00:18:57,788
we still have the last mile of getting inflation down to 2% and that's still
384
00:18:57,788 --> 00:18:59,706
turning out to be a lot harder.
385
00:18:59,706 --> 00:19:03,960
And that's why Kevin Warsh, and therefore FMC voted for unanimously, we need to do
386
00:19:03,960 --> 00:19:08,213
more to get inflation down and importantly to raise interest rates to slow the interest
387
00:19:08,213 --> 00:19:12,467
rates since and parts of the economy down in a very, very gradual and
388
00:19:12,467 --> 00:19:13,509
hopefully controlled way.
389
00:19:13,509 --> 00:19:17,388
And that's why of course the economy is as we look into next year, expected
390
00:19:17,388 --> 00:19:19,473
to begin to moderate in terms of growth.
391
00:19:19,473 --> 00:19:19,973
Right.
392
00:19:19,973 --> 00:19:24,477
I mean it just seems to me like it risks a little bit more of
393
00:19:24,477 --> 00:19:28,981
a K-shaped economy headed into the midterms at a time where their monetary policy as
394
00:19:28,981 --> 00:19:32,108
a blunt instrument can't fix that K shape.
395
00:19:32,108 --> 00:19:36,403
And if you had to fix it, it would really be the fiscal, but now
396
00:19:36,403 --> 00:19:40,740
you have a fiscal where if you added more to the pain you're you're also
397
00:19:40,740 --> 00:19:43,326
adding more pain to the higher cost of interest.
398
00:19:43,367 --> 00:19:45,828
So how do you think about that burden?
399
00:19:45,828 --> 00:19:46,120
Yeah,
400
00:19:46,120 --> 00:19:49,998
The real challenge is the Federal Reserve cannot do anything else, rather other than raise
401
00:19:49,998 --> 00:19:51,583
interest rates or lower interest rates.
402
00:19:51,583 --> 00:19:54,377
And as you say, exactly, that's a very blunt tool compared to a lot of
403
00:19:54,377 --> 00:19:55,878
the other things that are going on.
404
00:19:55,878 --> 00:20:00,382
So that's why fiscal policy is of course, in many cases a way to try
405
00:20:00,382 --> 00:20:04,844
to address some of the challenges, everything from the K-shaped economy, everything from what you
406
00:20:04,844 --> 00:20:08,430
might wanna do on taxes, what you wanna do on expenditures.
407
00:20:08,430 --> 00:20:13,142
But the problem is, as we all know that it's becoming more and more complicated
408
00:20:13,142 --> 00:20:16,562
to see any major national compromise on a fiscal deal.
409
00:20:16,562 --> 00:20:19,272
The challenge is that debt levels are just going to go up.
410
00:20:19,272 --> 00:20:23,359
And as we know at the moment, a lot of the spending by the government
411
00:20:23,359 --> 00:20:24,944
is just really, really significant.
412
00:20:24,944 --> 00:20:30,073
And an important statistic is that roughly two thirds of government spending is discretionary spending
413
00:20:30,073 --> 00:20:31,282
on entitlements.
414
00:20:31,282 --> 00:20:34,285
Entitlements meaning spending on Medicare, Medicaid and social security.
415
00:20:34,285 --> 00:20:39,039
And one very simple problem is that about 75 million people today get social security.
416
00:20:39,039 --> 00:20:42,500
So it's become very, very difficult politically to solve this problem.
417
00:20:42,500 --> 00:20:45,919
And you and we and us, everyone has now to deal with the cards that
418
00:20:45,919 --> 00:20:49,297
we've been dealt and say, well, there's not a great hand we have, but at
419
00:20:49,297 --> 00:20:52,717
the moment, well if interest rates are going to stay higher because the fiscal situation
420
00:20:52,717 --> 00:20:56,136
is going to continue to be bad, I just need to plan accordingly as an
421
00:20:56,136 --> 00:20:56,345
investor.
422
00:20:56,345 --> 00:21:00,932
And that just means cutting coupons in fixed income, benefiting from higher quality credit, still
423
00:21:00,932 --> 00:21:03,684
paying you will continues to be a good strategy.
424
00:21:03,684 --> 00:21:07,479
Because this is ultimately what is the consequence, both of short-term interest rates going up
425
00:21:07,479 --> 00:21:08,980
and long-term interest rates going up.
426
00:21:08,980 --> 00:21:12,316
How much do you think it encourages saving versus investing?
427
00:21:12,316 --> 00:21:15,569
I think this is very important because a lot of money has gone into money
428
00:21:15,569 --> 00:21:18,822
market funds and you and I spoke about this also several years ago, that money
429
00:21:18,822 --> 00:21:21,074
market funds have seen a very significant increase.
430
00:21:21,074 --> 00:21:25,035
Interestingly enough, if you look at the ETF flows, households are very, very eager to
431
00:21:25,035 --> 00:21:29,039
buy short-term government debt, T-bills, meaning money market funds because that has also gone
432
00:21:29,039 --> 00:21:30,039
up in yield.
433
00:21:30,039 --> 00:21:34,668
But households are actually selling over the last several years long-term government debt.
434
00:21:34,668 --> 00:21:38,213
So there are really not many households who are willing to lend money to the
435
00:21:38,213 --> 00:21:40,089
US government for the next 30 years.
436
00:21:40,089 --> 00:21:42,842
So that has meant that interest rates have gone up, but most of this has
437
00:21:42,842 --> 00:21:44,593
gone through the front end of the yield curve.
438
00:21:44,593 --> 00:21:47,137
Does that create a financial stability issue?
439
00:21:47,137 --> 00:21:48,221
I still think that we are.
440
00:21:48,221 --> 00:21:48,555
Okay.
441
00:21:48,555 --> 00:21:53,225
And I still think that because we have an AI boom and we still have
442
00:21:53,225 --> 00:21:57,854
a significant tailwind in growth, that the fiscal issues are still helped by significant tax
443
00:21:57,854 --> 00:22:02,483
payments, of course significant growth that will help ultimately still provide some more tax revenue
444
00:22:02,483 --> 00:22:03,859
on the fiscal side.
445
00:22:03,859 --> 00:22:08,238
But the problem is that if the economy does enter a recession for whatever reason,
446
00:22:08,238 --> 00:22:12,575
then you normally see the public deficit go from at the moment around 5% and
447
00:22:12,575 --> 00:22:14,952
you normally add about 4% points.
448
00:22:14,952 --> 00:22:19,080
So that means that we would get from basically having a 5% deficit for the
449
00:22:19,080 --> 00:22:20,915
government to basically a 10% deficit.
450
00:22:20,915 --> 00:22:24,501
And that's a very dramatic need to raise more money because if you have a
451
00:22:24,501 --> 00:22:28,129
recession, you need to pay more in unemployment benefits and of course you get less
452
00:22:28,129 --> 00:22:28,880
tax revenue.
453
00:22:28,880 --> 00:22:32,091
So therefore the deterioration will come along in the government finances.
454
00:22:32,091 --> 00:22:35,010
So the problem today is that we have a really good economy and we still
455
00:22:35,010 --> 00:22:35,886
have a significant deficit.
456
00:22:35,886 --> 00:22:41,057
So I don't think that will create fiscal and financial instability problems, but I'm still,
457
00:22:41,057 --> 00:22:43,892
of course, we are watching very carefully what the consequences are and especially if we
458
00:22:43,892 --> 00:22:47,896
have an environment where rates are higher for longer because then the debt servicing costs
459
00:22:47,896 --> 00:22:50,606
for the government will also be higher for longer.
460
00:22:50,606 --> 00:22:55,027
So what do you make of the Treasury’s moves in relation to what's happening at
461
00:22:55,027 --> 00:22:56,069
the Federal Reserve?
462
00:22:56,069 --> 00:22:59,155
Yeah, I I, I have seen that like, and you and I also talk about
463
00:22:59,155 --> 00:23:01,991
this, it's a very important move because it is telling you that this
464
00:23:01,991 --> 00:23:04,451
To buy back long bonds absolutely using T-bills,
465
00:23:04,451 --> 00:23:08,538
Trump and the Treasury was trying to say, well, we are worried about long-term interest
466
00:23:08,538 --> 00:23:12,958
rates, so let's instead take from $2 billion in buybacks to now $6 billion buybacks.
467
00:23:12,958 --> 00:23:17,253
And this is really putting a cloud over the rates market and saying there is
468
00:23:17,253 --> 00:23:21,590
a chance we could do something, but 2 billion, 6 billion are very, very small
469
00:23:21,590 --> 00:23:24,259
numbers when total government debt is 40 trillion.
470
00:23:24,259 --> 00:23:27,095
So that's a different way of saying it's not so much the numbers that have
471
00:23:27,095 --> 00:23:27,887
had an impact.
472
00:23:27,887 --> 00:23:31,682
It's really more the threatening of hanging a cloud over rates market is that we
473
00:23:31,682 --> 00:23:35,477
could suddenly come with some headline on your Bloomberg screen that will say, well we
474
00:23:35,477 --> 00:23:36,770
are doing this now.
475
00:23:36,770 --> 00:23:39,856
And if that happens, then you want to be prepared that, well if
476
00:23:39,856 --> 00:23:42,941
that risk certainly will come out of the blue, well maybe then I should be
477
00:23:42,941 --> 00:23:45,777
a little bit more worried about betting on, on rates continues to going up.
478
00:23:45,777 --> 00:23:49,155
But nevertheless, we have still seen interest rates going up and we have now seen
479
00:23:49,155 --> 00:23:50,614
the Fed also raise interest rates.
480
00:23:50,614 --> 00:23:53,867
So the direction of travel for interest rates both in the front end because of
481
00:23:53,867 --> 00:23:57,162
inflation is higher and in the long end also because of the fiscal problems is
482
00:23:57,162 --> 00:23:57,579
also higher.
483
00:23:57,620 --> 00:24:02,374
And there is really, unfortunately not much the Treasury Department can do about that.
484
00:24:02,374 --> 00:24:05,752
You don't need to have a view on this I guess, but do you have
485
00:24:05,752 --> 00:24:08,755
a feeling about where that ceiling is on rates on the long end?
486
00:24:08,755 --> 00:24:12,466
Because I think that that's what's kept a lot of people away this idea that
487
00:24:12,466 --> 00:24:16,177
why would I buy a 10 year at 4.8 or 4.9 if I think it's
488
00:24:16,177 --> 00:24:17,345
it'll go to 5.5.
489
00:24:17,345 --> 00:24:17,845
Exactly.
490
00:24:17,887 --> 00:24:21,640
So that's from an investing perspective, also extremely important because now it becomes important question,
491
00:24:21,640 --> 00:24:25,393
namely when are rates peaking, when do we get to a point where it is
492
00:24:25,393 --> 00:24:29,105
optimal for you and me to take a hundred dollars and put into interest rates
493
00:24:29,105 --> 00:24:30,856
because now they begin to go down.
494
00:24:30,940 --> 00:24:31,273
Right.
495
00:24:31,565 --> 00:24:35,235
I still think at this point, given the AI growth story is so strong that
496
00:24:35,235 --> 00:24:38,946
it's a little bit too early to do that, but it's very clear to watch
497
00:24:38,946 --> 00:24:42,616
the other parts of the economy, housing, autos, other things that are interest rate sensitive
498
00:24:42,616 --> 00:24:46,077
because they continue to be dragged down if interest rates continue to go up.
499
00:24:46,161 --> 00:24:49,538
So if that's the case, it becomes a race where on the one hand high
500
00:24:49,538 --> 00:24:52,916
hand you have both AI doing well, but on the other hand you have other
501
00:24:52,916 --> 00:24:55,585
parts of the economy naming housing and autos doing less well.
502
00:24:55,585 --> 00:24:59,297
So in that race we are watching very carefully not only AI, which has had
503
00:24:59,297 --> 00:25:03,008
a lot of hooray, hooray for a long time, but also watching the parts that
504
00:25:03,008 --> 00:25:04,009
are not doing well.
505
00:25:04,009 --> 00:25:05,760
Namely what are they saying, what are they doing?
506
00:25:05,760 --> 00:25:09,472
And if it gets worse, how big weight should we give to those parts that
507
00:25:09,472 --> 00:25:11,348
are dragging down in my GDP forecast?
508
00:25:11,348 --> 00:25:15,101
And therefore, ultimately the answer to your question is those things that are interest rate
509
00:25:15,101 --> 00:25:18,896
sensitive, will they become so big that it will begin to ultimately become a problem
510
00:25:18,896 --> 00:25:19,939
for the overall economy?
511
00:25:19,939 --> 00:25:23,316
That's not what we're seeing at the moment, but if you have a combination both
512
00:25:23,316 --> 00:25:26,694
of interest rates going up and AI slowing down at the same time, then you
513
00:25:26,694 --> 00:25:30,072
could have both the interest rates since its passed, pulling lower, and also AI beginning
514
00:25:30,072 --> 00:25:31,073
to pull lower.
515
00:25:31,073 --> 00:25:34,659
So what does this mean for one of the most highly talked about areas of
516
00:25:34,659 --> 00:25:36,077
the market this year?
517
00:25:36,077 --> 00:25:40,831
Private credit? On one hand, as a floating rate asset class on balance, it actually
518
00:25:40,831 --> 00:25:43,416
kind of becomes more attractive it feels like.
519
00:25:43,416 --> 00:25:47,628
But on the other hand, you have a whole host of borrowers that will be
520
00:25:47,628 --> 00:25:49,630
beholden to higher debt servicing costs.
521
00:25:49,630 --> 00:25:49,838
Yes.
522
00:25:49,838 --> 00:25:54,009
And there it becomes extremely important therefore to look under the hood and say, what
523
00:25:54,009 --> 00:25:58,179
are different types of credit doing in this environment when we have interest rates higher
524
00:25:58,179 --> 00:25:59,054
for longer?
525
00:25:59,054 --> 00:26:03,266
And the simple answer to that is that businesses in private credit that have been
526
00:26:03,266 --> 00:26:07,478
lent to that are very, very sensitive to interest rates higher for longer is in
527
00:26:07,478 --> 00:26:08,687
particular software.
528
00:26:08,687 --> 00:26:12,482
Software makes up 25% of all direct lending outstanding.
529
00:26:12,482 --> 00:26:17,361
It's an asset class direct lending that is about $2 trillion and roughly $500 billion
530
00:26:17,361 --> 00:26:22,240
of that $2 trillion is lending to software companies in particular lending to application software
531
00:26:22,240 --> 00:26:26,119
companies and less lending to infrastructure and less lending to cybersecurity.
532
00:26:26,119 --> 00:26:29,872
The problem with this is that AI is now a threat in terms of disrupting
533
00:26:29,872 --> 00:26:32,124
the business model for a lot of software companies.
534
00:26:32,124 --> 00:26:35,335
Because if I want to learn a language I would use to download an app,
535
00:26:35,335 --> 00:26:38,546
but now I can learn the language also of course on Claude or ChatGPT.
536
00:26:38,546 --> 00:26:41,632
If I have to take the ACT, the SAT or a test.
537
00:26:41,632 --> 00:26:44,968
Now I can also just use Claude, uh, of course ChatGPT for that.
538
00:26:44,968 --> 00:26:48,971
So that means there's an AI disruption risk, which is threatening, especially applications.
539
00:26:48,971 --> 00:26:53,683
But there's also the higher for longer risk because software companies generally are characterized
540
00:26:53,683 --> 00:26:54,809
by two things.
541
00:26:54,809 --> 00:26:59,104
They generally have a lot of leverage and they generally have very low coverage ratios.
542
00:26:59,104 --> 00:27:04,234
Coverage ratio measures my earnings divided by my debt servicing cost, and the software sector
543
00:27:04,234 --> 00:27:09,405
stands out dramatically from everything else in private credit because the other parts of credit
544
00:27:09,405 --> 00:27:13,241
that are in the private markets is hotels, restaurants, machinery, construction.
545
00:27:13,241 --> 00:27:17,620
They generally have much less leverage and they generally have much higher coverage ratio, much
546
00:27:17,620 --> 00:27:19,496
better ability to service that debt.
547
00:27:19,496 --> 00:27:22,123
So that's a different way of saying inside credit.
548
00:27:22,123 --> 00:27:26,544
When you look under the hood in private credit, different managers have different strategies and
549
00:27:26,544 --> 00:27:30,964
managers that have strategies that are more vulnerable to interest rates higher for longer are
550
00:27:30,964 --> 00:27:32,424
going to have more headwinds.
551
00:27:32,424 --> 00:27:36,761
Because these strategies, especially in software that have high levels of debt and very little
552
00:27:36,761 --> 00:27:41,098
ability to service that debt, they of course will have more problems with servicing that
553
00:27:41,098 --> 00:27:43,391
debt servicing cost when interest rates are high.
554
00:27:43,391 --> 00:27:47,853
Whereas again, for example, hotels, restaurants, they have revenue, they have earning, they have customers,
555
00:27:47,853 --> 00:27:49,897
whereas software is waiting to develop that.
556
00:27:49,897 --> 00:27:51,356
They only get customers down the road.
557
00:27:51,356 --> 00:27:55,026
So that means that those types of loans in private credit are actually going to
558
00:27:55,026 --> 00:27:55,860
do relatively better.
559
00:27:55,860 --> 00:28:00,322
So that's why it is not about private credit as an asset class, it's really
560
00:28:00,322 --> 00:28:03,700
about what did different private credit managers actually lend money to?
561
00:28:03,700 --> 00:28:08,412
Which type of businesses was it and were those businesses sensitive to interest rates higher for longer?
562
00:28:08,412 --> 00:28:10,580
It's been hard for people to understand what's been going on under the hood.
563
00:28:10,580 --> 00:28:15,835
There was this great story by my former colleagues over at Bloomberg, private credit defaults
564
00:28:15,835 --> 00:28:19,338
are 1%, 6% or 19% depending on who you ask.
565
00:28:19,338 --> 00:28:20,005
Yeah.
566
00:28:20,005 --> 00:28:22,340
So what is your view on where they're headed?
567
00:28:22,340 --> 00:28:26,302
This is important because this is exactly a matter of where do I measure?
568
00:28:26,302 --> 00:28:27,761
If I lend to a software company?
569
00:28:27,761 --> 00:28:31,514
Default rates of course have been higher if I lend to hotels and restaurants, default
570
00:28:31,514 --> 00:28:32,682
rates have been lower.
571
00:28:32,682 --> 00:28:36,143
So that's why if you measure in certain parts of the leverage spectrum, in some
572
00:28:36,143 --> 00:28:39,604
parts of the leverage spectrum, there's been more vulnerability in other parts of the leverage
573
00:28:39,604 --> 00:28:41,064
spectrum, there's been lower vulnerability.
574
00:28:41,064 --> 00:28:45,318
So this is a very difficult issue because it becomes a matter of discussion around
575
00:28:45,318 --> 00:28:49,571
what type of industries are we talking about or what areas of the economy are
576
00:28:49,571 --> 00:28:52,991
doing well, what areas of the economy are doing not so well?
577
00:28:52,991 --> 00:28:56,619
And in this case, for the rates higher for longer discussion, what areas of the
578
00:28:56,619 --> 00:29:00,247
economy is it that are more sensitive to interest rates and what areas of the
579
00:29:00,247 --> 00:29:02,665
economy is it that are less sensitive to interest rates?
580
00:29:02,665 --> 00:29:06,544
So put in different words, the vintages that were underwritten in 2021 and 22, for
581
00:29:06,544 --> 00:29:07,836
a lot of private credit.
582
00:29:07,836 --> 00:29:08,253
Mm-hmm .
583
00:29:08,253 --> 00:29:12,465
Those who underwrote vintages with a seven year bullet, meaning a seven year fixed interest
584
00:29:12,465 --> 00:29:16,677
rate, they need to reset now as we get to 2028 and 2029 and those
585
00:29:16,677 --> 00:29:20,889
private credit managers and those parts of credit that did do vintages in those years
586
00:29:20,889 --> 00:29:25,059
under the assumption that interest rates would never go up, those are the areas where
587
00:29:25,059 --> 00:29:26,769
we have seen higher default rates.
588
00:29:26,769 --> 00:29:30,563
Whereas those that were underwritten with their assumption that maybe interest rates could go up,
589
00:29:30,563 --> 00:29:32,357
then they of course are less vulnerable.
590
00:29:32,357 --> 00:29:35,443
And that's why default rates have been lower in those parts of the economy.
591
00:29:35,443 --> 00:29:37,903
Same question applies to private equity that Absolutely.
592
00:29:37,903 --> 00:29:38,570
That backs a lot of this.
593
00:29:38,570 --> 00:29:39,613
Last question for you.
594
00:29:39,613 --> 00:29:43,616
What do higher interest rates mean for private equity, especially when there's still a number
595
00:29:43,616 --> 00:29:46,035
of deals that were written in those vintages?
596
00:29:46,035 --> 00:29:46,327
Yeah,
597
00:29:46,327 --> 00:29:50,038
Because those vintages that were underwritten also in 2020 and 21 and 22, of course
598
00:29:50,038 --> 00:29:53,791
they were likely underwritten with the idea that interest rates are not going to go
599
00:29:53,791 --> 00:29:54,542
up.
600
00:29:54,542 --> 00:29:58,086
And raising the bar for private equity also means that companies that have cash flows
601
00:29:58,086 --> 00:30:01,798
far, far out in the future are now going to be more vulnerable, both in
602
00:30:01,798 --> 00:30:03,841
public markets and also in private markets.
603
00:30:03,841 --> 00:30:08,345
That means in private equity, it really is very essential to ask the question if
604
00:30:08,345 --> 00:30:12,849
the debt servicing costs are high, are the companies that I'm investing in private
605
00:30:12,849 --> 00:30:16,351
equity, are they able to pay the higher debt servicing costs?
606
00:30:16,351 --> 00:30:19,604
In other words, this is all about growth versus value.
607
00:30:19,604 --> 00:30:23,941
Value means that our companies today that you invest in is companies that have higher
608
00:30:23,941 --> 00:30:28,278
ability to pay for the higher debt servicing cost that comes along when interest rates
609
00:30:28,278 --> 00:30:29,446
are higher for longer.
610
00:30:29,487 --> 00:30:33,866
So the short answer to your question is in private equity, what has been outperforming
611
00:30:33,866 --> 00:30:38,245
is high quality companies that have, of course, ability to pay their debt servicing cost.
612
00:30:38,245 --> 00:30:42,415
Whereas companies that do not have much earnings and have cashflows in the future, they
613
00:30:42,415 --> 00:30:43,582
have been struggling more.
614
00:30:43,582 --> 00:30:43,874
Yeah.
615
00:30:43,874 --> 00:30:48,211
It's time to pull out the AI tools with all of those DCF capabilities.
616
00:30:48,211 --> 00:30:49,337
Exactly.
617
00:30:49,337 --> 00:30:50,380
Torsten, thank you for joining.
618
00:30:50,380 --> 00:30:51,547
That is Torsten Slok.
619
00:30:51,547 --> 00:30:56,051
He is the chief economist at Apollo Global Management at a critical inflection point in
620
00:30:56,051 --> 00:30:57,886
the macro and for artificial intelligence.
621
00:30:57,886 --> 00:31:00,471
And you've been watching The Bridge by iCapital.
00:00:00,010 --> 00:00:00,260
Mr.
2
00:00:00,260 --> 00:00:02,470
Higher for longer, for a very long time, .
3
00:00:02,470 --> 00:00:04,222
And he delivered on that promise.
4
00:00:05,181 --> 00:00:08,433
Torsten Slok is one of the most highly watched economists on Wall Street, and he's
5
00:00:08,433 --> 00:00:12,478
been saying for a while that interest rates would be higher for longer.
6
00:00:12,478 --> 00:00:15,898
And now that's certainly the case as the Fed starts a hiking cycle.
7
00:00:15,898 --> 00:00:20,610
So what does that mean for some of the most critical parts of the economy,
8
00:00:20,610 --> 00:00:22,153
particularly artificial intelligence?
9
00:00:22,153 --> 00:00:25,990
This AI thing better work out because otherwise we'll have problems both in the economy and
10
00:00:25,990 --> 00:00:27,991
also in financial markets.
11
00:00:44,213 --> 00:00:46,173
Welcome to the Bridge by iCapital.
12
00:00:46,173 --> 00:00:51,052
I'm Sonali Basak, the Chief Investment Strategist at iCapital, and today I am joined by
13
00:00:51,052 --> 00:00:54,221
Torsten Slok, the chief economist at Apollo Global Management.
14
00:00:54,221 --> 00:00:57,057
He has been focusing a lot on artificial intelligence.
15
00:00:57,099 --> 00:01:01,144
And Torsten, thank you for joining because we're in this moment where you've been talking
16
00:01:01,144 --> 00:01:04,313
about it a lot, this concentration of AI in portfolios.
17
00:01:04,313 --> 00:01:06,023
Here's my question to you.
18
00:01:06,023 --> 00:01:09,567
We are in an era where borrowing costs have been rising.
19
00:01:09,567 --> 00:01:12,904
What does that mean for the AI trade?
20
00:01:12,904 --> 00:01:16,532
Well, the risk when interest rates go up is that any business that has cash
21
00:01:16,532 --> 00:01:20,160
flows far out in the future is going to be more vulnerable because when the
22
00:01:20,160 --> 00:01:23,746
discount rate, meaning interest rate starts to go up, that means that cash flows have
23
00:01:23,746 --> 00:01:26,165
to be discounted in a way where they're worth less.
24
00:01:26,165 --> 00:01:30,335
Because now I can take my money, put it into fixed income and get higher
25
00:01:30,335 --> 00:01:30,919
interest rate.
26
00:01:30,919 --> 00:01:35,130
So that raises the bar for cash flows, and especially for equity that pays off
27
00:01:35,130 --> 00:01:36,840
far out in the future.
28
00:01:36,840 --> 00:01:40,677
And that's very important for AI because a lot of the investments in AI, especially
29
00:01:40,677 --> 00:01:44,513
AI and venture capital, are only cash flows that come far, far out in the
30
00:01:44,513 --> 00:01:44,763
future.
31
00:01:44,763 --> 00:01:49,267
So that's why when interest rates go up, it does increase the bar for all
32
00:01:49,267 --> 00:01:53,479
other investments because suddenly I can get a risk-free interest rate or an interest rate
33
00:01:53,479 --> 00:01:57,691
or yield in public credit, private credit, fixed income, which is higher, and therefore raises
34
00:01:57,691 --> 00:02:00,485
the bar for what I can get in other investments.
35
00:02:00,485 --> 00:02:04,697
So higher interest rates has been making it a bit more difficult for the AI
36
00:02:04,697 --> 00:02:04,988
trade.
37
00:02:04,988 --> 00:02:05,239
Well,
38
00:02:05,239 --> 00:02:09,284
It's so interesting what you're saying it sounds like is just credit is on average
39
00:02:09,284 --> 00:02:12,953
or on balance, rather, a little more attractive than equities in the scenario.
40
00:02:12,953 --> 00:02:16,540
Yeah, and if you think about that from a Federal Reserve perspective, this is actually
41
00:02:16,540 --> 00:02:18,541
the whole idea with raising interest rates.
42
00:02:18,541 --> 00:02:21,752
If someone from the Fed were sitting here next to you and me, they would
43
00:02:21,752 --> 00:02:25,005
say, but the reason why we are raising interest rates is that we want to
44
00:02:25,005 --> 00:02:27,507
slow the economy down to get inflation to slowly come down.
45
00:02:27,507 --> 00:02:27,757
Yeah.
46
00:02:27,757 --> 00:02:31,469
So that's why they raise in small increments of 25 basis points because they want
47
00:02:31,469 --> 00:02:33,220
to do this very, very gradually.
48
00:02:33,220 --> 00:02:37,474
And when you think about that from a 60/40 asset allocation perspective, that means
49
00:02:37,474 --> 00:02:41,727
that the Fed is really saying, please take a little bit money out of public
50
00:02:41,727 --> 00:02:45,981
equities and put it instead into fixed income, higher quality credit, things that pay you
51
00:02:45,981 --> 00:02:50,068
a higher level of yield because that is now more attractive in relative terms.
52
00:02:50,068 --> 00:02:50,443
We'll
53
00:02:50,443 --> 00:02:54,530
Talk more about the Federal Reserve because to your credit, you've been talking about higher
54
00:02:54,530 --> 00:02:58,616
for longer, for a very long time, well before the market started even baking in
55
00:02:58,616 --> 00:03:00,493
interest rate hikes for example.
56
00:03:00,493 --> 00:03:04,997
But on AI, what is at stake if the, if the trade really does slow
57
00:03:04,997 --> 00:03:05,831
down?
58
00:03:05,831 --> 00:03:06,957
That's a lot at stake.
59
00:03:06,957 --> 00:03:10,460
If the AI trade does slow down, first of all, a lot of GDP growth
60
00:03:10,460 --> 00:03:12,336
is driven by AI spending at the moment.
61
00:03:12,419 --> 00:03:16,965
We calculate that GDP growth, which is normally 2% today, roughly 1% is point of
62
00:03:16,965 --> 00:03:17,632
that namely
63
00:03:17,632 --> 00:03:19,509
half of that is because of AI.
64
00:03:19,509 --> 00:03:23,595
And it's not only the building of data centers and energy associated with data centers,
65
00:03:23,595 --> 00:03:24,972
it's also spending on tokens.
66
00:03:24,972 --> 00:03:25,389
Mm-hmm .
67
00:03:25,389 --> 00:03:27,807
It's also spending on AI tools for consumers.
68
00:03:27,807 --> 00:03:30,143
It's also spending on AI tools for businesses.
69
00:03:30,143 --> 00:03:34,271
And it's also the wealth effect that has come along with ai naming home price
70
00:03:34,271 --> 00:03:38,441
in San Francisco have been going up because there's a lot of wealth generated as
71
00:03:38,441 --> 00:03:39,567
a result of ai.
72
00:03:39,567 --> 00:03:43,695
So the first answer to your question is if AI does begin to slow down,
73
00:03:43,695 --> 00:03:45,989
it would also slow down the economy.
74
00:03:45,989 --> 00:03:50,910
And most importantly, really is also that AI is absolutely also now everywhere in people's
75
00:03:50,910 --> 00:03:55,872
asset allocation. The equity market, the 10 biggest stocks, they make up 40% of the
76
00:03:55,872 --> 00:03:56,331
index.
77
00:03:56,331 --> 00:03:59,876
And most of the returns in s and p 500 in the last several years
78
00:03:59,876 --> 00:04:01,252
has been driven by ai.
79
00:04:01,252 --> 00:04:05,630
Similarly, in fixed income, if you look at especially public credit, you also have that
80
00:04:05,630 --> 00:04:07,423
hyperscalers have been issuing more debt.
81
00:04:07,423 --> 00:04:08,758
That is also now more AI.
82
00:04:08,758 --> 00:04:13,929
And finally, if you also own venture capital, venture capital used to be pharma, biotech,
83
00:04:13,929 --> 00:04:17,891
prescription drugs, but today 87% of venture capital is also AI.
84
00:04:17,891 --> 00:04:21,936
So the answer to your question is, if AI does not work out, we will
85
00:04:21,936 --> 00:04:24,062
have some problems for the GDP outlook.
86
00:04:24,062 --> 00:04:26,272
We will also have some problems from an investing perspective.
87
00:04:26,272 --> 00:04:30,359
So the conclusion is this AI thing better work out because otherwise we'll have problems
88
00:04:30,359 --> 00:04:33,237
both in the economy and also in financial markets.
89
00:04:33,237 --> 00:04:34,780
Let's define workout, right?
90
00:04:34,780 --> 00:04:36,781
Because you're saying there's a slowdown.
91
00:04:36,781 --> 00:04:39,158
A slowdown is very different from a crash.
92
00:04:39,158 --> 00:04:42,494
I think a lot of people are very nervous out there about the worst scenarios
93
00:04:42,494 --> 00:04:44,079
that could happen in a slowdown here.
94
00:04:44,121 --> 00:04:47,582
And I think that's the art of the moment that we're in is crafting that
95
00:04:47,582 --> 00:04:49,000
landing.
96
00:04:49,000 --> 00:04:53,545
So when you think about what a slowdown really means, you know, does it mean
97
00:04:53,545 --> 00:04:57,674
that, you know, we might see some companies fall into really meaningful distress.
98
00:04:57,674 --> 00:05:02,719
Does it mean that we'll see even a probability of a recession start to rise?
99
00:05:02,719 --> 00:05:03,804
I mean, what does that actually mean to you?
100
00:05:03,804 --> 00:05:04,221
Yeah, so
101
00:05:04,221 --> 00:05:07,307
There are four different buckets of investing in AI.
102
00:05:07,307 --> 00:05:11,810
One is of course the frontier labs and the models and those of course that
103
00:05:11,810 --> 00:05:15,105
are exposed directly to the consumer and the user of AI.
104
00:05:15,105 --> 00:05:16,856
These are Open AI, Anthropic,
105
00:05:16,856 --> 00:05:17,565
For example.
106
00:05:17,565 --> 00:05:21,235
They have negative profit margins at the moment because they're developing the technology.
107
00:05:21,235 --> 00:05:22,653
And that of course is expensive.
108
00:05:22,653 --> 00:05:27,365
That's why those that are facing the consumer, which is unusual, have negative profit margins.
109
00:05:27,365 --> 00:05:29,909
The next step in the value chain is the hyperscalers.
110
00:05:29,909 --> 00:05:33,954
Those who are building the data centers, they have positive profit margins, but they of
111
00:05:33,954 --> 00:05:38,040
course are also an important part of whether AI working out or not is going
112
00:05:38,040 --> 00:05:39,333
to be impacting them.
113
00:05:39,333 --> 00:05:41,085
The next step is energy and grit.
114
00:05:41,085 --> 00:05:43,628
So that means that this provides the energy to AI.
115
00:05:43,628 --> 00:05:47,382
They are of course also at risk to the upside and downside as a function
116
00:05:47,382 --> 00:05:49,300
of whether AI is succeeding or not.
117
00:05:49,300 --> 00:05:53,345
And finally we have silicon and equipment, meaning those who built the chips and the
118
00:05:53,345 --> 00:05:57,390
semis, and they of course at the ultimate end because they're the, the ones that
119
00:05:57,390 --> 00:06:01,477
all the models are running on, whether it's closed source, open source, of course the
120
00:06:01,477 --> 00:06:04,521
chips are ultimately what is solving all the problems in AI.
121
00:06:04,521 --> 00:06:08,733
And the answer to your question is different parts of that value chain can absolutely
122
00:06:08,733 --> 00:06:12,652
get differently impacted as a result of what AI not working out exactly means.
123
00:06:12,652 --> 00:06:16,656
Let's talk about that a little more because you had mentioned that Frontier Labs typically
124
00:06:16,656 --> 00:06:19,241
have been losing money from a margin perspective, right?
125
00:06:19,241 --> 00:06:22,828
Um, hyperscalers though that that buffer is much thicker.
126
00:06:22,828 --> 00:06:26,873
Does that mean in an environment like this, there's so many people worried about the
127
00:06:26,873 --> 00:06:31,376
free cash flow impacts the declines in free cash flow, but they're relatively insulated actually.
128
00:06:31,376 --> 00:06:33,712
So are you actually really worried about them raising money?
129
00:06:33,712 --> 00:06:34,004
Yeah.
130
00:06:34,004 --> 00:06:34,170
This
131
00:06:34,170 --> 00:06:34,504
In the market?
132
00:06:34,504 --> 00:06:38,382
Yeah, this, this is a really important discussion because at the end of the day
133
00:06:38,372 --> 00:06:41,458
today, the demand for compute is really going up significantly.
134
00:06:41,468 --> 00:06:46,597
Not only because of all of us writing questions into ChatGPT and to Claude,
135
00:06:46,597 --> 00:06:50,142
but also because of agentic AI becoming much more pronounced.
136
00:06:50,142 --> 00:06:53,186
That means that demand for compute is rising very, very significantly.
137
00:06:53,186 --> 00:06:57,315
Meaning people want to run on chips in data centers, but at the same time,
138
00:06:57,315 --> 00:07:00,400
the competitive landscape is now also changing for the models.
139
00:07:00,400 --> 00:07:04,821
Namely open source models are becoming much more used and much more pronounced in the
140
00:07:04,821 --> 00:07:06,572
landscape of those who use AI.
141
00:07:06,572 --> 00:07:11,326
Whereas closed source models, of course, are now getting more competitive pressure from the open
142
00:07:11,326 --> 00:07:12,160
source models.
143
00:07:12,160 --> 00:07:15,496
This raises all kinds of questions about what is the competitive edge for closed source
144
00:07:15,496 --> 00:07:17,331
models for relative to open source models.
145
00:07:17,331 --> 00:07:21,335
But the answer to your question is this race is demand for compute is going
146
00:07:21,335 --> 00:07:25,338
up very, very rapidly, but at the same time, the price of that compute and
147
00:07:25,338 --> 00:07:29,299
the price of tokens and ultimately the competitive landscape for those who use it has
148
00:07:29,299 --> 00:07:31,718
actually been under more downward and more competitive pressure.
149
00:07:31,718 --> 00:07:35,430
And that's the race in ai that is the most critical part of understanding whether
150
00:07:35,430 --> 00:07:37,390
it's working out or not over the next several
151
00:07:37,390 --> 00:07:37,932
Years.
152
00:07:37,932 --> 00:07:41,476
So very simply, when you watch all these hyperscalers, the, you know, um, Amazons of
153
00:07:41,476 --> 00:07:44,229
the world go out and raise debt in markets, right?
154
00:07:44,229 --> 00:07:45,396
Uh, Alphabet, right?
155
00:07:45,396 --> 00:07:47,773
Do you, do you worry about it?
156
00:07:47,773 --> 00:07:51,818
Well, markets are already voting with their feet on that question because CDS spreads on,
157
00:07:51,818 --> 00:07:54,612
uh, several of these names have been widening out.
158
00:07:54,612 --> 00:07:58,407
CDS spreads of course is measuring what is the probability that these companies might not
159
00:07:58,407 --> 00:08:01,701
be able to pay back their debt in the next five years.
160
00:08:01,701 --> 00:08:05,246
And that means that that premium meaning the price of insuring yourself against the risk
161
00:08:05,246 --> 00:08:08,791
that they will not be paying that debt, has been moving a little bit higher,
162
00:08:08,791 --> 00:08:12,377
which some people are beginning to look at, including me and say, well, we need
163
00:08:12,377 --> 00:08:13,920
to understand better why that's happening.
164
00:08:13,920 --> 00:08:18,757
Similarly, we've also seen spreads on hyperscale debt also widen out, which is also saying
165
00:08:18,757 --> 00:08:20,884
that the market has not much.
166
00:08:20,884 --> 00:08:24,554
But started to worry a little bit more about what is the trajectory that we are
167
00:08:24,554 --> 00:08:25,346
on here?
168
00:08:25,346 --> 00:08:26,972
Are the revenues going to increase?
169
00:08:26,972 --> 00:08:31,601
And if you look at the consensus expectations from FactSet today, the operating cash flow
170
00:08:31,601 --> 00:08:34,312
in the five hyperscalers is about $600 billion.
171
00:08:34,312 --> 00:08:38,065
And the consensus is expecting that over the next several years will grow from
172
00:08:38,065 --> 00:08:39,566
$600 billion to about $2 trillion.
173
00:08:39,566 --> 00:08:44,445
In other words, a very dramatic increase in the revenues and the cash flows for
174
00:08:44,445 --> 00:08:45,112
the hyperscalers.
175
00:08:45,112 --> 00:08:48,574
And that is the question, if that revenue comes in faster, comes in slower, is
176
00:08:48,574 --> 00:08:49,783
the answer to the question.
177
00:08:49,783 --> 00:08:52,744
Namely, what is the scenario we have ahead of us for
178
00:08:52,744 --> 00:08:53,119
AI?
179
00:08:53,119 --> 00:08:56,580
And it's interesting because we are talking about this as it pertains to the hyperscalers,
180
00:08:56,580 --> 00:08:58,916
but the hyperscalers we just said have a bigger buffer.
181
00:08:58,916 --> 00:08:59,624
Right?
182
00:08:59,624 --> 00:09:01,376
But what about the frontier labs?
183
00:09:01,376 --> 00:09:05,254
At the end of the day, you saw something happen where you didn't really see
184
00:09:05,254 --> 00:09:09,132
this happen in prior years where pre IPO companies are actually raising fairly meaningful amounts
185
00:09:09,132 --> 00:09:10,217
of debt.
186
00:09:10,217 --> 00:09:11,384
Is that healthy?
187
00:09:11,384 --> 00:09:14,971
Well, one first observation about this, that these companies are private.
188
00:09:14,971 --> 00:09:18,265
So this tells you and me a very important conclusion that you and I have
189
00:09:18,265 --> 00:09:21,559
talked about for many years, namely, that a very significant part of the economy is
190
00:09:21,559 --> 00:09:23,102
outside the S&P 500, right?
191
00:09:23,102 --> 00:09:26,147
And there are a lot of private companies that are wildly successful.
192
00:09:26,147 --> 00:09:30,150
In fact, if you look at businesses in the U.S. that have revenue of more
193
00:09:30,150 --> 00:09:33,611
than a hundred million dollars, 90% of them are actually not public.
194
00:09:33,611 --> 00:09:36,905
So there's a different way of saying there's a lot of private companies including an
195
00:09:36,905 --> 00:09:40,158
Open AI and Anthropic that are very successful at what they do and generate a lot of
196
00:09:40,158 --> 00:09:40,784
revenue.
197
00:09:40,784 --> 00:09:44,161
So in that sense, they are not publishing the same way.
198
00:09:44,161 --> 00:09:47,915
Of course the companies in S&P 500 are doing, but if they are about to
199
00:09:47,915 --> 00:09:51,668
do IPO, of course this will give much more insight, much more light and transparency
200
00:09:51,668 --> 00:09:55,379
in terms of where are they making money and how are they making money?
201
00:09:55,379 --> 00:09:59,257
And that will give the market a better idea about this competitive landscape that we're
202
00:09:59,257 --> 00:10:01,676
seeing between open source models and closed source models.
203
00:10:01,676 --> 00:10:05,137
The reason I ask this also is because in private markets there's the venture capital
204
00:10:05,137 --> 00:10:08,640
aspect, which we've talked about the, you know, the the growth equity ven, uh, private
205
00:10:08,640 --> 00:10:10,684
equity type aspect that we were talking about.
206
00:10:10,684 --> 00:10:14,687
But even data centers to a large degree are tied to the credit ratings of
207
00:10:14,687 --> 00:10:17,689
these firms and the credit health of these firms.
208
00:10:17,689 --> 00:10:20,734
And so there's a whole ecosystem that hinges on this question it feels like.
209
00:10:20,734 --> 00:10:25,404
And and that's why ultimately in particular from a GDP perspective, data center buildout
210
00:10:25,404 --> 00:10:28,907
is absolutely critical for anyone's forecast for GDP growth.
211
00:10:28,907 --> 00:10:33,327
So normally people would say, oh, CapEx and business spending, which data center, the data
212
00:10:33,327 --> 00:10:37,122
center buildout is, is normally a smaller issue in my forecast.
213
00:10:37,122 --> 00:10:41,209
Normally it's all about the consumer, but now your entire forecast depends almost completely on
214
00:10:41,209 --> 00:10:43,961
what is your view on the data center buildout.
215
00:10:43,961 --> 00:10:47,756
And it's very clear, yeah, that the data center buildout depends on whether the
216
00:10:47,756 --> 00:10:51,551
hyperscalers are going to generate all the revenue on the back of all the compute
217
00:10:51,551 --> 00:10:52,552
that's being created.
218
00:10:52,552 --> 00:10:56,847
So yes, there's unlimited demand for compute, but the question is over time whether the
219
00:10:56,847 --> 00:11:00,850
price of that compute is at risk potentially of eventually beginning to go down.
220
00:11:00,850 --> 00:11:04,145
Because if that begins to go down, then you will have two opposite forces.
221
00:11:04,145 --> 00:11:08,023
Yes, we all want to do more at Agentic AI and use AI, but at
222
00:11:08,023 --> 00:11:11,859
the same time if the price of that usage, especially with the competition from open
223
00:11:11,859 --> 00:11:15,738
source models starts to go down, then of course there will be some questions exactly
224
00:11:15,738 --> 00:11:19,366
about what does that then mean for the overall data center build out
225
00:11:19,366 --> 00:11:21,242
Bullish or bearish?
226
00:11:21,242 --> 00:11:24,870
I am still bullish on the economic output because I still think that the most
227
00:11:24,870 --> 00:11:27,497
important impact of AI is also on the labor market.
228
00:11:27,497 --> 00:11:31,667
That yes, there will be some people that likely will lose their jobs, but I
229
00:11:31,667 --> 00:11:35,879
think that a much more important effect on the labor market is that the moment
230
00:11:35,879 --> 00:11:40,049
today in the census data, we are seeing the most businesses created in U.S. economy
231
00:11:40,049 --> 00:11:41,467
ever in U.S. history.
232
00:11:41,467 --> 00:11:46,305
In other words, the business dynamics that have appeared because of AI have just been
233
00:11:46,305 --> 00:11:46,847
incredible.
234
00:11:46,847 --> 00:11:49,474
People are leaving finance to open their own business.
235
00:11:49,474 --> 00:11:54,144
They're leaving consulting, they're leaving media, they're leading or leaving of course legal services and
236
00:11:54,144 --> 00:11:58,565
they go home to their basement and they basically start a solo founder firm.
237
00:11:58,565 --> 00:12:01,192
There are many more solo founder firms now than multi founder.
238
00:12:01,192 --> 00:12:04,695
And that is of course if some of these businesses, it's just a fraction of
239
00:12:04,695 --> 00:12:08,156
the many business that are created at the moment end up creating some success, they
240
00:12:08,156 --> 00:12:09,657
will also create more jobs.
241
00:12:09,657 --> 00:12:13,244
And I think when we put that up on the scale, that will be positive
242
00:12:13,244 --> 00:12:13,911
for employment.
243
00:12:13,911 --> 00:12:14,119
And
244
00:12:14,119 --> 00:12:14,870
You think that sticks?
245
00:12:14,870 --> 00:12:15,370
I truly, even
246
00:12:15,370 --> 00:12:19,457
The last, I truly think AI is a miracle drug that will both create higher
247
00:12:19,457 --> 00:12:20,833
productivity and also higher employment.
248
00:12:20,833 --> 00:12:23,210
And that's by the way also what the data is showing so far.
249
00:12:23,210 --> 00:12:26,671
The employment report, as you and I have talked about, continues to be strong, has
250
00:12:26,671 --> 00:12:30,133
been strong for the last six months so far at this point, there is no
251
00:12:30,133 --> 00:12:33,594
effect of the labor displacement effect, meaning the effect that people are losing their jobs
252
00:12:33,594 --> 00:12:35,012
because of AI being particularly significant.
253
00:12:35,012 --> 00:12:38,181
Okay, one more, one more question on AI before we move to the macro.
254
00:12:38,181 --> 00:12:40,600
We'll put some AI in there too because it's so related.
255
00:12:40,600 --> 00:12:45,229
But a tough question I had gotten this week from a large investor in your
256
00:12:45,229 --> 00:12:47,272
space was what happens?
257
00:12:47,272 --> 00:12:51,442
You have multiple companies now, three of them at least, that have surged past a
258
00:12:51,442 --> 00:12:55,612
trillion dollars in value in a timeframe we have never seen before reaching almost
259
00:12:55,612 --> 00:12:58,781
$2 trillion in some of those cases.
260
00:12:58,781 --> 00:13:02,243
What happens if one of those companies doesn't exist in the next five years?
261
00:13:02,243 --> 00:13:02,535
Yeah,
262
00:13:02,535 --> 00:13:03,911
This is really, really important.
263
00:13:03,911 --> 00:13:08,164
So that's also why the famous discussion at the moment is about the moat.
264
00:13:08,164 --> 00:13:12,626
What is really the moat that different businesses, especially in the tech world have?
265
00:13:12,626 --> 00:13:16,838
Because if the tech companies, the broadly speaking, the magnificent seven and the hyperscalers, if
266
00:13:16,838 --> 00:13:21,008
they do not have a moat where they can protect themselves, then of course they
267
00:13:21,008 --> 00:13:24,553
will ultimately begin to shrink and they will be more at risk.
268
00:13:24,553 --> 00:13:28,264
So the question of course to ask will all the hyperscalers survive?
269
00:13:28,264 --> 00:13:32,893
Are some of them at risk of going down because they might not have the
270
00:13:32,893 --> 00:13:37,564
right moat or are we seeing all of them surviving and growing into even much
271
00:13:37,564 --> 00:13:38,356
bigger companies?
272
00:13:38,356 --> 00:13:42,526
And that becomes hugely a function of this question around is it open source models
273
00:13:42,526 --> 00:13:43,694
that are winning?
274
00:13:43,694 --> 00:13:45,487
Is it closed source models that are winning?
275
00:13:45,487 --> 00:13:47,322
What is regulation going to look like?
276
00:13:47,322 --> 00:13:48,615
Are we going to see regulation?
277
00:13:48,615 --> 00:13:52,284
There's a lot of talk at the moment about human extinction, maybe everyone losing their
278
00:13:52,284 --> 00:13:52,910
jobs.
279
00:13:52,910 --> 00:13:56,996
I mean all these impacts of AI on the economy ultimately will be very critical
280
00:13:56,996 --> 00:14:01,041
for the answer to the question of who are the winners and who are the
281
00:14:01,041 --> 00:14:01,292
losers.
282
00:14:01,292 --> 00:14:05,420
So another way of saying this is many in Silicon Valley talk about technology as
283
00:14:05,420 --> 00:14:06,129
an S-curve.
284
00:14:06,129 --> 00:14:09,549
First we invented the technology and now we're trying to figure out how do we
285
00:14:09,549 --> 00:14:10,466
use this technology?
286
00:14:10,466 --> 00:14:11,634
How do we implement it?
287
00:14:11,634 --> 00:14:15,303
If that implementation phase takes six months, well then we're certainly off to the races
288
00:14:15,303 --> 00:14:18,931
and saying this is how the technology should be used, this is going to pay
289
00:14:18,931 --> 00:14:19,432
off.
290
00:14:19,432 --> 00:14:23,060
But if that implementation phase takes like 2, 3, 4 years, then we have a
291
00:14:23,060 --> 00:14:25,729
risk that the takeoff is only going to come later.
292
00:14:25,729 --> 00:14:29,940
And if there is an S-curve in implementation, maybe to your question, there's also an
293
00:14:29,940 --> 00:14:30,775
S-curve in valuations.
294
00:14:30,775 --> 00:14:31,192
Hmm.
295
00:14:31,192 --> 00:14:34,986
And if that's the case, maybe the companies that invented the technology are not the
296
00:14:34,986 --> 00:14:37,905
ones that ultimately will be winning once the technology is used.
297
00:14:37,905 --> 00:14:42,117
So that's a different way of saying the magnificent seven and hyperscalers did very well
298
00:14:42,117 --> 00:14:46,329
initially, but maybe what we're waiting for now is the s and p 4 93,
299
00:14:46,329 --> 00:14:50,541
meaning the rest of the economy that now needs to take off and show the
300
00:14:50,541 --> 00:14:53,627
returns of AI adoption in healthcare in airlines, in consumer products.
301
00:14:53,627 --> 00:14:57,630
Because if that takeoff takes a longer time, then of course we will not get
302
00:14:57,630 --> 00:15:01,091
the payoff from all the investments that were done to develop the technology.
303
00:15:01,091 --> 00:15:01,508
Yeah, absolutely.
304
00:15:01,508 --> 00:15:05,261
So lemme say there was a very slightly different words, let's think about the invention
305
00:15:05,261 --> 00:15:06,221
of Ozempic.
306
00:15:06,221 --> 00:15:10,432
Ozempic was invented and therefore there was several years where those who invented Ozempic were
307
00:15:10,432 --> 00:15:14,769
able to generate higher profits because this was the only GLP-1 in the market.
308
00:15:14,769 --> 00:15:17,271
Then later on other GLP-1s came to the market.
309
00:15:17,271 --> 00:15:21,859
But that long period generated a situation where you could create some pricing power to
310
00:15:21,859 --> 00:15:23,568
pay back your development cost.
311
00:15:23,568 --> 00:15:27,697
But today the difference between the closed source models and the open source models is
312
00:15:27,697 --> 00:15:29,907
only like three, four months in terms of
313
00:15:29,907 --> 00:15:30,407
capability.
314
00:15:30,407 --> 00:15:32,451
It's like generic drugs versus premium drugs.
315
00:15:32,451 --> 00:15:32,951
Exactly.
316
00:15:32,951 --> 00:15:36,871
That's why there is a risk that we don't have a long period where you
317
00:15:36,871 --> 00:15:40,749
can squeeze out high, high pricing power because they're just right behind you, the open
318
00:15:40,749 --> 00:15:41,333
source models.
319
00:15:41,333 --> 00:15:44,836
And that creates all kinds of corporate finance questions around, normally our textbook would say,
320
00:15:44,836 --> 00:15:48,381
we invent you and I a new product, therefore we have that on the market
321
00:15:48,381 --> 00:15:49,298
for a period.
322
00:15:49,298 --> 00:15:52,843
But that period becomes really, really important for what were the revenue and the profitability
323
00:15:52,843 --> 00:15:56,429
be of our business in particular if that period is at risk of being relatively
324
00:15:56,429 --> 00:15:56,804
short.
325
00:15:56,804 --> 00:15:57,096
Okay.
326
00:15:57,096 --> 00:16:00,683
So I could be completely off on this, but I do want to kind of
327
00:16:00,683 --> 00:16:04,227
give you the victory lap here on the higher for longer thing because when the
328
00:16:04,227 --> 00:16:07,772
feds, uh, summary of economic projections came out and I saw that higher neutral rate
329
00:16:07,772 --> 00:16:11,858
baked in, I just wanted to sign your name at the bottom of it, .
330
00:16:11,858 --> 00:16:16,112
And so what does this all mean at the end when the Federal Reserve unanimously
331
00:16:16,112 --> 00:16:20,366
voted for a higher interest rate in the most recent FLMC, um, they're projecting another
332
00:16:20,366 --> 00:16:23,868
rate hike for the rest of the year at least one.
333
00:16:23,868 --> 00:16:28,289
Um, and they're looking at longer term rates being higher.
334
00:16:28,289 --> 00:16:29,665
What does that amount to in the end?
335
00:16:29,665 --> 00:16:29,957
Yeah,
336
00:16:29,957 --> 00:16:33,168
This is really important because the economy is doing well because of AI spending and
337
00:16:33,168 --> 00:16:36,421
they wanna Big Beautiful Bill, but that's another part of the economy that
338
00:16:36,421 --> 00:16:37,296
is not doing well.
339
00:16:37,296 --> 00:16:40,966
And that's the part that's sensitive to interest rates and that part that's sensitive to
340
00:16:40,966 --> 00:16:43,468
interest rates is housing is not doing well.
341
00:16:43,468 --> 00:16:48,013
Mortgage rates during the pandemic where 2.7 normally mortgage rates are at a level of
342
00:16:48,013 --> 00:16:51,683
around three and four, but today mortgage rates are now seven.
343
00:16:51,683 --> 00:16:55,311
So that means we've seen a very significant increase in mortgage rates that is weighing
344
00:16:55,311 --> 00:16:57,396
dramatically on the housing market at the moment.
345
00:16:57,396 --> 00:17:01,692
Similarly, auto sector and car sales have also been dragged down by interest rates going
346
00:17:01,692 --> 00:17:05,987
up because signing a lease to buy a car, the monthly payment has just gone
347
00:17:05,987 --> 00:17:08,072
up because interest rates have gone up.
348
00:17:08,072 --> 00:17:11,992
So we have a bifurcated economy where the boom is happening in AI spending and
349
00:17:11,992 --> 00:17:14,369
in the one Big Beautiful Bill the fiscal expansion.
350
00:17:14,369 --> 00:17:17,872
But we have another part of the economy that is dragging things down, namely housing
351
00:17:17,872 --> 00:17:18,581
and autos.
352
00:17:18,581 --> 00:17:23,293
The good news from a macroeconomic perspective is that the Fed still is experiencing that
353
00:17:23,293 --> 00:17:28,047
the tailwind coming from AI spending is so strong that they still see more hikes
354
00:17:28,047 --> 00:17:30,757
coming along because the economy still does well.
355
00:17:30,757 --> 00:17:34,510
But that just gets back to the discussion around, well, if AI does not continue
356
00:17:34,510 --> 00:17:38,264
to do well, if there is any type of slowdown, either because the models are
357
00:17:38,264 --> 00:17:42,017
dangerous or because there's risk to the models in terms of security and safety or
358
00:17:42,017 --> 00:17:45,728
if that's the case, that may also mean that the tailwind to the economy from
359
00:17:45,728 --> 00:17:49,481
AI would also be weaker and that could potentially begin to, to ultimately slow the
360
00:17:49,481 --> 00:17:50,148
economy down.
361
00:17:50,148 --> 00:17:53,902
Maybe sometime in our view, sometimes in the middle of 2027.
362
00:17:53,902 --> 00:17:57,905
But at this point where we're sitting here still in 2026, we still have strong
363
00:17:57,905 --> 00:18:01,950
growth from AI coming and that is the reason why the Fed is seeing higher
364
00:18:01,950 --> 00:18:06,037
inflation combined also with higher oil prices and combined with some delayed effects of tariffs.
365
00:18:06,037 --> 00:18:09,081
But all that is arguing for still a strong economy, at least for the next
366
00:18:09,081 --> 00:18:09,915
six to nine months.
367
00:18:09,915 --> 00:18:10,332
Right.
368
00:18:10,332 --> 00:18:15,086
The, the job market seemed to have kept enough room for a rate hike to
369
00:18:15,086 --> 00:18:19,047
happen, but do you ultimately agree that a hiking cycle is warranted?
370
00:18:19,047 --> 00:18:22,217
Well, the risk of course with beginning to raise rates because markets, as you said,
371
00:18:22,217 --> 00:18:25,136
are out pricing a hike this year and then two more hikes next year.
372
00:18:25,136 --> 00:18:25,595
Yeah.
373
00:18:25,595 --> 00:18:29,348
The risk is of course that the interest rate sensitive parts of the economy begins
374
00:18:29,348 --> 00:18:30,974
to slow down even more.
375
00:18:30,974 --> 00:18:33,601
And if that's the case, then of course is it warranted.
376
00:18:33,601 --> 00:18:37,354
Well, you are going to still continue to see a significant drag downwards in housing
377
00:18:37,354 --> 00:18:40,065
and autos because those sectors are sensitive to interest rates.
378
00:18:40,065 --> 00:18:43,484
So that's why it's a very difficult balance that they FMC is striking here.
379
00:18:43,484 --> 00:18:43,943
Right.
380
00:18:43,943 --> 00:18:47,404
The main thing that they are telling us at the last meeting is that, well,
381
00:18:47,404 --> 00:18:50,866
we still have inflation today, which is in round number three and a half at
382
00:18:50,866 --> 00:18:54,327
the peak during the pandemic it was nine, now it's three and a half, but
383
00:18:54,327 --> 00:18:57,788
we still have the last mile of getting inflation down to 2% and that's still
384
00:18:57,788 --> 00:18:59,706
turning out to be a lot harder.
385
00:18:59,706 --> 00:19:03,960
And that's why Kevin Warsh, and therefore FMC voted for unanimously, we need to do
386
00:19:03,960 --> 00:19:08,213
more to get inflation down and importantly to raise interest rates to slow the interest
387
00:19:08,213 --> 00:19:12,467
rates since and parts of the economy down in a very, very gradual and
388
00:19:12,467 --> 00:19:13,509
hopefully controlled way.
389
00:19:13,509 --> 00:19:17,388
And that's why of course the economy is as we look into next year, expected
390
00:19:17,388 --> 00:19:19,473
to begin to moderate in terms of growth.
391
00:19:19,473 --> 00:19:19,973
Right.
392
00:19:19,973 --> 00:19:24,477
I mean it just seems to me like it risks a little bit more of
393
00:19:24,477 --> 00:19:28,981
a K-shaped economy headed into the midterms at a time where their monetary policy as
394
00:19:28,981 --> 00:19:32,108
a blunt instrument can't fix that K shape.
395
00:19:32,108 --> 00:19:36,403
And if you had to fix it, it would really be the fiscal, but now
396
00:19:36,403 --> 00:19:40,740
you have a fiscal where if you added more to the pain you're you're also
397
00:19:40,740 --> 00:19:43,326
adding more pain to the higher cost of interest.
398
00:19:43,367 --> 00:19:45,828
So how do you think about that burden?
399
00:19:45,828 --> 00:19:46,120
Yeah,
400
00:19:46,120 --> 00:19:49,998
The real challenge is the Federal Reserve cannot do anything else, rather other than raise
401
00:19:49,998 --> 00:19:51,583
interest rates or lower interest rates.
402
00:19:51,583 --> 00:19:54,377
And as you say, exactly, that's a very blunt tool compared to a lot of
403
00:19:54,377 --> 00:19:55,878
the other things that are going on.
404
00:19:55,878 --> 00:20:00,382
So that's why fiscal policy is of course, in many cases a way to try
405
00:20:00,382 --> 00:20:04,844
to address some of the challenges, everything from the K-shaped economy, everything from what you
406
00:20:04,844 --> 00:20:08,430
might wanna do on taxes, what you wanna do on expenditures.
407
00:20:08,430 --> 00:20:13,142
But the problem is, as we all know that it's becoming more and more complicated
408
00:20:13,142 --> 00:20:16,562
to see any major national compromise on a fiscal deal.
409
00:20:16,562 --> 00:20:19,272
The challenge is that debt levels are just going to go up.
410
00:20:19,272 --> 00:20:23,359
And as we know at the moment, a lot of the spending by the government
411
00:20:23,359 --> 00:20:24,944
is just really, really significant.
412
00:20:24,944 --> 00:20:30,073
And an important statistic is that roughly two thirds of government spending is discretionary spending
413
00:20:30,073 --> 00:20:31,282
on entitlements.
414
00:20:31,282 --> 00:20:34,285
Entitlements meaning spending on Medicare, Medicaid and social security.
415
00:20:34,285 --> 00:20:39,039
And one very simple problem is that about 75 million people today get social security.
416
00:20:39,039 --> 00:20:42,500
So it's become very, very difficult politically to solve this problem.
417
00:20:42,500 --> 00:20:45,919
And you and we and us, everyone has now to deal with the cards that
418
00:20:45,919 --> 00:20:49,297
we've been dealt and say, well, there's not a great hand we have, but at
419
00:20:49,297 --> 00:20:52,717
the moment, well if interest rates are going to stay higher because the fiscal situation
420
00:20:52,717 --> 00:20:56,136
is going to continue to be bad, I just need to plan accordingly as an
421
00:20:56,136 --> 00:20:56,345
investor.
422
00:20:56,345 --> 00:21:00,932
And that just means cutting coupons in fixed income, benefiting from higher quality credit, still
423
00:21:00,932 --> 00:21:03,684
paying you will continues to be a good strategy.
424
00:21:03,684 --> 00:21:07,479
Because this is ultimately what is the consequence, both of short-term interest rates going up
425
00:21:07,479 --> 00:21:08,980
and long-term interest rates going up.
426
00:21:08,980 --> 00:21:12,316
How much do you think it encourages saving versus investing?
427
00:21:12,316 --> 00:21:15,569
I think this is very important because a lot of money has gone into money
428
00:21:15,569 --> 00:21:18,822
market funds and you and I spoke about this also several years ago, that money
429
00:21:18,822 --> 00:21:21,074
market funds have seen a very significant increase.
430
00:21:21,074 --> 00:21:25,035
Interestingly enough, if you look at the ETF flows, households are very, very eager to
431
00:21:25,035 --> 00:21:29,039
buy short-term government debt, T-bills, meaning money market funds because that has also gone
432
00:21:29,039 --> 00:21:30,039
up in yield.
433
00:21:30,039 --> 00:21:34,668
But households are actually selling over the last several years long-term government debt.
434
00:21:34,668 --> 00:21:38,213
So there are really not many households who are willing to lend money to the
435
00:21:38,213 --> 00:21:40,089
US government for the next 30 years.
436
00:21:40,089 --> 00:21:42,842
So that has meant that interest rates have gone up, but most of this has
437
00:21:42,842 --> 00:21:44,593
gone through the front end of the yield curve.
438
00:21:44,593 --> 00:21:47,137
Does that create a financial stability issue?
439
00:21:47,137 --> 00:21:48,221
I still think that we are.
440
00:21:48,221 --> 00:21:48,555
Okay.
441
00:21:48,555 --> 00:21:53,225
And I still think that because we have an AI boom and we still have
442
00:21:53,225 --> 00:21:57,854
a significant tailwind in growth, that the fiscal issues are still helped by significant tax
443
00:21:57,854 --> 00:22:02,483
payments, of course significant growth that will help ultimately still provide some more tax revenue
444
00:22:02,483 --> 00:22:03,859
on the fiscal side.
445
00:22:03,859 --> 00:22:08,238
But the problem is that if the economy does enter a recession for whatever reason,
446
00:22:08,238 --> 00:22:12,575
then you normally see the public deficit go from at the moment around 5% and
447
00:22:12,575 --> 00:22:14,952
you normally add about 4% points.
448
00:22:14,952 --> 00:22:19,080
So that means that we would get from basically having a 5% deficit for the
449
00:22:19,080 --> 00:22:20,915
government to basically a 10% deficit.
450
00:22:20,915 --> 00:22:24,501
And that's a very dramatic need to raise more money because if you have a
451
00:22:24,501 --> 00:22:28,129
recession, you need to pay more in unemployment benefits and of course you get less
452
00:22:28,129 --> 00:22:28,880
tax revenue.
453
00:22:28,880 --> 00:22:32,091
So therefore the deterioration will come along in the government finances.
454
00:22:32,091 --> 00:22:35,010
So the problem today is that we have a really good economy and we still
455
00:22:35,010 --> 00:22:35,886
have a significant deficit.
456
00:22:35,886 --> 00:22:41,057
So I don't think that will create fiscal and financial instability problems, but I'm still,
457
00:22:41,057 --> 00:22:43,892
of course, we are watching very carefully what the consequences are and especially if we
458
00:22:43,892 --> 00:22:47,896
have an environment where rates are higher for longer because then the debt servicing costs
459
00:22:47,896 --> 00:22:50,606
for the government will also be higher for longer.
460
00:22:50,606 --> 00:22:55,027
So what do you make of the Treasury’s moves in relation to what's happening at
461
00:22:55,027 --> 00:22:56,069
the Federal Reserve?
462
00:22:56,069 --> 00:22:59,155
Yeah, I I, I have seen that like, and you and I also talk about
463
00:22:59,155 --> 00:23:01,991
this, it's a very important move because it is telling you that this
464
00:23:01,991 --> 00:23:04,451
To buy back long bonds absolutely using T-bills,
465
00:23:04,451 --> 00:23:08,538
Trump and the Treasury was trying to say, well, we are worried about long-term interest
466
00:23:08,538 --> 00:23:12,958
rates, so let's instead take from $2 billion in buybacks to now $6 billion buybacks.
467
00:23:12,958 --> 00:23:17,253
And this is really putting a cloud over the rates market and saying there is
468
00:23:17,253 --> 00:23:21,590
a chance we could do something, but 2 billion, 6 billion are very, very small
469
00:23:21,590 --> 00:23:24,259
numbers when total government debt is 40 trillion.
470
00:23:24,259 --> 00:23:27,095
So that's a different way of saying it's not so much the numbers that have
471
00:23:27,095 --> 00:23:27,887
had an impact.
472
00:23:27,887 --> 00:23:31,682
It's really more the threatening of hanging a cloud over rates market is that we
473
00:23:31,682 --> 00:23:35,477
could suddenly come with some headline on your Bloomberg screen that will say, well we
474
00:23:35,477 --> 00:23:36,770
are doing this now.
475
00:23:36,770 --> 00:23:39,856
And if that happens, then you want to be prepared that, well if
476
00:23:39,856 --> 00:23:42,941
that risk certainly will come out of the blue, well maybe then I should be
477
00:23:42,941 --> 00:23:45,777
a little bit more worried about betting on, on rates continues to going up.
478
00:23:45,777 --> 00:23:49,155
But nevertheless, we have still seen interest rates going up and we have now seen
479
00:23:49,155 --> 00:23:50,614
the Fed also raise interest rates.
480
00:23:50,614 --> 00:23:53,867
So the direction of travel for interest rates both in the front end because of
481
00:23:53,867 --> 00:23:57,162
inflation is higher and in the long end also because of the fiscal problems is
482
00:23:57,162 --> 00:23:57,579
also higher.
483
00:23:57,620 --> 00:24:02,374
And there is really, unfortunately not much the Treasury Department can do about that.
484
00:24:02,374 --> 00:24:05,752
You don't need to have a view on this I guess, but do you have
485
00:24:05,752 --> 00:24:08,755
a feeling about where that ceiling is on rates on the long end?
486
00:24:08,755 --> 00:24:12,466
Because I think that that's what's kept a lot of people away this idea that
487
00:24:12,466 --> 00:24:16,177
why would I buy a 10 year at 4.8 or 4.9 if I think it's
488
00:24:16,177 --> 00:24:17,345
it'll go to 5.5.
489
00:24:17,345 --> 00:24:17,845
Exactly.
490
00:24:17,887 --> 00:24:21,640
So that's from an investing perspective, also extremely important because now it becomes important question,
491
00:24:21,640 --> 00:24:25,393
namely when are rates peaking, when do we get to a point where it is
492
00:24:25,393 --> 00:24:29,105
optimal for you and me to take a hundred dollars and put into interest rates
493
00:24:29,105 --> 00:24:30,856
because now they begin to go down.
494
00:24:30,940 --> 00:24:31,273
Right.
495
00:24:31,565 --> 00:24:35,235
I still think at this point, given the AI growth story is so strong that
496
00:24:35,235 --> 00:24:38,946
it's a little bit too early to do that, but it's very clear to watch
497
00:24:38,946 --> 00:24:42,616
the other parts of the economy, housing, autos, other things that are interest rate sensitive
498
00:24:42,616 --> 00:24:46,077
because they continue to be dragged down if interest rates continue to go up.
499
00:24:46,161 --> 00:24:49,538
So if that's the case, it becomes a race where on the one hand high
500
00:24:49,538 --> 00:24:52,916
hand you have both AI doing well, but on the other hand you have other
501
00:24:52,916 --> 00:24:55,585
parts of the economy naming housing and autos doing less well.
502
00:24:55,585 --> 00:24:59,297
So in that race we are watching very carefully not only AI, which has had
503
00:24:59,297 --> 00:25:03,008
a lot of hooray, hooray for a long time, but also watching the parts that
504
00:25:03,008 --> 00:25:04,009
are not doing well.
505
00:25:04,009 --> 00:25:05,760
Namely what are they saying, what are they doing?
506
00:25:05,760 --> 00:25:09,472
And if it gets worse, how big weight should we give to those parts that
507
00:25:09,472 --> 00:25:11,348
are dragging down in my GDP forecast?
508
00:25:11,348 --> 00:25:15,101
And therefore, ultimately the answer to your question is those things that are interest rate
509
00:25:15,101 --> 00:25:18,896
sensitive, will they become so big that it will begin to ultimately become a problem
510
00:25:18,896 --> 00:25:19,939
for the overall economy?
511
00:25:19,939 --> 00:25:23,316
That's not what we're seeing at the moment, but if you have a combination both
512
00:25:23,316 --> 00:25:26,694
of interest rates going up and AI slowing down at the same time, then you
513
00:25:26,694 --> 00:25:30,072
could have both the interest rates since its passed, pulling lower, and also AI beginning
514
00:25:30,072 --> 00:25:31,073
to pull lower.
515
00:25:31,073 --> 00:25:34,659
So what does this mean for one of the most highly talked about areas of
516
00:25:34,659 --> 00:25:36,077
the market this year?
517
00:25:36,077 --> 00:25:40,831
Private credit? On one hand, as a floating rate asset class on balance, it actually
518
00:25:40,831 --> 00:25:43,416
kind of becomes more attractive it feels like.
519
00:25:43,416 --> 00:25:47,628
But on the other hand, you have a whole host of borrowers that will be
520
00:25:47,628 --> 00:25:49,630
beholden to higher debt servicing costs.
521
00:25:49,630 --> 00:25:49,838
Yes.
522
00:25:49,838 --> 00:25:54,009
And there it becomes extremely important therefore to look under the hood and say, what
523
00:25:54,009 --> 00:25:58,179
are different types of credit doing in this environment when we have interest rates higher
524
00:25:58,179 --> 00:25:59,054
for longer?
525
00:25:59,054 --> 00:26:03,266
And the simple answer to that is that businesses in private credit that have been
526
00:26:03,266 --> 00:26:07,478
lent to that are very, very sensitive to interest rates higher for longer is in
527
00:26:07,478 --> 00:26:08,687
particular software.
528
00:26:08,687 --> 00:26:12,482
Software makes up 25% of all direct lending outstanding.
529
00:26:12,482 --> 00:26:17,361
It's an asset class direct lending that is about $2 trillion and roughly $500 billion
530
00:26:17,361 --> 00:26:22,240
of that $2 trillion is lending to software companies in particular lending to application software
531
00:26:22,240 --> 00:26:26,119
companies and less lending to infrastructure and less lending to cybersecurity.
532
00:26:26,119 --> 00:26:29,872
The problem with this is that AI is now a threat in terms of disrupting
533
00:26:29,872 --> 00:26:32,124
the business model for a lot of software companies.
534
00:26:32,124 --> 00:26:35,335
Because if I want to learn a language I would use to download an app,
535
00:26:35,335 --> 00:26:38,546
but now I can learn the language also of course on Claude or ChatGPT.
536
00:26:38,546 --> 00:26:41,632
If I have to take the ACT, the SAT or a test.
537
00:26:41,632 --> 00:26:44,968
Now I can also just use Claude, uh, of course ChatGPT for that.
538
00:26:44,968 --> 00:26:48,971
So that means there's an AI disruption risk, which is threatening, especially applications.
539
00:26:48,971 --> 00:26:53,683
But there's also the higher for longer risk because software companies generally are characterized
540
00:26:53,683 --> 00:26:54,809
by two things.
541
00:26:54,809 --> 00:26:59,104
They generally have a lot of leverage and they generally have very low coverage ratios.
542
00:26:59,104 --> 00:27:04,234
Coverage ratio measures my earnings divided by my debt servicing cost, and the software sector
543
00:27:04,234 --> 00:27:09,405
stands out dramatically from everything else in private credit because the other parts of credit
544
00:27:09,405 --> 00:27:13,241
that are in the private markets is hotels, restaurants, machinery, construction.
545
00:27:13,241 --> 00:27:17,620
They generally have much less leverage and they generally have much higher coverage ratio, much
546
00:27:17,620 --> 00:27:19,496
better ability to service that debt.
547
00:27:19,496 --> 00:27:22,123
So that's a different way of saying inside credit.
548
00:27:22,123 --> 00:27:26,544
When you look under the hood in private credit, different managers have different strategies and
549
00:27:26,544 --> 00:27:30,964
managers that have strategies that are more vulnerable to interest rates higher for longer are
550
00:27:30,964 --> 00:27:32,424
going to have more headwinds.
551
00:27:32,424 --> 00:27:36,761
Because these strategies, especially in software that have high levels of debt and very little
552
00:27:36,761 --> 00:27:41,098
ability to service that debt, they of course will have more problems with servicing that
553
00:27:41,098 --> 00:27:43,391
debt servicing cost when interest rates are high.
554
00:27:43,391 --> 00:27:47,853
Whereas again, for example, hotels, restaurants, they have revenue, they have earning, they have customers,
555
00:27:47,853 --> 00:27:49,897
whereas software is waiting to develop that.
556
00:27:49,897 --> 00:27:51,356
They only get customers down the road.
557
00:27:51,356 --> 00:27:55,026
So that means that those types of loans in private credit are actually going to
558
00:27:55,026 --> 00:27:55,860
do relatively better.
559
00:27:55,860 --> 00:28:00,322
So that's why it is not about private credit as an asset class, it's really
560
00:28:00,322 --> 00:28:03,700
about what did different private credit managers actually lend money to?
561
00:28:03,700 --> 00:28:08,412
Which type of businesses was it and were those businesses sensitive to interest rates higher for longer?
562
00:28:08,412 --> 00:28:10,580
It's been hard for people to understand what's been going on under the hood.
563
00:28:10,580 --> 00:28:15,835
There was this great story by my former colleagues over at Bloomberg, private credit defaults
564
00:28:15,835 --> 00:28:19,338
are 1%, 6% or 19% depending on who you ask.
565
00:28:19,338 --> 00:28:20,005
Yeah.
566
00:28:20,005 --> 00:28:22,340
So what is your view on where they're headed?
567
00:28:22,340 --> 00:28:26,302
This is important because this is exactly a matter of where do I measure?
568
00:28:26,302 --> 00:28:27,761
If I lend to a software company?
569
00:28:27,761 --> 00:28:31,514
Default rates of course have been higher if I lend to hotels and restaurants, default
570
00:28:31,514 --> 00:28:32,682
rates have been lower.
571
00:28:32,682 --> 00:28:36,143
So that's why if you measure in certain parts of the leverage spectrum, in some
572
00:28:36,143 --> 00:28:39,604
parts of the leverage spectrum, there's been more vulnerability in other parts of the leverage
573
00:28:39,604 --> 00:28:41,064
spectrum, there's been lower vulnerability.
574
00:28:41,064 --> 00:28:45,318
So this is a very difficult issue because it becomes a matter of discussion around
575
00:28:45,318 --> 00:28:49,571
what type of industries are we talking about or what areas of the economy are
576
00:28:49,571 --> 00:28:52,991
doing well, what areas of the economy are doing not so well?
577
00:28:52,991 --> 00:28:56,619
And in this case, for the rates higher for longer discussion, what areas of the
578
00:28:56,619 --> 00:29:00,247
economy is it that are more sensitive to interest rates and what areas of the
579
00:29:00,247 --> 00:29:02,665
economy is it that are less sensitive to interest rates?
580
00:29:02,665 --> 00:29:06,544
So put in different words, the vintages that were underwritten in 2021 and 22, for
581
00:29:06,544 --> 00:29:07,836
a lot of private credit.
582
00:29:07,836 --> 00:29:08,253
Mm-hmm .
583
00:29:08,253 --> 00:29:12,465
Those who underwrote vintages with a seven year bullet, meaning a seven year fixed interest
584
00:29:12,465 --> 00:29:16,677
rate, they need to reset now as we get to 2028 and 2029 and those
585
00:29:16,677 --> 00:29:20,889
private credit managers and those parts of credit that did do vintages in those years
586
00:29:20,889 --> 00:29:25,059
under the assumption that interest rates would never go up, those are the areas where
587
00:29:25,059 --> 00:29:26,769
we have seen higher default rates.
588
00:29:26,769 --> 00:29:30,563
Whereas those that were underwritten with their assumption that maybe interest rates could go up,
589
00:29:30,563 --> 00:29:32,357
then they of course are less vulnerable.
590
00:29:32,357 --> 00:29:35,443
And that's why default rates have been lower in those parts of the economy.
591
00:29:35,443 --> 00:29:37,903
Same question applies to private equity that Absolutely.
592
00:29:37,903 --> 00:29:38,570
That backs a lot of this.
593
00:29:38,570 --> 00:29:39,613
Last question for you.
594
00:29:39,613 --> 00:29:43,616
What do higher interest rates mean for private equity, especially when there's still a number
595
00:29:43,616 --> 00:29:46,035
of deals that were written in those vintages?
596
00:29:46,035 --> 00:29:46,327
Yeah,
597
00:29:46,327 --> 00:29:50,038
Because those vintages that were underwritten also in 2020 and 21 and 22, of course
598
00:29:50,038 --> 00:29:53,791
they were likely underwritten with the idea that interest rates are not going to go
599
00:29:53,791 --> 00:29:54,542
up.
600
00:29:54,542 --> 00:29:58,086
And raising the bar for private equity also means that companies that have cash flows
601
00:29:58,086 --> 00:30:01,798
far, far out in the future are now going to be more vulnerable, both in
602
00:30:01,798 --> 00:30:03,841
public markets and also in private markets.
603
00:30:03,841 --> 00:30:08,345
That means in private equity, it really is very essential to ask the question if
604
00:30:08,345 --> 00:30:12,849
the debt servicing costs are high, are the companies that I'm investing in private
605
00:30:12,849 --> 00:30:16,351
equity, are they able to pay the higher debt servicing costs?
606
00:30:16,351 --> 00:30:19,604
In other words, this is all about growth versus value.
607
00:30:19,604 --> 00:30:23,941
Value means that our companies today that you invest in is companies that have higher
608
00:30:23,941 --> 00:30:28,278
ability to pay for the higher debt servicing cost that comes along when interest rates
609
00:30:28,278 --> 00:30:29,446
are higher for longer.
610
00:30:29,487 --> 00:30:33,866
So the short answer to your question is in private equity, what has been outperforming
611
00:30:33,866 --> 00:30:38,245
is high quality companies that have, of course, ability to pay their debt servicing cost.
612
00:30:38,245 --> 00:30:42,415
Whereas companies that do not have much earnings and have cashflows in the future, they
613
00:30:42,415 --> 00:30:43,582
have been struggling more.
614
00:30:43,582 --> 00:30:43,874
Yeah.
615
00:30:43,874 --> 00:30:48,211
It's time to pull out the AI tools with all of those DCF capabilities.
616
00:30:48,211 --> 00:30:49,337
Exactly.
617
00:30:49,337 --> 00:30:50,380
Torsten, thank you for joining.
618
00:30:50,380 --> 00:30:51,547
That is Torsten Slok.
619
00:30:51,547 --> 00:30:56,051
He is the chief economist at Apollo Global Management at a critical inflection point in
620
00:30:56,051 --> 00:30:57,886
the macro and for artificial intelligence.
621
00:30:57,886 --> 00:31:00,471
And you've been watching The Bridge by iCapital.
