Do you think it's a bubble or just the beginning of the AI era?
I don't see it as negative. And are people overspending? Yes. But the people who are making all these investments in America can afford to lose it. The capital allocators in America, they understand something called the power law. Reports suggest that OpenAI and Anthropic alone account for nearly 90% of all startup revenues.
But your question was not about tokens. Your question was about the revenue generated from tokens. If we look at all tokens since Kimmy came out this week, most tokens on the internet are going through their dark tokens. And what's going to happen is they're going to run on your laptop and then there'll be even smaller ones on your phone. When that comes, 90% of tokens are going to run locally. Nobody will make any money from them. They'll be free tokens.
Oh, it's good to be here.
Thank you, everybody. Thank you so much for coming to this show. Welcome to Pivot.
Yeah. So, you are the legend of Android investor.
You know, Ali, investor of Uber. Yeah. And super popular podcaster and founder.
No, I'm so excited to be here also.
Yeah. Thank you. I'm excited to be here as well.
What's the impression of Japanese founders?
I'll step back for one second and I'll tell you why I always send founders and friends of mine to Japan and encourage them to come to Tokyo and experience Japanese culture. If you're a founder, you probably have many of the traits of the Japanese culture and people. And the Japanese culture is actually a very founder culture.
But they don't recognize it. They don't know it.
Everybody in Japan who I meet, whether they work at the Subway or they work at 7-Eleven or they work at Uniqlo, or they make pancakes or they're a barista and make coffee, they all have a sense of pride in their work. They also care deeply about the aesthetic and the design and the experience. When I take my daughters, my daughters are with me for the first time here, or when I bring founders, they're always blown away that everything works and it works well and the people who are operating whatever it happens to be care deeply. In America and in other countries, people don't have the same sense of pride, right? And they also don't have that sense of kaizen where they're always improving. And they don't also have ikigai where they understand what the world needs and what they're good at and what people will pay for, right? And what they enjoy doing.
But founders understand that and those concepts are also built into Japanese culture. Also in Japanese culture there's a giri, like the sense of honor and purpose, right? And so when I bring them here, I tell them, 'Go here and have the French toast. Go here and have the tempura. Go here, have the tonkatsu, katsu curry. And go here and go here and go here and have coffee.' They come back to me and they say, 'I've never been to a society and seen everybody care so much about what they're doing and have such pride and detail work.' Cuz when you're a founder, even in America, you feel strange because you meet 10 people or 20, maybe you meet 20 people and only one of them thinks like a founder or a Japanese person. But people in Japan don't know. They're all founders. They all think like a founder. But there are some things that are different. There's also in Japan a sense of, with that honor and giri, that if you fail, it is terrible. The sense of honor that you failed, dishonor that you failed, is too heavy.
In startups, it's okay to fail. In America, people have no shame in failing.
In Japan, people work harder and have more detail oriented and the kaizen and the giri and very good focus on design, but they care too much about failing.
So, each society is slightly different. The entrepreneurs here for the last maybe 20 years I've been coming, they care too much about failure and they have not been bold enough to go after the world. But the history of Japan from Sony to Nintendo to Toyota, many great companies, the goal was to take on the world. And I watched in the 80s when I was growing up and I worked at Sony in America. And I watched Sony just inspire everybody with the Walkman.
With movies, with the laptops, you know, just every product one after the next after the next. But then something happened in the 90s and early 2000s. The Japanese customer was so good, it was easier to just sell to them. So in America, we didn't hear from the Japanese founders.
But something in the last 10 years has changed again. Uniqlo taking over the world. Japanese animation taking over the world. Still Nintendo doing very well. In talking to the founders here, the thing that changed was people's parents and what their hope is for their child.
So I was talking to a journalist yesterday and we're both Gen X, both born in the 1970s. And I said, if you told your parents you were going to go work at this very prestigious newspaper or you were going to be a founder, would they be scared and happy? They said, 'Oh, very scared that I would be a founder. They'd be very happy that I had a safe job and I was a salary man, right?' And I said, 'Well, do you have kids?' He said, 'Yeah, two kids, 17 and 20 years old. If they came to you want to do a startup, would you be scared or would you be happy?' 'Oh, I'd be so happy.' So this generational change has happened and it's reasonable. Parents want the best for their kid. It used to be the thinking, oh, at least in Japan, the best thing for my kid is safe, safe job, 30 years.
But now I think people realize, well, we know what's going to happen on that path. It's going to be safe, but there's no chance to break out. It's good, but it's not great.
And then if you do a startup, you'll learn faster. And if you fail, you start again. If you fail, you do a third company. And if it doesn't work out, you still can go to the big company.
And they will covet you more because you tried to be an entrepreneur and you learned a lot. So the even the worst case scenario, you fail at three companies and you're 35 years old and you wasted 15 years of your life, you are going to be the best employee Sony ever had, you know, or Nintendo ever had. You're going to be super special.
And so that's why I'm here. I think Japan is going to have the best 20 years, for the next 20 years will be the best 20 years of startups.
It's great to hear that. So we have potential, we have strength.
Unlimited potential because of the discipline to make great products and to care deeply. And the people who are going to win in this new AI era are the ones who don't quit. So the number one reason a startup fails is not because they run out of money. It's not because they don't get product market fit. It's not because their product is not good enough. It's because the founder quits.
I've seen many companies lose customers, the product doesn't work, they run out of money, but the founder doesn't quit. And then they figure it out and they find 10 customers. And those 10 customers tell 10 others. And then they raise the price and then all of a sudden the company is saved and they make something brilliant.
Cuz they pivot and pivot again.
And they figure it out.
And we call this like understanding the idea maze is the concept. Every market and every problem set and product plus consumers equals this maze and some people get in the maze and they give up and then other ones keep fighting and fighting and they don't leave the maze. They just keep going and going until they get to the other side and they get the cheese.
That ability I think is very unique to Japanese culture as well as American. The doggedness, never quitting.
As you mentioned the AI era, so I'd like to talk about, you know, AI bubble or not.
A bubble has a negative connotation. I don't see it as negative. I see it as like waves.
Yeah.
So when the waves are coming in and they're very strong like this, I think a lot of the people get excited and they all go to try to ride the wave. And some people get knocked out, some people miss the wave, but some people figure it out. So whenever you have this bubble and it pops, okay, it pops, but there's like something very strong underneath it and then it goes a little bit, it pops, boom, but it keeps growing. The waves keep growing. So there's so much excitement because there's something important here.
And are people overspending? Yes. But the people who are making all these investments in America can afford to lose it. The capital allocators in America, they understand something called the power law. Most of the returns come from a small number of companies or investments. So you have 100 people try to make an LLM. You have 50 people try to make software for accountants. We have one tax GPT is doing very well. You have a hundred companies try to do training data. And we have one micro one that's doing very well and this happens over and over and over again and most fail. But if of the hundred, you know, 95 fail and five succeed, the five who succeed will return more capital than was invested in the hundred. That's the power law. 100 companies raise 10 million each just for argument sake. And 99 of them go out of business.
Yeah. If that one becomes worth 10 billion, 1 billion in investment turned into 10 billion. For people who don't understand what just happened, they look at it and go, 'Oh my god, everyone failed. It's a total failure.' If you are a capital allocator, you go, 'Oh, Amazon, Google, Apple, Microsoft, Tesla, Nvidia.' So that's the power law.
Yeah. Almost in every case a market reduces down to one or two big winners and then you know some smaller ones who carve out some space there.
But generally we see a duopoly, sometimes three, but usually one or two big winners.
This ecosystem of Silicon Valley I think everybody in the world wants to replicate it.
But sometimes the press, the media, people not in the industry, they look at it and go, 'Oh, it's a bubble. Oh, no. I can't wait for the next bubble. I like to be on top of the bubble.' I look at it like a cappuccino. Lots of little bubbles. It's okay. It's delicious.
It's okay to have the bubble. As long as something makes it through to the other side that changes the world.
So, compared with 1999 bubble burst.
Can you see some, you know, similarity or?
Sure. At that time there was a mania and for a good reason. Investors understood the internet was incredibly powerful and would change humanity forever. So the thesis was correct. The internet would change everything. However, the internet needed to be built out more. There was not high speed and fiber. Storage was very expensive. Compute was very expensive. And people wouldn't put their credit cards on the internet yet. There were many things that needed to be built. And so what happened was when those companies became overvalued, they were not profitable yet. They didn't have customers paying and the infrastructure wasn't there. So in that case, when the bubble burst, most of the companies went out of business. This time when the bubble deflates, and it will, some people say it's happening right now when we're taping this, that it's deflating a little bit. What you'll see happen is, oh yeah, this company had no customers and it went to zero. But you'll look at something like Claude and Anthropic or you'll look at something like OpenAI or Nvidia and you'll look at these companies and go, 'Oh yeah, they had 20% their revenue was flat one year or their revenue went from growing 10 times year-over-year to two times year-over-year. Oh, they lost money. They did layoffs.' But the amount of demand for intelligence, I believe there is no upper bound. It is unlimited the desire for intelligence on demand.
Just like energy has proven to be essentially unlimited for humans. Humans will take as much energy as possible. Now there have been moments where people stop, gas becomes too expensive, people get a hybrid car, people ride a bicycle like you know during the 70s oil crisis or during COVID and supply shock. So there are times where you have up and down but up and down and trending up.
And yeah then people make refrigerators that use you know half as much electricity. They put solar on their roof. They get batteries and fill them up.
There's all kinds of ways that people make it more efficient along that journey, but the demand for electricity continues.
Same thing is true of intelligence, but times 100 or times a thousand with energy. Everybody has questions that they want to get an answer for all the time.
Everybody wakes up every day with problems they want to solve. Every business wakes up every day with an infinite number of problems they want to solve. There is no world in which I think the demand for intelligence will go down. Not in a hundred years. Like this is the greatest product ever created, on-demand intelligence. In other words, problems being solved by computers with very little human intervention is the last product with infinite demand. Now the prices will change, the profit will change. Open-source versus frontier model, Nvidia versus, you know, other hardware solutions, energy coming from nuclear, solar, natural gas. There'll be many different factors that come into the price. But I don't think humans are going to stop wanting to ask questions or solve problems.
We are problem-solving monkeys at our core. The nature of humans is we make tools to solve problems.
And if we don't have problems to solve, we go crazy.
The worst thing you could do to a human is put them in a room with unlimited food and shut the door and give them nothing to do. They'll lose their mind. The greatest form of torture is solitary confinement.
A person will go mad with no purpose in life.
So that's why humans constantly look for a new problem to solve.
Interesting. So AI can dramatically lower making products, right?
Yes. So it's very positive for founders.
For startups, when I started in the industry, you would raise 3 million, 4 million, 5 million in venture capital in the 90s and in the early maybe, you know, until 2000. You would raise three or 4 million. You would work 12, 18 months and release a product and you'd have 30 people at your startup. So 18 months and 3 million to release a product. Then cloud computing happened. Co-working spaces like WeWork, on-demand services like HR, legal, accounting on demand. You didn't have to hire an HR team.
You didn't have to hire and buy servers anymore. Just was in the cloud. You didn't have to get an office manager anymore. You didn't have to put down 250,000 to rent an office. You could just put your credit card down and rent two desks. And then you need a third desk. You can get a third desk. That person works from home. Now you go back down to two desks.
So then in the 2000s, we saw people build a startup for 200 maybe 300,000. They'd have 300,000. They'd have five people and they would have their product out in 6 months.
So from 3 million and 18 months to 300,000, right? 90% less money. Today, people will start a company for 0. Just one person or two people, sweat equity, vibe coding, using Claude Code, using Lovable, Grok, whatever it is. They build a product. They put it in market in 48 hours. I could build a product with you right now,
And have it in under three days and I could make 100 and get my first customer.
For no dollars. That's the future. And we actually see some entrepreneurs now, they start a company and they build three products at once. They test all three products, then they turn off all three. They get a new idea, they test three more and then they shut down two and have one.
We see them do that over three months. They just iterate, iterate, iterate.
And they don't need a developer anymore. They can just build it themselves. They don't need a designer. They don't need an HR person. They don't need an accountant. They don't need a VC. There'll be many companies started without venture capital that get to a million dollars in revenue.
But also, I'd like to ask about OpenAI and Anthropic.
Yeah. So reports suggest that OpenAI and Anthropic alone account for nearly 90% of all startup revenues. So what is your take on this situation?
They are the first two people to come out with models that are very good. Probably right now 1.5 billion people using AI, maybe you know a couple of hundred million on ChatGPT and Claude and then maybe a billion on Google. And the people on Google are just getting AI search results but there's hundreds of millions of people in business, I would say maybe 500 million minimum in business globally paying for or using AI. We've never seen that before in the history of business. We've never seen in three years it to go from zero, all the and zero customers for a large language model, a new product all the way to 1 billion people paying 20 a month.
If there's one billion people paying 20 a month, that's every year 240 billion.
Yeah, that's huge. It's a lot of money.
In just three years. And it started out with 10 million people. Then it went to probably a hundred million. And now it's probably, you know, 500 million people paying, maybe even a billion. And it's going to go to, I would say, every person on the planet will be paying for AI either paying through advertising like on Google or paying on their credit card or their business paying. And it's also very inspiring. So, just like a consumer, if they get asked to pay 20 or 10, like, sure, I'm having fun. This is cool. And that's how much value is being created. The problem with this is if you were to look at tokens being used, there's more tokens being used now by open source. But your question was not about tokens. Your question was about the revenue generated from tokens. And yes, those two are a lot of it. If we look at all tokens since Kimmy came out this week,
It flipped again to most tokens on the internet are going through their dark tokens. You don't know. You can't see them because they're not from a company. They're open source. And what's going to happen is they're going to run on your laptop.
I need to go to Apple store. Mine now has 16.
Yes. Immediately.
I have the M5.
48 gigs of RAM. Only 2,500.
The next one I'm going to get is the Mac Studio with 512 gigs or maybe even when M6 and M7 come out, it'll have 1 TB and it'll be 10,000. People are starting to load the language model on their laptop or their Mac Studio. Those tokens will be running 24 hours a day on your local machine. And then there'll be even smaller ones on your phone. So the next generation of iPhone is going to have a small language model, SLM.
SLM, small language model.
And then they're building vertical small language models. So here's a model only for video editing. Here's a model only for legal, a model only for images, a model only for voice, a model only for photos to organize your phone photo library. That's coming. When that comes, 90% of tokens are going to run locally. Nobody will make any money from them. They'll be free tokens.
If I asked you, how much do you spend on internet access and storage this year?
Per year, per month? Do you know?
Of course. Why? Because it's close to free.
You spend more on coffee than you do on storage.
You spend more on coffee than you do on storage or bandwidth.
Mhm.
Because it's so free.
And every year it gets cheaper. If it's getting 15, 20% cheaper every year, every 3 years it's half the price.
On the token front it's getting 90% cheaper per year. 10 this year, 1 cent. 10 to 1 cent in four years. It's going to drop that fast. Yeah, it's already dropping that fast. The hardware is getting more powerful, cheaper. The software is getting more optimized, so therefore it can do more with less energy, less storage, less memory. It's an unlimited amount of demand for intelligence, and it's going to get so cheap that you're going to not think about it.
When you got your computer in the 90s and you dialed up to the internet, it was 5 an hour, then 3, then unlimited.
And then when you had your computer, you had to have extra hard drives. And if you were recording a studio show, your editors would be constantly coming to the boss and saying, 'Can I get another hard drive? Can I get more storage? Can I get more zip discs? Can I get more CD-ROMs?' Storage was a big issue. Now, nobody even thinks about storage.
Storage is like it's free. It's like oxygen or like water coming from a tap.
How much did you spend on water last year?
You don't know. Nobody knows.
Yeah. I don't have to care. Maybe.
That it doesn't matter. So, when we worry about water today, people like, 'Oh my god, we're using so much water.' It's a very simple solution.
Make it one penny more per gallon until people notice.
But right now nobody notices their water use.
Nobody thinks about their electricity use. Nobody thinks about their storage use. Nobody thinks about their bandwidth use. And do we think about our food usage?
Yeah, people think about food. Food is still a little bit expensive.
You might think about it, especially if you're having steak, you know, or sushi, something that's a limited supply.
Tokens, all this is going to be very cheap.
It's going to be so cheap that people are not going to think about it.
Mhm. So it's definitely the big wave as you mentioned.
But you know a lot of viewers are interested in you know potential IPOs. Do you think they go to public?
They will go to public. Of course.
