WEBVTT
1
00:00:11.180 --> 00:00:14.460
Good morning. It is great to be back at Moscone.
2
00:00:15.840 --> 00:00:20.460
You can almost feel the tectonic plates shifting underneath us.
3
00:00:20.520 --> 00:00:21.353
Yesterday,
4
00:00:21.700 --> 00:00:26.480
Will walked you through the 288 new products that we're launching to help you as
5
00:00:26.500 --> 00:00:30.880
the economy replatforms around AI. But today,
6
00:00:31.500 --> 00:00:32.980
I want to take you to my happy place,
7
00:00:33.280 --> 00:00:37.900
getting deep into the economic data and I have a few big trends that we're
8
00:00:37.920 --> 00:00:41.780
seeing that I want to talk you through. But first,
9
00:00:42.620 --> 00:00:46.500
let's take stock of what's happened since we were together here at Sessions last
10
00:00:46.540 --> 00:00:50.880
year. So since we last got together,
11
00:00:52.040 --> 00:00:54.540
it's been a little wobbly. In January,
12
00:00:54.820 --> 00:00:59.640
the software sector lost a trillion dollars of market value in under 30 days.
13
00:01:00.860 --> 00:01:03.800
You're all familiar with the worry. AI makes software more abundant,
14
00:01:03.860 --> 00:01:07.840
more substitutable, potentially less sticky than prior models had assumed.
15
00:01:09.180 --> 00:01:12.060
But interestingly, this is a perspective worry.
16
00:01:12.180 --> 00:01:17.180
It's not revenue softness that's happening right now.
SaaS payment volumes
17
00:01:17.520 --> 00:01:21.300
and Stripe, they're actually a good chunk higher today than before the sell-off.
18
00:01:21.360 --> 00:01:23.700
So SaaS is still growing just fine.
19
00:01:26.280 --> 00:01:29.360
What does seem to have come back into fashion though is profitability and not
20
00:01:29.380 --> 00:01:32.680
just in software, but across the entire equity market.
21
00:01:33.260 --> 00:01:38.100
So this is a chart showing how much market cap has historically accrued
22
00:01:38.460 --> 00:01:41.300
to the most profitable and the least profitable companies in the index.
23
00:01:41.640 --> 00:01:42.560
So you see in the middle here,
24
00:01:42.660 --> 00:01:46.300
this is the dot-com boom where the markets went a bit mad for a moment,
25
00:01:46.400 --> 00:01:46.960
and briefly,
26
00:01:46.960 --> 00:01:50.260
the least profitable companies were worth more than the most profitable ones.
27
00:01:51.060 --> 00:01:55.440
But if you go forward to today, now, despite all the talk about bubbles,
28
00:01:55.520 --> 00:01:58.520
it's the opposite. The markets are being tediously rational.
29
00:01:58.580 --> 00:02:02.360
The most profitable companies are the ones being rewarded with outsize
30
00:02:02.660 --> 00:02:03.493
valuations.
31
00:02:06.500 --> 00:02:09.100
We heard a lot about trade policy over the past year. Believe it or not,
32
00:02:09.280 --> 00:02:13.100
it's only been a year since Liberation Day. It feels like eight years,
33
00:02:13.560 --> 00:02:14.460
but one year.
34
00:02:15.500 --> 00:02:19.540
And the common wisdom by now is that the tariffs were kind of economically
35
00:02:19.580 --> 00:02:22.340
speaking "a nothing burger," the dog that didn't bark.
36
00:02:23.460 --> 00:02:24.660
And you can make that case.
37
00:02:25.180 --> 00:02:28.700
Trade flows ended the year only a little bit below 2024 levels,
38
00:02:30.000 --> 00:02:33.620
but there were higher prices to absorb as a result of tariffs.
39
00:02:34.620 --> 00:02:38.920
And what happened is early on, you had businesses eating the higher costs,
40
00:02:39.760 --> 00:02:42.880
but now quarter by quarter, they're starting to pass them through.
41
00:02:43.280 --> 00:02:45.720
So we can throw up durable goods prices here.
42
00:02:46.260 --> 00:02:49.760
And this is how they behave intrayear in normal years. You might be wondering,
43
00:02:49.880 --> 00:02:53.560
"Why is there this downward trend?" It's because you have deflation in durable
44
00:02:53.600 --> 00:02:56.000
goods. Manufacturing productivity generally increases,
45
00:02:56.340 --> 00:02:57.520
global trade adds competition,
46
00:02:57.580 --> 00:03:01.080
and your flat-screen TV gets cheaper or your washing machine gets better for the
47
00:03:01.120 --> 00:03:04.200
same price.
So you usually, in most years, get this gentle,
48
00:03:04.260 --> 00:03:08.840
predictable slope downward. And you see this only goes up to '24. If we add '25,
49
00:03:10.020 --> 00:03:13.640
totally different. Bucks the trend. And if we add '26,
50
00:03:14.520 --> 00:03:18.240
it's only three months in, but same thing is happening and even more so.
51
00:03:19.600 --> 00:03:23.700
So our view is that tariff costs are still working their way through to the
52
00:03:23.740 --> 00:03:25.400
consumer. The story is not written yet.
53
00:03:27.880 --> 00:03:31.120
Another phenomenon you might've heard about is this K-shaped economy.
54
00:03:31.440 --> 00:03:34.100
The idea that wealthy consumers are holding things up.
55
00:03:34.180 --> 00:03:36.460
They're an increasingly large share of the spending.
56
00:03:37.000 --> 00:03:39.320
So maybe you saw this go viral on Twitter.
57
00:03:39.380 --> 00:03:42.720
This is the new United 787-9 seat map.
58
00:03:42.820 --> 00:03:47.800
And you can see that most of the cabin is now dedicated to business class seats.
59
00:03:49.420 --> 00:03:51.560
Or you can look at Delta's ticket revenue.
60
00:03:51.920 --> 00:03:56.180
So they reported that premium seat revenue is now bigger than economy for the
61
00:03:56.200 --> 00:03:57.033
first time.
62
00:03:59.240 --> 00:04:01.800
This stuff gets a lot of retweets,
63
00:04:02.020 --> 00:04:05.260
but we don't actually see it in Stripe's data.
64
00:04:05.900 --> 00:04:10.360
So if we look at the ratio of high-income to
65
00:04:10.460 --> 00:04:14.740
low-income consumer spending in our data, in a K-shaped economy,
66
00:04:15.100 --> 00:04:17.460
you'd expect this chart to be going up, but it's the opposite.
67
00:04:17.520 --> 00:04:18.980
The line is trending gradually downwards.
68
00:04:19.080 --> 00:04:23.380
The gap between high-income and low-income spending has been shrinking.
69
00:04:23.380 --> 00:04:25.340
So we don't actually see, when we look for it,
70
00:04:25.600 --> 00:04:29.600
this overreliance of the economy on high-end consumers that everyone has been
71
00:04:29.620 --> 00:04:30.453
talking about.
72
00:04:32.840 --> 00:04:37.040
The other major economic topic is the worry that AI is taking all the jobs.
73
00:04:38.140 --> 00:04:41.580
And if you look at the numbers, it's true the labor market is cooling slightly.
74
00:04:42.120 --> 00:04:46.140
US unemployment rates are ticking up about a percentage point over the past
75
00:04:46.180 --> 00:04:50.600
three years. But you can ask the question: how much of that is actually AI?
76
00:04:51.680 --> 00:04:54.960
Part of it is the delayed pandemic hangover. If you look at the hiring rate,
77
00:04:55.760 --> 00:04:58.080
companies really binged on hiring in '21 and '22,
78
00:04:58.200 --> 00:04:59.680
and they're still unwinding that,
79
00:05:00.260 --> 00:05:03.280
and tighter immigration played a bit of a role. And then of course,
80
00:05:03.480 --> 00:05:04.313
you have interest rates.
81
00:05:04.840 --> 00:05:09.720
So the Fed's latest forecast for the end of 2026 is half a point higher than
82
00:05:09.780 --> 00:05:12.040
they predicted at this time two years ago.
83
00:05:12.420 --> 00:05:15.280
So money is tighter and tight money slows hiring.
84
00:05:16.720 --> 00:05:20.400
So AI is in the mix. It's presumably going to have a big future impact,
85
00:05:21.300 --> 00:05:26.140
but it's one force amongst many and probably not the dominant one in the labor
86
00:05:26.180 --> 00:05:29.160
market numbers you've seen to date.
Now,
87
00:05:29.700 --> 00:05:33.140
you're probably wondering what we're seeing at Stripe and how it corresponds to
88
00:05:33.220 --> 00:05:34.780
what's going on in the wider world.
89
00:05:35.840 --> 00:05:38.340
Stripe now processes almost 2% of global GDP,
90
00:05:38.740 --> 00:05:42.020
so we've a really useful window into the forefront of the entire economy.
91
00:05:43.180 --> 00:05:46.440
There's three areas that I want to dive into deeper with you.
92
00:05:48.920 --> 00:05:53.300
The first is that we're seeing a structural increase in economic dynamism.
93
00:05:54.380 --> 00:05:57.600
More firms are getting started. They're keeping head count lean for longer,
94
00:05:58.240 --> 00:06:01.240
but they're scaling revenue faster than we've ever seen before.
95
00:06:02.280 --> 00:06:06.020
To explore this more, I'd like to call on some help from an actual economist,
96
00:06:06.560 --> 00:06:08.540
Stripe's head of data and AI, Emily Sands.
97
00:06:18.720 --> 00:06:19.553
Thanks, John.
98
00:06:22.700 --> 00:06:26.380
There's plenty of talk about AI not showing up in the macrodata yet,
99
00:06:26.580 --> 00:06:28.100
like the labor stats we just saw.
100
00:06:29.040 --> 00:06:31.500
If you're as AI-pilled as most of us in this room,
101
00:06:31.820 --> 00:06:35.860
that probably creates some cognitive dissonance. But here's the thing:
102
00:06:36.640 --> 00:06:40.520
AI actually is showing up in the macrodata. You just have to know where to look.
103
00:06:41.620 --> 00:06:44.400
Here's US business formations over the past 20 years.
104
00:06:45.160 --> 00:06:48.320
You can see a huge bump during the pandemic, not surprising,
105
00:06:48.920 --> 00:06:53.200
but you can also see it reaccelerating now. Basically,
106
00:06:53.400 --> 00:06:57.780
all of the recent growth is coming from what the Census Bureau calls nonemployer
107
00:06:57.860 --> 00:07:02.140
firms, or what we all call solopreneurs. That's the blue line here.
108
00:07:03.260 --> 00:07:06.100
Now, these didn't used to be considered serious businesses,
109
00:07:06.780 --> 00:07:10.420
but AI means more of these businesses are getting to real scale.
110
00:07:11.580 --> 00:07:16.140
This is the number of solopreneurs in the US doing over $100,000 in revenue.
111
00:07:17.060 --> 00:07:17.860
Yes,
112
00:07:17.860 --> 00:07:22.620
solopreneurism is now how close to five million Americans earn their living.
113
00:07:23.980 --> 00:07:28.100
And we're fortunate to have many thriving solopreneurs building with Stripe.
114
00:07:36.700 --> 00:07:40.080
It's a great crew. This isn't just an American story.
115
00:07:40.460 --> 00:07:43.380
New business registrations are up 40% in Australia,
116
00:07:43.820 --> 00:07:47.240
up 70% in Finland, up 80% in France.
117
00:07:47.500 --> 00:07:52.420
So the surge in dynamism is happening across advanced economies. At Stripe,
118
00:07:52.880 --> 00:07:53.820
we're seeing this firsthand.
119
00:07:54.580 --> 00:07:57.780
Yeah. Stripe Atlas is the simplest way to get incorporated in the US.
120
00:07:58.280 --> 00:08:01.920
And last week, we celebrated our 100,000th Atlas-founded business.
121
00:08:06.220 --> 00:08:09.600
So somewhat topically for this talk, that business is Amperical,
122
00:08:09.760 --> 00:08:14.000
which is building AI software that optimizes profitability for battery energy
123
00:08:14.240 --> 00:08:16.440
storage systems. Sounds like something we're going to need.
124
00:08:16.860 --> 00:08:20.200
And I actually think Rachana, the founder, is here today. But the point is,
125
00:08:20.640 --> 00:08:25.020
we hit 100,000 Atlas incorporations way earlier than we expected.
126
00:08:25.080 --> 00:08:29.780
The number has exploded. There's more of them and they're scaling up like crazy.
127
00:08:30.100 --> 00:08:31.900
They really are. So in aggregate,
128
00:08:32.340 --> 00:08:37.320
Atlas companies incorporated in 2025 are raking in twice as much revenue as
129
00:08:37.340 --> 00:08:41.540
the class of '24 was by this point. Nothing shabby about a doubling,
130
00:08:42.440 --> 00:08:45.760
but it's the class of '26 that's really cooking.
131
00:08:45.880 --> 00:08:49.560
So let's zoom in a little bit and you can see that just a few months into the
132
00:08:49.620 --> 00:08:50.453
year,
133
00:08:51.240 --> 00:08:54.990
the class of '26 is tracking to five times the revenue of last year's
134
00:08:55.960 --> 00:09:00.080
cohort. These are very unusual growth rates,
135
00:09:00.240 --> 00:09:04.000
and they're driven in part by a generational shift in how companies grow
136
00:09:04.140 --> 00:09:08.260
internationally. So one simple way to see this is to ask,
137
00:09:09.100 --> 00:09:09.440
"Across Stripe,
138
00:09:09.440 --> 00:09:13.080
how many companies are earning most of their revenue outside their home
139
00:09:13.120 --> 00:09:16.800
country?" Five years ago, that was 11.6%.
140
00:09:18.120 --> 00:09:20.040
Since then, it's doubled.
141
00:09:20.920 --> 00:09:23.160
And they aren't just selling in the obvious places.
142
00:09:23.520 --> 00:09:26.000
Among companies making most of their money cross-border,
143
00:09:26.280 --> 00:09:31.240
around a quarter are making most outside the top 10 global markets.
144
00:09:31.240 --> 00:09:32.840
It's a real long tail story.
145
00:09:34.480 --> 00:09:36.400
The international pecking order has also flipped.
146
00:09:36.680 --> 00:09:40.640
It used to be that young companies sold mostly at home and only ventured out to
147
00:09:40.720 --> 00:09:44.000
the open sea of global commerce once they've gotten big. Now,
148
00:09:44.360 --> 00:09:48.400
it's the newborns on Stripe who are globetrotting their way through the long
149
00:09:48.400 --> 00:09:53.260
tail of markets. Take the top 100 AI startups on Stripe.
150
00:09:53.620 --> 00:09:58.600
The median earns most of its revenue internationally and sells into 55 countries
151
00:09:58.800 --> 00:10:00.440
within its first year of existence.
152
00:10:01.360 --> 00:10:04.160
Emergent Labs was founded in the US in 2024,
153
00:10:04.800 --> 00:10:08.720
but already nearly 70% of their revenue comes from abroad.
154
00:10:08.780 --> 00:10:10.140
They're doing material business in many markets.
155
00:10:11.360 --> 00:10:15.160
No less than 16 countries drive at least 1% of Emergence revenue.
156
00:10:15.560 --> 00:10:19.680
So this is a totally new startup playbook. You launch globally on day one,
157
00:10:19.880 --> 00:10:23.360
you keep head count very lean, and you automate aggressively.
158
00:10:24.520 --> 00:10:29.440
But not every business can be a solopreneur or a top AI startup.
So
159
00:10:29.960 --> 00:10:32.280
what does all of this mean for the median firm?
160
00:10:33.320 --> 00:10:38.160
Despite the fact that one of us has a PhD in economics and the other dropped out
161
00:10:38.280 --> 00:10:43.000
of college and now hosts a podcast in a pub,
162
00:10:45.390 --> 00:10:48.640
John is going to teach us a little bit about economic history.
163
00:10:57.880 --> 00:11:00.980
I think you might be surprised to learn just how many economic theories have
164
00:11:01.000 --> 00:11:04.760
been devised over a pint, Dr. Sands. OK.
165
00:11:05.020 --> 00:11:08.400
Some economic history. Are we ready? In 1931,
166
00:11:08.840 --> 00:11:13.260
a 21-year-old British student took a road trip across Depression era America.
167
00:11:13.820 --> 00:11:18.480
And back then, as now, a bunch of young people-pretty into communism.
168
00:11:19.200 --> 00:11:22.860
Central planning had this real intellectual prestige at the time.
169
00:11:22.920 --> 00:11:26.520
A lot of smart people thought that top-down coordination might be the way to do
170
00:11:26.600 --> 00:11:29.500
things much better than messy, chaotic markets.
171
00:11:30.440 --> 00:11:33.640
And the student noticed something interesting. Within a firm,
172
00:11:34.180 --> 00:11:36.140
resources aren't allocated by prices.
173
00:11:36.200 --> 00:11:38.220
There's no little market inside the company.
174
00:11:38.720 --> 00:11:43.380
Companies themselves are centrally planned. And so he wondered,
175
00:11:44.020 --> 00:11:45.460
"If markets are so efficient,
176
00:11:46.040 --> 00:11:49.880
why do we keep building little islands of central planning inside markets?"
177
00:11:51.120 --> 00:11:52.520
His name was Ronald Coase,
178
00:11:53.480 --> 00:11:56.980
and his answer was that firms exist because coordinating inside a company is
179
00:11:57.020 --> 00:12:01.420
often actually cheaper and easier than coordinating through markets.
180
00:12:01.420 --> 00:12:03.680
And that answer, won him the Nobel Prize...
181
00:12:04.380 --> 00:12:07.620
60 years later. They don't rush into decisions over there in Stockholm.
182
00:12:08.300 --> 00:12:09.420
I know what you're thinking.
183
00:12:09.760 --> 00:12:14.360
John's little economic story is heavy on story and light on
184
00:12:14.540 --> 00:12:17.180
economics, but here's where we're going.
185
00:12:18.000 --> 00:12:21.840
What's the Coasean reading of AI? In the near term,
186
00:12:22.120 --> 00:12:24.240
the within firm effect is the most obvious.
187
00:12:24.840 --> 00:12:28.500
Companies have shared context and systems of record and aligned incentives,
188
00:12:28.740 --> 00:12:32.220
and all of that allows for easier coordination with AI.
189
00:12:32.640 --> 00:12:35.140
You're probably feeling this in all your own companies already.
190
00:12:36.320 --> 00:12:40.960
But in the medium term, external markets are likely to get even more efficient.
191
00:12:41.900 --> 00:12:43.880
Agents are already great at discovery.
192
00:12:44.580 --> 00:12:47.620
They make it trivially easy to integrate a new piece of software.
193
00:12:48.160 --> 00:12:51.420
They make contracting much more straightforward. And,
194
00:12:51.620 --> 00:12:52.460
as we'll discuss in a minute,
195
00:12:52.520 --> 00:12:57.080
they can transact much more frequently and for smaller amounts than humans can.
196
00:12:58.020 --> 00:12:58.853
Together,
197
00:12:58.920 --> 00:13:03.500
all of this should bring coordination costs between firms down by quite a bit.
198
00:13:04.140 --> 00:13:06.800
Now, some of these changes will take a little while to play out,
199
00:13:07.380 --> 00:13:11.840
but some of them are already here.
So on net, with AI,
200
00:13:12.240 --> 00:13:16.100
we expect fewer people per firm, more output per firm,
201
00:13:16.780 --> 00:13:18.160
just more firms,
202
00:13:18.740 --> 00:13:22.340
and more coordination happening through market-like mechanisms.
203
00:13:23.080 --> 00:13:24.400
Coase would have liked that idea.
204
00:13:24.760 --> 00:13:27.700
He always thought firms were an inefficient setup.
205
00:13:27.760 --> 00:13:30.500
The more we can do via markets, the better.
206
00:13:31.440 --> 00:13:34.300
An indie hacker from another era. Yeah. Thanks, Emily.
207
00:13:41.700 --> 00:13:46.660
OK. So that was the explosion in business dynamism that we're seeing.
208
00:13:48.500 --> 00:13:52.800
The second trend I want to talk about is how commerce itself is becoming
209
00:13:53.140 --> 00:13:55.500
agentic. At Stripe,
210
00:13:55.880 --> 00:14:00.060
we think about this in increasing levels of autonomy in the purchasing flow from
211
00:14:00.180 --> 00:14:04.300
simple help all the way up to significant agency. But you might be wondering,
212
00:14:04.580 --> 00:14:07.500
"Where have we actually gotten to in 2026?" Well,
213
00:14:08.180 --> 00:14:09.640
level one is already here.
214
00:14:09.720 --> 00:14:12.540
This is software schlepping through forms on your behalf.
215
00:14:13.340 --> 00:14:16.680
The in-app checkout experience we're doing with Meta is actually a good example
216
00:14:16.760 --> 00:14:21.220
of this. So maybe you find something, maybe in an ad, you express interest,
217
00:14:21.540 --> 00:14:23.580
and the agent, it has your details already,
218
00:14:23.940 --> 00:14:28.380
and it can complete the checkout for you. Super handy, really convenient,
219
00:14:28.440 --> 00:14:30.600
but not exactly science fiction. I mean,
220
00:14:30.660 --> 00:14:33.720
you mightn't even think of this as agentic, even though strictly speaking,
221
00:14:33.800 --> 00:14:36.940
the software is doing the purchasing for you. It is your agent.
222
00:14:38.600 --> 00:14:42.600
Level two is the shift from plain old keyword search that we've had for decades
223
00:14:43.020 --> 00:14:47.120
to a shopping assistant that can actually reason within constraints and find
224
00:14:47.140 --> 00:14:52.000
products accordingly. Think about when you do some shopping on ChatGPT.
225
00:14:52.160 --> 00:14:54.330
Imagine totally hypothetically you say,
226
00:14:54.930 --> 00:14:59.650
"I need a birthday gift for my brother." He's 38 and has kind of weirdly at this
227
00:14:59.730 --> 00:15:01.750
late age gotten very into calisthenics,
228
00:15:02.120 --> 00:15:05.130
but he already actually has a lot of calisthenics gear.
229
00:15:05.190 --> 00:15:08.680
And so what's a nonobvious gift that I could get for him for under $100,
230
00:15:10.030 --> 00:15:10.863
hypothetically.
231
00:15:12.450 --> 00:15:16.770
It's not obvious what keywords you would put in to get these results.
232
00:15:18.230 --> 00:15:22.650
Wayfair does something similar. You can describe a room or a style or a feeling,
233
00:15:22.970 --> 00:15:24.950
and the agent rummages through the catalog for you.
234
00:15:25.950 --> 00:15:28.370
And you're still in control here. You're making the buying decision.
235
00:15:28.430 --> 00:15:31.950
You don't have some awful vase you didn't want showing up unexpected,
236
00:15:32.690 --> 00:15:35.270
but you've better ways to find what you're actually looking for.
237
00:15:36.850 --> 00:15:38.670
Most people say they already shop this way. I mean,
238
00:15:38.760 --> 00:15:40.670
you've probably all already done this.
239
00:15:42.430 --> 00:15:45.230
But what does it take to get to levels three, four, and five,
240
00:15:45.630 --> 00:15:49.470
where agents are making and executing purchasing decisions with real autonomy?
241
00:15:51.330 --> 00:15:54.750
Well, one way to peer into the future is OpenClaw, and there,
242
00:15:54.870 --> 00:15:57.070
the demand for autonomous commerce is really palpable.
243
00:15:57.450 --> 00:16:00.910
This is the cumulative downloads of payment-related skills on ClawHub,
244
00:16:01.550 --> 00:16:03.210
125,000 in 12 weeks.
245
00:16:04.010 --> 00:16:06.890
And this is despite the fact that OpenClaw is still pretty hard to use for
246
00:16:06.950 --> 00:16:11.010
regular folks. So the question isn't whether there's demand; there is.
247
00:16:12.030 --> 00:16:15.330
It's how do we get what's already live at the frontier to go mainstream?
248
00:16:18.330 --> 00:16:22.030
Will talked yesterday about how we need the economic infrastructure for AI.
249
00:16:22.270 --> 00:16:23.310
Agents need to be able to pay,
250
00:16:23.410 --> 00:16:25.590
businesses need to be able to accept payments from those agents,
251
00:16:25.610 --> 00:16:27.230
and the whole thing needs a trust layer,
252
00:16:28.190 --> 00:16:31.010
and Stripe is working hard and getting all of this deployed.
253
00:16:32.410 --> 00:16:35.950
There's actually one corner of commerce where things are already moving quickly,
254
00:16:36.010 --> 00:16:39.890
which is software buying from software. Let me show you what I mean.
255
00:16:43.510 --> 00:16:46.950
So previously,
256
00:16:47.630 --> 00:16:48.950
yesterday you saw buying stuff, but here,
257
00:16:49.010 --> 00:16:53.110
just imagine we have an agent that we want to help us do some research.
258
00:16:53.170 --> 00:16:57.710
So I have a question I've been wondering. Hey, Claude,
259
00:16:57.910 --> 00:17:02.310
how is AI demand affecting commodity prices and supply and demand for different
260
00:17:02.410 --> 00:17:03.243
energy sources?
261
00:17:07.770 --> 00:17:10.890
You guys are probably still typing to your Claude, but you can just talk to it.
262
00:17:11.950 --> 00:17:15.190
So what's going on here?
263
00:17:15.530 --> 00:17:20.430
You've probably heard so much about how the AI build-out is this massive CapEx
264
00:17:20.510 --> 00:17:22.290
boom. We're building all these new data centers,
265
00:17:22.670 --> 00:17:26.240
and at various points in the past, it's actually been power-constrained.
266
00:17:26.830 --> 00:17:29.960
And so we need to plug these data centers into something,
267
00:17:30.390 --> 00:17:32.530
but for many, many years,
268
00:17:32.610 --> 00:17:35.690
we haven't actually expanded the US grid and now we're adding all this new
269
00:17:35.710 --> 00:17:38.240
demand. And the electricity grid is a market.
270
00:17:38.370 --> 00:17:40.710
It's got a supply and demand equilibrium.
271
00:17:40.770 --> 00:17:45.030
And so just when you have this equilibrium that's existed for many years,
272
00:17:45.090 --> 00:17:48.450
and then you plunk, this new demand comes on and cannonballs into the pool,
273
00:17:48.740 --> 00:17:49.960
just what happens?
274
00:17:50.530 --> 00:17:55.490
And so I asked Claude to go research this and it's going
275
00:17:55.570 --> 00:17:57.610
off and it's like finding things.
And here, OK,
276
00:17:58.070 --> 00:17:59.750
it said it's picture is striking.
277
00:18:00.470 --> 00:18:03.050
I'm going to need some commodity and equity data to ground this analysis.
278
00:18:03.470 --> 00:18:07.730
Alpha Vantage has what I need. I'd like to buy this stuff. Total is 4¢. Yes,
279
00:18:07.910 --> 00:18:08.810
that is within our budget.
280
00:18:14.750 --> 00:18:17.110
OK. And it is blanching.
281
00:18:19.410 --> 00:18:22.410
So what did you see here? The agent analyzed my question.
282
00:18:22.870 --> 00:18:26.530
It looked for the relevant sources. It found a paid source,
283
00:18:26.990 --> 00:18:31.630
and now it's off buying and downloading that data autonomously.
284
00:18:31.830 --> 00:18:35.530
And you're probably used to your AI doing lots of thinking and building,
285
00:18:35.590 --> 00:18:37.290
but then it's asking you to carry out the grunt work.
286
00:18:37.350 --> 00:18:39.550
It wants you to do the deployment or the checkout flow or the sign-up.
287
00:18:40.050 --> 00:18:44.350
But where we're rapidly headed is the agent doing that work for you.
288
00:18:45.250 --> 00:18:49.550
In his demo yesterday, we used the Link CLI to pay the API reviewer. Here,
289
00:18:49.610 --> 00:18:51.730
as you can see from some of the Tempo requests at the top,
290
00:18:51.810 --> 00:18:55.790
we're actually using the Tempo CLI because my agent has a stablecoin wallet.
291
00:18:56.230 --> 00:19:00.930
Machine payments can use fiat, but for these tiny purchases, for micropayments,
292
00:19:00.990 --> 00:19:03.470
you need a different type of infrastructure. You need stablecoins,
293
00:19:03.810 --> 00:19:05.270
which have near zero transaction costs,
294
00:19:05.590 --> 00:19:10.590
making it viable for the first time.
So it's blanching away here. See,
295
00:19:10.730 --> 00:19:12.130
this is where you really need a fellow with a
296
00:19:12.170 --> 00:19:16.290
guitar.
297
00:19:16.290 --> 00:19:18.490
Just wrong timing.
298
00:19:21.590 --> 00:19:24.670
But while we did splurge for fast mode for you guys,
299
00:19:24.850 --> 00:19:26.110
it's still blanching away there.
300
00:19:26.430 --> 00:19:29.550
And so I just thought I'd show you some cool tabs
301
00:19:31.770 --> 00:19:35.050
while we're waiting. This is from Works in Progress,
302
00:19:35.210 --> 00:19:36.850
Stripe's magazine about progress.
303
00:19:37.150 --> 00:19:39.930
You've probably seen it at the cafe out there and everything.
304
00:19:40.210 --> 00:19:42.850
And just to the discussion of all these new energy sources,
305
00:19:43.470 --> 00:19:46.630
we have this cool article about how Britain made a lot of progress and then
306
00:19:46.730 --> 00:19:50.910
forgot it all in nuclear. What else do we have? We have the Guinndex,
307
00:19:51.030 --> 00:19:52.970
the first real-world application of AI,
308
00:19:53.590 --> 00:19:58.490
where an AI agent called every pub in Ireland to have a real-time tracking
309
00:19:58.610 --> 00:20:02.710
of the cost of a Guinness. So finally, something useful.
310
00:20:07.790 --> 00:20:08.730
Oh, OK. We're done.
311
00:20:10.150 --> 00:20:14.110
So Claude has given me my output. It's opened in my browser.
312
00:20:14.470 --> 00:20:18.270
And so you see here, again, I gave it a single prompt here,
313
00:20:18.510 --> 00:20:22.290
just my kind of one-word question, and it spat out this report.
314
00:20:22.350 --> 00:20:23.183
And what's it saying?
315
00:20:23.290 --> 00:20:27.120
US data center electricity demand is projected to nearly triple by '28.
316
00:20:28.570 --> 00:20:31.040
Hyperscaler CapEx expenditure has surged 62%.
317
00:20:34.470 --> 00:20:36.720
Natural gas prices have surged 104%.
318
00:20:37.530 --> 00:20:40.910
Natural gas is actually picking up a lot of this. So anyway, super interesting.
319
00:20:41.530 --> 00:20:44.490
We don't have time. I would love to just actually read all this.
320
00:20:44.570 --> 00:20:45.670
I'm not going to read all it in front of you,
321
00:20:46.610 --> 00:20:47.830
but I'll be interested to read it later.
322
00:20:47.970 --> 00:20:49.850
You guys will probably be interested to read it later.
323
00:20:49.930 --> 00:20:53.070
And so what I can actually do is go back to my Claude and say,
324
00:20:54.910 --> 00:20:57.890
"Hey, Claude, publish and sell this report.
325
00:20:57.890 --> 00:21:01.390
Price it as you see fit for other agents and humans to find and buy
326
00:21:01.430 --> 00:21:02.263
it."
327
00:21:08.480 --> 00:21:12.470
Great. So it's off working and you saw our vibe-deploying yesterday,
328
00:21:12.550 --> 00:21:16.030
and so it's going to go off and make a website where any of you can buy it.
329
00:21:16.230 --> 00:21:19.110
And actually, now that I say that, I should maybe check.
330
00:21:21.590 --> 00:21:24.190
Check the licensing terms for this Alpha Vantage dataset.
331
00:21:24.770 --> 00:21:27.750
Do I actually have the rights to commercially redistribute the final
332
00:21:27.770 --> 00:21:32.770
report?
333
00:21:33.850 --> 00:21:36.930
Phew! OK. Yes. Check the terms of service, blah, blah, blah, blah, blah.
334
00:21:37.080 --> 00:21:37.913
We're fine.
335
00:21:46.530 --> 00:21:47.430
OK, so it's doing its thing there,
336
00:21:47.490 --> 00:21:49.310
but while we wait for the report to get published,
337
00:21:49.370 --> 00:21:54.250
what should you take away from this? Well, agentic commerce, again, it is here,
338
00:21:55.230 --> 00:21:58.810
and we think there'll be a really big first mover advantage or an early mover
339
00:21:58.850 --> 00:21:59.370
advantage.
340
00:21:59.370 --> 00:22:02.850
It's one of the reasons we're moving so quickly at Stripe to enable you to do
341
00:22:02.930 --> 00:22:07.270
this. If your product or your platform can possibly support machine-to-machine
342
00:22:07.290 --> 00:22:11.790
payments, we think you should build for it now.
And it's still going.
343
00:22:12.050 --> 00:22:16.150
It's blanching. And again,
344
00:22:16.450 --> 00:22:21.410
previously you probably had to go get API keys or go poke around in the Vercel
345
00:22:21.550 --> 00:22:25.090
interface or anything like this. Again, now, thanks to Stripe Projects,
346
00:22:25.610 --> 00:22:27.630
it can orchestrate all of this for you.
347
00:22:29.590 --> 00:22:30.630
Do you have any other good tabs here?
348
00:22:30.870 --> 00:22:33.950
These are some of the companies that already support agentic commerce.
349
00:22:34.230 --> 00:22:37.470
This is Parallel and Browserbase for
350
00:22:39.730 --> 00:22:42.770
agentic web browsing. You have PostalForm,
351
00:22:42.870 --> 00:22:45.450
which your agent can mail a letter for you.
352
00:22:45.650 --> 00:22:49.170
So if you want some compatibility between the new way of working and the old way
353
00:22:49.190 --> 00:22:51.370
of working. OK, here we go.
354
00:22:52.030 --> 00:22:55.950
It is live at johnsreport.vercel.app. So if I just open that, you see here,
355
00:22:56.350 --> 00:22:58.670
you can go to johnsreport.vercel.app.
356
00:22:59.370 --> 00:23:03.870
You can click "purchase report," and I see I'm getting a Link confirmation
357
00:23:03.910 --> 00:23:05.890
there. That's great. That's all working. Stripe Checkout.
358
00:23:05.960 --> 00:23:10.750
But also if I go to llms.txt,
359
00:23:12.450 --> 00:23:17.190
you see here it also constructed an llms.txt for us with instructions for how
360
00:23:17.250 --> 00:23:19.390
with a single Tempo request, agents
361
00:23:21.550 --> 00:23:25.270
can buy the product.
So I would welcome you, indeed, I would encourage you,
362
00:23:25.390 --> 00:23:29.840
I would beseech you to please buy my report for the princely sum of $5.
363
00:23:31.570 --> 00:23:33.710
That will help me with my token budget.
364
00:23:36.830 --> 00:23:39.210
There we go. That is it.
365
00:23:46.650 --> 00:23:49.590
Give them the confetti. These are Gen Zs. They need... There we go. Yeah.
366
00:23:52.010 --> 00:23:55.150
So that's agent commerce live today.
367
00:23:55.430 --> 00:24:00.090
You can go check it out.
And it raises a really interesting question,
368
00:24:00.670 --> 00:24:02.090
which is my third topic.
369
00:24:03.150 --> 00:24:08.070
In a world where intelligence can do all of that and everyone has access
370
00:24:08.110 --> 00:24:11.730
to that kind of intelligence, what actually becomes more valuable?
371
00:24:11.730 --> 00:24:16.190
There's an old rule in economics. When something gets cheap,
372
00:24:16.630 --> 00:24:18.650
its complements get more valuable.
373
00:24:19.510 --> 00:24:23.030
So when containerized shipping collapsed the cost of moving goods,
374
00:24:23.610 --> 00:24:27.390
ports that could handle the ships became much more valuable.
375
00:24:28.730 --> 00:24:32.310
The first radio spectrum auctions in the 1990s raised hundreds of millions of
376
00:24:32.330 --> 00:24:35.490
dollars, but then mobile phones got cheap, loads of people had them,
377
00:24:35.870 --> 00:24:39.790
and the same airwaves were suddenly worth orders of magnitude more.
378
00:24:39.850 --> 00:24:43.150
Governments started auctioning them off for tens of billions of dollars.
379
00:24:43.150 --> 00:24:47.450
So these things have joint demand curves.
And so one question we
380
00:24:47.470 --> 00:24:50.410
should be asking about AI is, what are AI's complements?
381
00:24:51.070 --> 00:24:52.510
What are the complements to intelligence?
382
00:24:52.950 --> 00:24:55.990
What becomes more valuable as intelligence becomes cheaper?
383
00:24:57.710 --> 00:24:59.890
Some of the answers are obvious. For example,
384
00:24:59.950 --> 00:25:02.230
you can see this effect very clearly in chips.
385
00:25:03.130 --> 00:25:06.150
GPUs were really useful before AI,
386
00:25:07.370 --> 00:25:12.090
but it's clear from NVIDIA's deliveries and market cap that chips have become
387
00:25:12.150 --> 00:25:15.130
much more useful recently. They were previously majority gaming,
388
00:25:15.390 --> 00:25:19.010
and now you see the compute networking segment take off.
389
00:25:20.850 --> 00:25:21.683
Same goes for energy.
390
00:25:22.290 --> 00:25:25.430
Power is more valuable if you have intelligence to plug it into.
391
00:25:25.890 --> 00:25:28.070
Nuclear power is undergoing a renaissance,
392
00:25:28.330 --> 00:25:32.790
largely because we need nuclear power to power data centers. In the meantime,
393
00:25:32.930 --> 00:25:35.070
we're going to need a lot of gas turbines,
394
00:25:35.130 --> 00:25:36.590
which you can read about in the report.
395
00:25:37.590 --> 00:25:40.130
You can see the order volumes taking off here.
396
00:25:40.130 --> 00:25:43.010
It's also reflected in the market value of the handful of companies that make
397
00:25:43.030 --> 00:25:46.370
them. This is Siemens Energy, one of the big gas turbine makers.
398
00:25:48.470 --> 00:25:51.750
But a less-discussed complement to AI is proprietary data,
399
00:25:52.170 --> 00:25:55.810
which gets much more valuable when you can let superintelligent agents reason
400
00:25:55.870 --> 00:25:56.703
over it.
401
00:25:58.070 --> 00:26:00.750
One way you can tell data's getting more valuable is that companies that used to
402
00:26:00.790 --> 00:26:03.190
give it away for free have stopped doing that.
403
00:26:04.010 --> 00:26:07.030
This chart shows the percentage of various parts of the internet that have been
404
00:26:07.090 --> 00:26:09.150
shut off to AI crawling, and instead,
405
00:26:09.410 --> 00:26:11.270
those companies are starting to monetize it.
406
00:26:12.050 --> 00:26:13.850
Reddit has always had a ton of data,
407
00:26:14.110 --> 00:26:17.610
all the comments and the subreddits were always there, but before AI,
408
00:26:17.670 --> 00:26:21.850
it was a dormant asset on the balance sheet. And today their non-ad revenue,
409
00:26:21.910 --> 00:26:25.930
which comes from data agreements is $35 million a quarter.
410
00:26:27.470 --> 00:26:30.050
We see the same dynamic with our own data. Take Stripe Radar.
411
00:26:30.710 --> 00:26:34.530
So Stripe Radar has always been able to reason across Stripe's entire corpus of
412
00:26:34.550 --> 00:26:37.670
data, but in recent years, as the AI models have gotten better,
413
00:26:37.730 --> 00:26:39.910
the underlying data is then more valuable.
414
00:26:42.470 --> 00:26:44.330
Network effect is another one we should talk about.
415
00:26:44.950 --> 00:26:48.410
Buyers and sellers still need places to meet, and if switching costs are low,
416
00:26:48.710 --> 00:26:52.070
the network effects are even more important. To understand this better,
417
00:26:52.830 --> 00:26:56.390
we took a look at the public take rate for 10 top marketplaces.
418
00:26:56.890 --> 00:26:59.990
What you see is take rates flat for a few years and then you get this bump up
419
00:27:00.310 --> 00:27:03.550
during the pandemic and then a steady increase over the past three years as
420
00:27:03.630 --> 00:27:06.810
marketplaces see higher returns to better AI techniques.
421
00:27:10.230 --> 00:27:13.890
The last complement I want to call out is just companies that have figured out
422
00:27:14.010 --> 00:27:19.010
the complex interactions between software systems and real-world execution.
423
00:27:19.390 --> 00:27:23.410
I think we'll see that that is an enduring moat. Take John Deere.
424
00:27:24.190 --> 00:27:26.410
If I asked you to name an AI beneficiary,
425
00:27:27.230 --> 00:27:30.030
John Deere might not be the first name that you'd think of,
426
00:27:30.810 --> 00:27:34.050
but they've spent years integrating GPS guidance and machine vision and sensor
427
00:27:34.070 --> 00:27:37.310
arrays into their equipment. And the defensible part, the most here,
428
00:27:37.550 --> 00:27:39.730
it's not the AI. Lots of people could build the AI.
429
00:27:40.830 --> 00:27:44.150
It's having the tractors in the fields across 130 countries.
430
00:27:46.410 --> 00:27:51.150
So those are five things we think probably increase in value alongside AI
431
00:27:51.630 --> 00:27:55.530
and therefore create even more durable competitive advantages in the years
432
00:27:55.550 --> 00:27:57.970
ahead. And for all of you,
433
00:27:58.370 --> 00:28:01.510
AI should change how you think about your own competitive advantages.
434
00:28:02.450 --> 00:28:06.070
You might previously have built the best software in your space,
435
00:28:06.550 --> 00:28:08.630
but you might be finding that software is not the name of the game anymore.
436
00:28:09.030 --> 00:28:12.930
But what you do have is powerful proprietary data, interesting network effects,
437
00:28:13.470 --> 00:28:16.830
real-world operations and tools that took a decade to get right,
438
00:28:17.510 --> 00:28:21.450
all sorts of advantages that hold their value or even become more valuable in a
439
00:28:21.510 --> 00:28:24.550
world of abundant intelligence. So as we wrap up,
440
00:28:25.290 --> 00:28:29.290
will you indulge me in just a little more economic history? I can't resist.
441
00:28:30.270 --> 00:28:31.103
In 1882,
442
00:28:31.690 --> 00:28:35.930
Thomas Edison lit up 82 customers in lower Manhattan using 6
443
00:28:36.410 --> 00:28:40.830
dynamos. Finally, electricity in Manhattan. And for decades after,
444
00:28:41.490 --> 00:28:43.030
even as electricity adoption grew,
445
00:28:43.630 --> 00:28:47.230
productivity growth barely budged or even slowed down as the railroad investment
446
00:28:47.290 --> 00:28:52.190
boom started to wear off.
And the economists were confused initially. I mean,
447
00:28:52.250 --> 00:28:54.110
we had this awesome technology and electricity.
448
00:28:54.250 --> 00:28:56.330
Why wasn't it showing up in the statistics?
449
00:28:57.130 --> 00:29:00.270
And the problem wasn't the technology. The electricity did work.
450
00:29:01.010 --> 00:29:04.570
It was the economy had to digest it. You see, factories,
451
00:29:04.630 --> 00:29:08.210
they'd been built around steam, the shafts and the belts and the floor plans.
452
00:29:08.410 --> 00:29:09.950
It was all wrong for electricity,
453
00:29:10.570 --> 00:29:14.170
and it wasn't until we redesigned factories from scratch that the productivity
454
00:29:14.230 --> 00:29:17.450
gains finally appeared. And people sometimes forget,
455
00:29:17.510 --> 00:29:21.870
but this took a full 30 years from 1882 all the way to the late 1910s.
456
00:29:22.390 --> 00:29:25.550
And then in the 1910s, in that single decade,
457
00:29:26.090 --> 00:29:28.790
the growth rate of output per worker more than doubled,
458
00:29:28.850 --> 00:29:30.430
but there was this big lag.
459
00:29:32.090 --> 00:29:35.030
We saw the same phenomenon again with the birth of computing. In fact,
460
00:29:35.550 --> 00:29:38.210
economists have since dubbed this whole phenomenon The Solow paradox,
461
00:29:38.450 --> 00:29:43.290
after Robert Solow's 1987 quip that computers were everywhere to be seen
462
00:29:43.710 --> 00:29:47.490
except in the productivity statistics.
And he was right,
463
00:29:48.150 --> 00:29:52.650
and they weren't to be seen and they wouldn't be until the mid-1990s.
464
00:29:54.550 --> 00:29:59.090
Transformative technology looks for a long time like it's not doing much.
465
00:29:59.170 --> 00:30:00.550
If you're looking at the economic gauges,
466
00:30:00.630 --> 00:30:02.710
you're kind of sitting there and you're tapping the gauge, is this thing on?
467
00:30:04.050 --> 00:30:07.910
I think this is what we're actually watching in real time with AI.
468
00:30:09.630 --> 00:30:12.250
You see the seeds of this in the phenomena I already mentioned.
469
00:30:12.670 --> 00:30:15.710
The minimum efficient size of a serious business is collapsing.
470
00:30:16.250 --> 00:30:18.570
Solopreneurs are scaling to seven figures and beyond.
471
00:30:19.130 --> 00:30:22.480
Agents are buying from agents. Companies are launching globally from day one,
472
00:30:22.700 --> 00:30:24.760
but none of that fits in the old model.
473
00:30:25.620 --> 00:30:29.060
These changes might yield productivity dividends tomorrow because we have to
474
00:30:29.100 --> 00:30:29.933
digest them.
475
00:30:31.300 --> 00:30:35.740
But they're the early indicators of an economy that's replatforming itself.
476
00:30:36.960 --> 00:30:40.840
The businesses you're all building now, they're not a footnote to AI history.
477
00:30:40.840 --> 00:30:42.180
They are the AI history.
478
00:30:42.240 --> 00:30:45.660
They're the story of the economic and productivity gains.
479
00:30:47.080 --> 00:30:49.980
Electrification took 30 years to reorganize the economy,
480
00:30:50.980 --> 00:30:55.760
but I suspect we won't need to wait anywhere as long as that for AI.
Thank you
481
00:30:55.960 --> 00:30:59.090
so much for being here. I hope you're getting a ton out of Sessions.
482
00:30:59.630 --> 00:31:00.730
Enjoy the rest of your morning,
483
00:31:00.930 --> 00:31:04.530
and I will see you back here for our final fireside with Daniel Gross and Nat
484
00:31:04.590 --> 00:31:05.770
Friedman this afternoon.