0:00
Jeff, welcome to Giant Ideas.
Search this video
0:02
>> Thank you very much, Cam. I appreciate it. Thanks for having me.
0:05
>> You've spent the the last 20 years building one of the the defining internet infrastructure companies, Twilio, and you've now shifted gears and are choosing to focus and try and solve one of the hardest problems in physics, fusion. Why?
0:18
>> Well, that's a great question.
0:18
You know, I had been interested in curious about fusion for a very long time.
0:25
Um, starting with just I took a bunch of physics classes in in college.
0:26
And you know, the reason is obvious, right?
0:30
The promise of fusion is that if you can have cheap energy that is made out of abundant resources on Earth that will never run out, that is both clean and safe. Right?
0:43
This is the perfect This is the holy grail of humanity's energy needs.
0:49
And if you succeeded at bringing this to market, you could transform not just, you know, everyone thinks about electricity as the number one thing and bringing down rate at the meters, which you certainly can.
0:58
Uh, and obviously put out, you know, zero emissions doing so.
1:04
Um, yeah, you think about the use for AI data centers in today's age.
1:06
Obviously, everybody is is worried about the role that AI data centers are going to play in and and and the grid and having enough energy for them and making sure that energy isn't isn't dirty.
1:15
Um, but not only that, you can transform other industries, too.
1:19
Uh, the heat generated by a fusion um, reaction is so high quality and so high temperature, you could use it to green the concrete industry, the steel industry, plastics, glass, all these industrial domains that run these blast furnaces that are emitting tons of carbon into the environment, you could instantly green all of these industries.
1:41
Okay, so you know, what what's what's preventing this?
1:46
Well, it turns out that for 75 years, people have been wanting to make fusion work.
1:49
It's just that the science hasn't cooperated.
1:51
Humanity just doesn't know enough about the behavior of these plasmas at extreme temperatures to make it work.
1:59
And at a high level for anyone who isn't familiar with fusion, fusion is the act of taking light atoms like hydrogen isotopes usually and under intense pressure and heat getting them to fuse into a helium atom.
2:15
And just like fission where you take these big heavy elements and split them, when you fuse two light elements together, they also release a ton of energy.
2:25
And so, this has been the promise.
2:25
So, several years ago, about 40 years ago, there was a breakthrough at Lawrence Livermore National Lab here in California where I am.
2:33
And this breakthrough was the first time that any experiment, any device, anywhere in the world actually succeeded in creating energy.
2:44
It gave off more energy than it took the experiment to start.
2:47
Because people have known how to create fusion reactions for 75 years.
2:51
It just always takes more energy to get the little bit of fusion energy out and a ton of energy in.
2:59
Uh it just makes no economic sense for why you would ever do this.
3:01
Uh but this experiment at Lawrence Livermore was the first time that scientists ever got one of those experiments to give off more energy than it took to >> Was that your eureka moment?
3:13
>> I mean, that I mean, for for the world, that was the eureka moment. It was a huge thing.
3:17
Well, anyway, so, I after moving on from Twilio, I was curious just to learn more about fusion and and see where this whole thing is going.
3:24
And I managed to connect with the lead designer of the experiment, the person who actually uh created all the parameters that allowed it to come together and finally work. This woman, Dr. Annie Krikorian.
3:35
And I was chatting with her about, you know, where all this is going.
3:40
And I you know, cuz I was expecting after this huge announcement, I mean, it really was uh a huge announcement.
3:45
One, you know, like a a true global.
3:47
It should probably win the Nobel Prize.
3:50
Um the the I was expecting to a few months later read, oh and then there was this spin-out and now there's a you know some company or some startup that's going to commercialize this amazing technology and I never read about such a company.
4:02
And I was really curious to understand why.
4:04
And so I was talking to Annie and you know and she said she she said, yeah I don't really understand it either but I'm not a business person and so I don't know why no one is commercializing it but I've been told that people just think it's too expensive to go uh commercialize this technology.
4:20
So they're continuing to look at you know other ways that have never actually worked but you know people think maybe will work one day.
4:26
And I was like, what that seems a bit odd like this eureka moment for society after 75 years they made it work and everyone's kind of ignoring it for commercial purposes. Well, why is that?
4:37
And what I found were two reasons that people generally gave.
4:39
They said you know first of all so so the the way the experiment works is they take this tiny little target of fuel.
4:45
It's about that big about the size of a peppercorn.
4:48
And they hit this tiny little thing full of uh deuterium and tritium which are just two isotopes of hydrogen.
4:51
And they hit it with the world's largest laser from all sides.
4:56
They hit it with all this laser light uh megajoules of energy uh in this one pulse.
5:01
And if you get it just right you compress this little sphere of fuel down to this tiny thing and inside you've created now you know huge temperatures and pressures that allow you to actually create that fusion then and then it creates all this energy.
5:14
And so the story was the reason why you couldn't commercialize it was because oh well building a laser for a a plant scale endeavor has to be a much bigger laser, a much more efficient laser and it's made out of diodes.
5:26
And people said, oh these diodes are so expensive.
5:30
In fact, it to try to go build a a power plant out of this will cost you a hundred billion dollars because these laser diodes are so expensive and you need so many of them.
5:40
And I kind of look at this and said, oh well that that you know that's a lot of money right a hundred billion dollars.
5:45
What's the deal with these diodes?
5:45
And I go look into it just to understand this better, and it turns out that this particular kind of laser diode that you would need is this tiny cottage industry.
5:55
>> [snorts] >> It's this, you know, there's only a few use cases for it.
5:57
The the total output of the global industry of this particular kind of diode is a 0.
6:03
1% of what you would need the industry to be for building a plant. Okay.
6:06
So, I'd have to scale up this diode, this semiconductor fabrication industry by about a thousand X what it is today to go build a plant.
6:16
And a lot of people are like, "Well, see, that's the reason why you can't do it."
6:18
I was like, "Well, actually, that's the whole name of the game, right?
6:22
If it's too expensive and it's this tiny scale thing today, and we scale it up a thousand X, guess what?
6:26
We all know what happens when you scale up semiconductor fabrication by orders of magnitude, the cost go like that, right? It's simple. Um now, it's not simple.
6:37
There's a lot of work to go do, but the economics of semiconductor scaling are well understood.
6:43
So, I was like, "Okay, well, that seems like a solvable problem on the grand scheme of things, you know, compared to solving basic science problems that have eluded humanity for 75 years.
6:53
Scaling up a little corner of the semiconductor industry with well-known process like Okay, that seems easy."
6:59
All right, what's the other problem?
7:00
There's another problem people said.
7:01
They said, "You know, Jeff, these targets that they shoot, each one costs half a million dollars to make.
7:09
And in order to create energy, you're shooting these targets 10 times a second.
7:14
And each time you hit them, they they basically explode and give off all this energy, so you need to replace it.
7:19
And so, they're like, "Jeff, you're going to blow up $5 million a second.
7:24
How are the economics of this power plant ever going to work out?"
7:28
And I said, you know, I was like, "Well, yeah, that doesn't sound very good.
7:29
I don't know I'm blowing up $5 million a second. Okay.
7:32
What's the deal with these targets?"
7:34
And start looking into it, and I'm like, "Okay, there's only really four components to these targets.
7:40
You can make these out of fairly, you know, cheap materials.
7:42
It's not like the materials cost.
7:44
Okay, like why are they so expensive?
7:46
And I'm talking to Annie and I say, "How many of these targets, you know, a year do you guys shoot?" And she says, "Six." I'm like, "What? Six million? Six billion?" She's like, "No, no. Six.
8:00
We shoot We We run this experiment about every other month.
8:08
And each time, by the way, we're doing science.
8:09
We're not just doing the same thing every time.
8:10
We're trying to learn new things all the time.
8:11
So, every time we make one of these targets, it's different.
8:15
Slightly different specifications in order to learn and see if we can, you know, make the experiment work even better."
8:21
And I'm like, "Oh, I know what these are called. Prototypes.
8:25
You guys are prototyping in the lab.
8:27
That's what you're doing, right?
8:28
You're making a bunch of different variations, each a little different, to see if you can get it to work."
8:32
And I'm like, "Okay, well, then what do you do when you have a prototype you like and you want to bring it to market?"
8:37
Well, you go take it and you go figure out how to make it in a factory.
8:41
And when you do, you now amortize the cost of, you know, the the targets aren't that expensive to make.
8:46
They're just expensive because you're amortizing big fixed costs over six units.
8:55
So, now you got to amortize your costs over, you know, millions or billions of units and that's how the economies of scale kick in.
9:00
And in fact, these things can be made quite cheaply.
9:01
You just have to figure out like all the process steps to go take something that's done by hand today and automate it.
9:06
today and automate it. And it turns out that there are a lot of people in the world, in fact, many of them right down the street from where we are in the Bay Area, that work at places like Apple, that every single year are figuring out
9:18
how do I bring some product to market with some new specifications that no one's ever made before, and we got to go figure out how to take some designer's whim and turn it into a billion units at a price point and a and a and a yield and everything else. So, I was like,
9:29
So, I was like, this is another thing that is, you know, not easily solved.
9:36
>> Maybe to radically simplify it for listeners, you basically looked at the problem, you said this has been solved scientifically, but let's reduce it to first principles.
9:42
There's kind of two core inputs that I think we could radically reduce the cost of production of if we were to do this at scale, we'll get huge economies of scale if we apply best practices that we've learned in the chip industry.
9:53
And so all these assumptions that the people saying this is too expensive, this can never work, they're now they're now false.
9:58
So what what's the key like challenge commercially?
10:01
>> Well, and that's what we set out to prove.
10:02
What we set out to prove.
10:02
So we started in inertia to go with the first step, prove that those two things that people thought were impediments to commercialization are in fact false.
10:14
>> And what do you think the key commercial challenges today to bring this science to scale?
10:20
>> Well, it's mostly about building up a supply chain and doing process development.
10:23
So for example, we've got to figure out how to scale up these diodes.
10:26
We're partnering with the major laser diode vendors in the world and basically doing process development with them and say, "Okay, well, you know, this part of the process here to make it is, you know, using a I don't know, a 3-in wafer.
10:37
Okay, what do we need to go take that up to a modern 6-in wafer process.
10:41
Or hey, this assembly step because, you know, you have to hold package them and process them after you make the semiconductors.
10:47
Okay, well, that step is done manually today by a person in a lab because, you know, the demand is so low, there's no incentive to go automate it."
10:55
Okay, but if you look at scaled things that we make, it's all automated.
10:59
The whole end-end process is completely automated.
11:01
Okay, well, what do we need to do to go automate this process?
11:04
And like build, you know, robotics in and automation and increase batch sizes and things like that.
11:08
And so it's basically bringing the best practices that everything we've learned about fabricating uh semiconductors and uh consumer electronics in other areas of in other disciplines and just bring them into this one. >> Fascinating.
11:23
Reminds me a lot of um the approach the founder at PsiQuantum's taking to to quantum computing, which we also had him on the podcast.
11:29
But, you know, other very well respected, well-regarded fusion companies like Commonwealth, Helion, they've taken a completely different approach. So, why are they wrong?
11:39
>> Look, I I'm not going to say they're wrong.
11:40
I'm not smart enough to to know if if what they're doing is going to work or not.
11:44
All I would say is that given the choice when it's time to go commercialize, given the choice between an approach where the physics is well understood and well known.
11:56
>> And yours is called inertial confinement, correct?
11:59
>> Yeah, that's the inertial confinement fusion.
12:00
In fact, it's it's I think people often confuse the fact that there's because this one experiment worked, all that says is the very pretty specific conditions that they created there are now a well-understood regime.
12:16
But, if you change things too much, now suddenly you've wandered back into the realm of of unknown physics.
12:20
And so, one of the things that Inertia focuses on is making sure that all the things we do in order to create this scale-up, we are always staying within the boundaries of what is this known physics regime.
12:32
And it's a very specific one.
12:34
You hit a certain kind of target with a certain kind of laser pulse. We know it works.
12:39
That took 75 years to figure out.
12:39
Okay, so one of the principles of our company is number one, start with understood physics.
12:46
Number two, don't mess up the physics.
12:48
And so, that's kind of how we think about the problem.
12:50
And in fact, you know, our road map is actually pretty simple.
12:54
You know, step one, start with known physics.
12:56
Number two, go build the world's most powerful and most energetic laser.
13:05
Number three, go build the world's first fuel fusion fuel target assembly line, like a factory.
13:12
And then, after you figure those two things out, go build them into the first-of-a-kind power plants to actually bring all these technologies together and produce energy.
13:22
And so, that's really our road map. It's quite simple.
13:24
>> you succeed, what is how does this change life of the average person?
13:29
>> Well, I think you'll see is, you know, fusion is going to it is really the holy grail of humanity's energy needs, right?
13:36
So, a lot of the consternation that we have about our energy being either too expensive, too dirty, or too unsafe, or non um uh renewable, those all go away.
13:46
And so, I think what you'll see is a lot of the um consternation around climate at least from the corner of the of the of the uh of the production, which is either based on, you know, burning petroleum in cars, now that we have EVs, uh whether it is how we get electricity to our home, whether it is how we are um uh creating concrete, plastic, steel, all these things can now be clean industries with zero emissions.
14:14
And between all of these things, you've now chipped away at the major sources of our our climate problems.
14:21
And so, not only does energy get cheaper, it also gets better in terms of every attribute you can possibly look at for how where we get our energy.
14:29
And even things that are un im- impossible today become possible, like you can have desalinization everywhere you need clean drinking water, because suddenly the energy you need to power it is essentially cheap and clean.
14:44
And so, you know, the other thing, by the way, is there's no unlike uh fission plants, we have proliferation risk and things like that.
14:50
There's nothing in fusion that could allow you to create weapons, create national security problems, um and even the inputs to it are so commodity that we will never run out of them.
15:03
In fact, you could power your entire life's energy needs with a bathtub full of water and a single laptop battery worth of lithium.
15:16
That's how efficient this process is. Can you imagine that? >> That is incredible. That's a step change.
15:20
But, some people would argue that you know, solar batteries, geothermal are getting more and more effective, improving rapidly.
15:27
The kind of the cost curves when you compare the two doesn't really make sense to be chasing this fusion dream.
15:32
What do you What would you say to that?
15:34
>> Well, the thing that I agree with there is the best form of energy.
15:36
In fact, you know, if you think about it, the most effective way of scaling up energy over the last 50 years, give or take, has been solar and batteries.
15:45
And there's a particular reason why.
15:47
It's because we can make them in factories.
15:51
You know, unlike big fission plants or even geothermal projects or even like you know, wind projects or or hydro, you know, these are all big projects where each project is a bespoke implementation.
16:05
And that's why they're slow, expensive, usually end up costing five times as much as the designers said it was going to cost the beginning of the project.
16:12
That's what makes them so hard.
16:13
But, solar and batteries have the amazing property which is we make them in factories.
16:18
And so, you can get better and better at them.
16:20
There's a learning curve.
16:21
And we you know, if each panel gets you know, twice as a twice as efficient and effective and half the cost every five or 10 years, you you can see these curves kick in. That is really powerful.
16:33
Now, the thing about solar and battery is there's only so much of the earth's surface that we're going to you [snorts] know, devote to taking over with solar panels.
16:43
You know, and if that wasn't the case, like if it wasn't the societal pushback on we don't want solar panels absolutely everywhere because you know, they can be an eyesore or whatever, then I think you'd already see solar having taken over our grid.
16:54
You just haven't seen it happen because there's enough pushback.
16:58
And so, certainly there are great applications of solar in places where there is a lot of sunlight and not a lot of people.
17:03
Like that's where you see it spring up.
17:04
But, in places where there isn't as much sunlight or there is more population density.
17:07
And it turns out electricity is hard to move.
17:09
So, you really do need your power generating somewhat close to your population centers because it's lost.
17:14
You lose energy across transmission.
17:17
Uh and so and I think that's why and I think that also the inputs to solar and batteries and batteries in particular, at least today's chemistries, you know, you do start to look at it and say, "Wow, we actually start to run into the availability of the materials needed to make these batteries."
17:33
Now, we're not like close to those limits today, but if you really try to deploy these resources at societal scale, you would start to say, "Hey, look, we actually don't have the right, you know, amount of these elements uh at least available in any reasonable way to produce as much battery as we need."
17:48
Now, over a long enough horizon, sure, battery chemistries can evolve and can change and become easier to produce.
17:53
Um so, I think it's a combination.
17:55
In fact, I I think that 50 years from now there will be two sources of energy on the grid.
18:02
Uh there will be solar plus battery >> Mhm.
18:05
>> and there will be fusion energy. >> Yeah.
18:06
>> And there will be no need to produce anything else. >> I think so.
18:10
>> Because solar and battery will be an effective way uh but I think fusion will be cheaper, uh easier to deploy, and um uh and actually just much more available in terms of its inputs.
18:21
And so, I think you will see those two forms win out.
18:24
>> My view is the more the merrier, but I agree with you about those two paradigms.
18:26
Um AI's been like a huge boon for boon for nuclear, right?
18:29
I mean, that's one of the reasons we've seen such a renewed interest in fusion and fission is we need this clean energy base load for data centers.
18:36
Um do you think that fusion's going to go the same way as some of these other critical technologies like AI and and minerals where it becomes a real sovereign race between between nations?
18:45
Is that kind of what you're seeing?
18:50
>> I think there is a bit of that going on, yes.
18:51
Uh you know, frankly, I think you do see nation states uh where you know, we're living in a world now where, you know, I think like the the war in Ukraine showed people that the global market for energy is not one we can take for granted uh because energy was cut off to a lot of uh Western Europe uh and uh and people realized that the precarious position they might be in.
19:13
And so I think that's why you see a lot of uh geopolitical concerns driving this race towards new forms of energy.
19:19
And in fact, you see the Chinese are investing a lot in fusion.
19:25
Uh your government uh the British government actually has just put a few billion uh pounds to work to develop fusion uh domestically for you.
19:30
Uh the Germans have been investing a lot because they have already committed to divesting themselves of fission plants.
19:39
Um kind of related around the time when their energy needs got more complicated by uh their their supplier of gas which turns out was Russia.
19:45
So, you know, you see a lot of geopolitical concerns here driving the need for every country to want to make sure that they are energy independent or they have a reasonable path to get there.
19:56
And you just can't do that with natural resources that don't exist everywhere, you know, not everywhere has coal, gas, had dinosaurs uh you know, a few hundred million years ago.
20:05
And so everyone's got to go figure out, okay, you know, what is something that we can power our society with that we do have domestically.
20:11
And it turns out water and a little bit of lithium, most people can get their hands on those things.
20:14
And so fusion is a great uh natural resource.
20:18
In fact, compared to fission which I agree is having a resurgence because of AI and everything else, but again, the inputs to it, you know, so you know, you highly enriched uranium.
20:26
Yeah, that turns out that isn't everywhere.
20:29
And even if it was everywhere, how to enrich it and whether the world wants you to enrich it uh is a whole other question.
20:35
And and again, that's why fusion is uh >> [laughter] >> is such a compelling uh technology here for for humanities.
20:41
>> So so this experiment at the National Institute was a huge moment, right?
20:42
Cuz you they should prove that you could use a laser to smash atoms together to create more energy than it took to to do so.
20:51
But the quip would be that fusion has always been 20 years away for the last 60 years.
20:55
So if I'm going to push you on commercialization timeline, when will we see a fusion plant delivering real energy to the grid, what would what would you say?
21:03
>> Our goal is to have it uh first commissioning of that plant be about 10 years from now, although it won't be fully uh fully I should caveat that, right?
21:10
I think it'll be 15 years before we're actually putting out serious energy on the grid.
21:15
So, you know, unlike others who say it's just around the it's you know, next year and they'll say that for 20 years.
21:18
I'm not going to say that.
21:20
I'm going to be realistic with you.
21:22
So, it's not an answer to like tomorrow's AI data center problem.
21:26
And I just think that as a entrepreneur, I've always thought that the best thing to do is under promise and over deliver.
21:35
So, do I hope that we can deliver energy before 15 years from now? Absolutely.
21:39
But I want to under promise and over deliver because I think that's that's the way to set expectations properly.
21:44
But you know, it's interesting and then you said fusion's always 20 years away.
21:48
In fact, I've always said 30 years away.
21:49
So, I'm glad that you're more optimistic or maybe that's just our you know, uh maybe that's just sort of a conversion factor for you know, imperial units.
21:55
I don't know, but >> [laughter] >> The the reason why is fascinating.
22:00
And you know, there's this analogy that that we like to tell.
22:05
If you go back uh to the found you know, the sort of first understanding of fusion in the 1940s, uh which is when it was first kind of really figured out the basics of it.
22:14
The the analogy is, okay, people heard about an island far off that was full of riches.
22:23
Gold, you know, endless amounts of gold and treasure.
22:25
Okay, so explorers set off in all directions to try and go find this island.
22:30
You know, some you know, one went north, one went east, one went west, one went south, right?
22:33
They all went in these different directions hoping to find their island.
22:36
And they kept believing that just over the horizon, they're going to find their island.
22:42
And of course, they sail another day and you know, the island hasn't appeared and they okay, well, I'm pretty sure it's just another day out, right?
22:46
And that's kind of been the the the the attempt to go find fusion and make it work.
22:50
Well, it turns out that this one approach, you know, hitting a tar a very specific kind of target with a very specific kind of laser, it turns out the people doing that are the ones who found the island.
23:02
They landed on the island. >> There we go.
23:04
>> And so therefore it's like now we're you're talking about a known quantity because if you're still looking for your island, it could be that it's just a day's sail away over the horizon.
23:13
It could be it's a year, it could be there's no island at all in your direction.
23:17
And so that's the thing that made it always seem 30 years away, which is a long enough period of time you're like, well, anything could be possible.
23:23
So 30 years is not actually a prediction, it's just an admission that they don't know. Well, guess what?
23:29
We're on the island now and we're looking up and there's a mountain we got to go climb on the island. Great.
23:32
We got our hiking gear out.
23:34
We're going up the mountain.
23:35
Uh but that's uh the value of having reached the island is now we know.
23:40
We know we and we know how tall the mountain is.
23:41
We have to go climb to go uh commercialize. >> Love that.
23:43
For our overachieving listeners, can you just give us a brief overview of the other approach, the magnetic approach?
23:47
Just so they understand what's been >> Yeah. Yeah.
23:51
>> So every approach diffusion is is the same fundamentals, which is you're trying to essentially heat up a plasma of fuel, typically isotopes of hydrogen, to some amazing temperature or pressure.
24:04
Uh so it's usually you know, give or take hundreds of millions of degrees and then keep it there for long enough that the uh atoms uh these isotopes start fusing in the helium.
24:12
And the longer you can keep it at those conditions, you know, very intense conditions, and that's where you get all this energy that that starts to get created.
24:21
Um and your goal is to burn as much of your fuel as you can.
24:23
Uh and then there's different ways.
24:25
So the magnetic way, which is actually the the the primary way most of scientists have actually looked at fusion over the decades.
24:32
Uh there's a reason for that.
24:34
I'll tell you in a second.
24:34
Uh the way that they the way they do it is they have this up they create a plasma.
24:37
And again, creating a plasma is not hard.
24:41
Everyone, you know, society is know how to do that.
24:42
You know, those little like devices where you put your hand in the ball and the lightning bolt goes to you.
24:46
Like you can have plasmas pretty much anywhere. That's pretty easy.
24:48
You create your plasma and then you heat it up to these extreme conditions.
24:53
And you use and and the plasma is magnetically charged.
24:55
Uh it's ions and so you've got an uh you've got a magnetic charge to it.
25:00
And so you can use giant magnets and you surround this chamber.
25:03
The chamber looks like a The most common design looks like a donut. It's called a tokamak. Yeah.
25:08
And it looks like a donut.
25:08
And so you've got these round magnets surrounding your chamber creating just the right kind of field to hold the plasma in place while you're heating it up to these absolutely extreme conditions.
25:21
And the idea is if you can heat it up while continuing to hold it.
25:22
Now, the problem they've run into for decades and decades is that a plasma heated to hundreds of millions of degrees is such a chaotic system that it's really hard, it turns out, to hold the plasma in place because the thing is kind of wild.
25:36
And it turns out when your plasma hundreds of millions of degrees, you know, you lose you lose containment of it and it like collides with the wall of your very expensive machine you just built.
25:45
It kind of destroys your machine.
25:47
And so it's a very hard problem in order to contain these plasmas.
25:52
>> to race between magnets and lasers.
25:52
I'm going to be on Caltech poly market. >> Basically.
25:56
Um before we shift gears to talk about your experience at Twilio, just one last one to wrap up on the on this topic.
26:03
What lessons do you bring from Twilio as a software founder to a very, very different space building hardware?
26:09
If Oh, you could argue, you know, there doesn't seem to be a lot of crossover, but I'm curious what you think.
26:17
>> Well, actually there's more than you'd expect.
26:18
First of all, you know, entrepreneurship is entrepreneurship in a lot of ways.
26:21
And so having built a company to billions in revenue and about 10,000 employees at our peak, well, you just learn a lot of things about company building that are fairly universal across, you know, wide variety of companies you can go build.
26:34
And it's really things about like investing in the culture of the company and how do you guide the culture in a way that you want?
26:41
How do you, you know, hire great people?
26:42
And there's a lot of things that are just, you know, process-wise.
26:46
And having seen a lot of talents, you just know what great talent looks like.
26:50
Um and things like that that you can bring over.
26:53
And so a lot of that is like the culture building and the team building stuff that happens.
26:56
The second thing that I'd point to is, you know, I think one of the innovations that has allowed deep tech companies to do some quite amazing things actually over the last, you know, couple decades has been porting in a lot of the learnings from the world of software.
27:15
So if you think about, you know, deep tech endeavors like space travel, they're much more driven by like, you know, nation states and governments and very kind of risk-averse, right?
27:22
In terms of, well, you know, if you make a mistake, you get hauled in front of Congress or Parliament and, you know, people yell at you because there's all these other departments who want the funding and so they they create a show for you so try to get shut you down and then they get your funding.
27:36
Like like there's all these perverse incentives in in a government setting that make it very risk risk-averse despite the fact that some of these endeavors are very risky by their very nature.
27:46
Whereas software companies, because software is so easy to write in the grand scheme of things.
27:51
Like building software is easy.
27:53
>> Yeah, come become easier.
27:55
>> Uh and yeah, it's like and it's and it's relatively low risk.
27:58
The only question is like, um you know, can we build sophisticated projects and can we build things that people are willing to pay us for?
28:05
Well, it turns out software figured out systems to do that well.
28:07
And you know, a lot of it comes down to agility.
28:11
You know, the whole agile system which is, okay, let's go build things iteratively, build things quickly, test, learn, iterate, and iterate your way towards better and better products, better and better solutions.
28:20
We don't have to go, you know, kind of do the whole thing at once, but actually learn, iterate, learn, iterate.
28:26
Like that's the pace that software has evolved to because you can in the world of software.
28:32
It turns out that just like software in hardware the nature of the problem is taking a large amount of ambiguity and trying to distill it down to answerable questions.
28:46
So in software you think of it like, "Oh, if I build software to solve this problem, would anyone pay me for it?" is the thing.
28:52
You know, like it's like anything is buildable.
28:54
The question is, is it worth building?
28:55
That's the fundamental question in the software world.
28:59
In the hardware world, the deep tech world, it's actually the exact opposite.
29:02
Usually, it's very clear that if you succeed at building a thing, society will pay you for it.
29:06
If I can go build cheap, safe, clean, abundant energy, the world, I'm pretty sure, will pay me for it.
29:11
The question though is the opposite, is it possible?
29:15
And it turns out the same tools you can use to go answer that kind of question, such as here are the roadblocks.
29:20
Earlier, I told you, right?
29:20
People said, "Oh, are we able to scale up this laser diode industry?
29:24
Are we able to mass manufacture these targets?" Okay.
29:25
Well, let's boil that down and actually operate in sprints. Okay.
29:28
What is it that we know today that we can make progress on towards solving these open questions?
29:38
Um and and and and and evolve that.
29:38
And then, 2 weeks later in the next sprint, look back and say, "Okay, what did we learn?
29:44
And what do we now know about this or other problems that should change our road map?"
29:48
And the thing that has plagued huge infrastructure-type projects, you know, think about your multi-billion-dollar projects, right?
29:53
Is essentially the Gantt chart.
29:57
>> [laughter] >> The Gantt chart is a shared lie that we all accept for some reason, right?
30:00
At the beginning of some huge decade-long project, you know, people meticulously lay out, "Okay, here's all the steps that have to go in order and the dependencies and all this kind of stuff."
30:11
And you get this huge Gantt chart, which is a fine endeavor to do.
30:15
Like, it's fine to actually make sure you understand all your dependencies and all your steps.
30:18
However, when you start to actually believe that this plan is a real plan, and then get upset or surprised when it doesn't go to plan, that's where you've gone wrong. That's the screw-up.
30:31
And so, what Agile does is to say, "Look, we still know where we're trying to get to over some long horizon.
30:38
We're still trying to build a company with this kind of software.
30:41
We're still trying to build a fusion power plant.
30:43
However, we accept that the path to get there is going to be a jagged line.
30:51
And we are building in the idea that the plan will change.
30:55
We're building it into our execution cadence.
30:57
So that every 2 weeks we pick our head up every sprint, pick our head up and say, "Okay, what changed and how do our priorities now change as a result of it?"
31:04
And constantly attacking the bottleneck.
31:08
Yeah, constantly reevaluating your plans.
31:10
Still with that long-term in in goal in sight.
31:14
But make reevaluating your path and how you're going to get there, make reevaluating it and re-communicating that to the various parties.
31:20
Like build that into the process.
31:22
And so I think the agile software development process brought into deep tech works as well if not even better because of the amount of uncertainty there is in in tackling such a big thing like, you know, bringing a fusion power plant to market. >> I love that.
31:39
I think that's demystified for a lot of people how we've managed to achieve some of the things in hardware over the past, you know, 5 10 years.
31:44
Do you have Do you find that you have to kind of reprogram those engineers that you're bringing on who come from a hardware background?
31:50
Or they've all typically now they've worked at places maybe like SpaceX or Tesla and that's kind of rippled through the industry. >> Or industry.
31:59
>> like, you know, there's a lot of people coming from different places.
32:00
And so I think you know, there is a bit of of education.
32:04
And you know, the common thing I'll tell you the the the the most common stumbling step here is you know, one is you have to explain it that way.
32:13
Because a lot of people are like, "Oh, we're just doing the software thing cuz software people think they know everything." No, it's it's not that.
32:18
There's actually a reason for it.
32:20
It's about how do you deal with ambiguity?
32:21
How do you deal with tremendous amount of unknowns when building anything?
32:25
And then the second thing I find is a real stumbling block is um people buy into it actually.
32:31
They're like, "Oh, okay, this iterative thing and every 2 weeks and we Okay, I get."
32:37
And they lose sight of oh, but wait a minute.
32:39
Don't we have this longer goal that we're trying to achieve?
32:40
[laughter] Like, yeah, we do.
32:42
Right, so we're not just heads down for 2 weeks at a time looking with blinders looking at nothing beyond our 2 week sprint.
32:49
The sprints are in service of quarters and years and a decade and and you have big goals we're trying to achieve and so always connecting the dots between okay, there's a thing we're focusing on for a short period of time that ladders up to bigger and bigger and bigger things we're trying to achieve and so always connecting those dots cuz you know, the funny thing is I've seen that in software, too.
33:09
People would always be like, well, how do we achieve things here that take more than 2 weeks to do because sometimes things are hard and take more than 2 weeks.
33:15
I'm like, yeah, of course.
33:17
That's that's you know, these two concepts are are not in opposition to each other.
33:22
>> So, shifting gears to Twilio, you went on a wild ride with that company. It's huge success.
33:26
It went from 10 to 70 billion dollars during the pandemic and now it's it's back being a public markets darling.
33:32
It's kind of defied the SAS apocalypse over the past couple months.
33:36
Um, yeah, we'd love to hear a bit about your journey and and what why did you choose to leave?
33:39
>> Well, you know, it's interesting that that that being an infrastructure company is like a blessing and a curse. >> Mhm.
33:46
>> Uh, it is a blessing cuz you know, you get the whole picks and shovels notion, right?
33:50
You can power everybody but you don't have to actually go find the gold yourself.
33:53
You just sell picks and shovels to all the people who are finding gold and some of them will, good for them. Some of them won't.
34:00
I still sold them a shovel, right?
34:00
Like that's the nature of being an infrastructure company.
34:04
And so I think at times we fully embraced being an infrastructure company and at other times it was a struggle, right?
34:09
We did see, you know, customers of ours going and creating SAS companies with much higher margins and much higher multiples and you know, feel a little bit of jealousy over you know, over that and feel like, well, it seems like we could be doing that, too.
34:21
We've got this huge customer base and customers often ask us, hey, do you mind if I take these building blocks and just finishing it for us a bit?
34:28
And uh so the allure was always there.
34:33
And I would say that, you know, we did we did we we did build some SAS-like software-looking things.
34:36
Now, we had a flavor of it.
34:38
They were always very API-driven.
34:39
It was very building-blocky, but we did put them together.
34:42
I think what you're seeing now is a real return to the notion of infrastructure companies are very valuable in this era of AI. For two reasons.
34:50
And I and I, you know, I I I think I was ahead of of in in calling these out uh actually several years ago in 2022 and 23, which was, you know, for two reasons.
35:02
One, infrastructure companies win when the world is building. Yeah. Right.
35:06
So, in, you know, 2010 when the mobile boom was just starting and Twilio was, you know, a 2-year-old company, that was a great tailwind for us because people are building everything and figuring out, okay, what are all the things we can do on these fancy gadgets we now have in our pockets. That was amazing.
35:22
Uh during COVID, you know, you mentioned that we got a great bump during COVID. Why?
35:26
Cuz the world had to go build to go solve some really hard problems with a the first pandemic of its kind.
35:32
And so, we saw an acceleration of our revenue in the first, you know, whatever 3 6 months because every company had to go figure out new solutions to things that broke because of the pandemic.
35:42
And now with the AI revolution, the world is building yet again.
35:46
There's a whole new slate of startups and incumbents and everyone's trying to figure out, okay, what is the AI um uh what does AI mean for our business and what do we have to go build to uh not just survive but thrive in the new AI era.
35:59
And that's the amazing thing about infrastructure companies.
36:01
The only thing that that's bad for infrastructure companies is when the world stops building.
36:05
So, for example, during the downturn that happened in 2022 and 2023, uh that was bad for our business because, you know, a lot of companies um you know, the economy was not doing well, and interest rates went through the roof, and so people stopped building.
36:21
There were a lot of layoffs, and you know, people culling projects and shutting things down and all that kind of stuff.
36:26
So, that was a bad period for uh for Twilio.
36:29
The other thing by the other interesting trend that that I I actually predicted is I think SaaS is going to die. >> You do? You're one of those?
36:37
You think it's completely dying? All of it?
36:42
>> It Look, >> Most of it. >> Sure.
36:43
If I'm on a podcast and I want a good quote, yes, SaaS is completely dead. >> Thank [laughter] you.
36:48
>> Okay, end the clip for social media.
36:50
Now, the real answer is no, it's not going to die, of course, but it will stop growing and may even shrink.
36:53
The reason why is Uh and certainly market caps will be affected, I I think, a lot because of this.
37:00
And it's not the reason people give.
37:01
Some people say, "Oh, I can just vibe code a solution to a problem."
37:03
I I don't actually think that's the reason why SaaS is going to die.
37:07
I I still think that, you know, people want a vendor and they want a throat to choke and all sorts of things like that.
37:13
Yeah, there'll be some amount of like, "Oh, I don't need to pay all this money and if I only need a small number of the features, I can just go have my people vibe code something ourselves, assuming vibe coding gets to the reliability, security, and scalability that we need," which I believe it will.
37:26
Um but I think the real reason is actually something different.
37:31
It is the fact that any business based on seats >> Yeah.
37:37
>> has a real innovator's dilemma. >> Yeah.
37:39
The business model is >> And it's like the the obvious first answer is like, "Oh, well, if I can lay off a third of my sales team and get just as good of results cuz I don't need you whatever BDRs anymore and those BDRs don't need a Salesforce licenses anymore."
37:51
Okay, well, that's the immediate impact which you can see and that's a negative, you know, that's a big headwind to SaaS company.
37:58
Uh but the second is when those companies realize that, and they'll they're all smart, they'll realize this and they say, "Oh, well, we got to go build a bunch of AI things."
38:09
The thing is they will have the innovator's dilemma.
38:11
They will hesitate to build the AI capabilities with the pricing model that customers demand and therefore they will be beaten in this AI era by companies who are pure play, who have no legacy to defend.
38:24
And I'll give you an example, right?
38:26
If I am Salesforce, I would say, "Hey, of course my customers want AI.
38:31
Of course I have to introduce AI into my products.
38:35
I'm introducing AI that is going to make your sales BDRs 20% more productive. Right?
38:42
And I'm going to charge you for this product and and call it a win and say, "Everyone wants more productive BDRs and AI is now making them All these AI features I added to the product."
38:50
Well, it turns out you don't need the 20% more productive BDRs, you need no BDRs at all.
38:56
And that's where the incumbent will be like, "Well, I I I don't really want to introduce the product that's going to cut 20, you know, cut all this revenue out because I don't I won't get the seats." Or maybe they will.
39:05
Maybe they're smart enough to say, "Ah, I know this is coming anyway." >> Totally agree.
39:08
I think the more horizontal your software platform is and the more seat-based your business model is, the more Yeah, the more screwed you are.
39:13
Um I think the I think the podcast quotable moment, by the way, is is is a a throat to squeeze. That's a good one.
39:18
Is there anything you would do differently as a CEO?
39:20
Were there any lessons or learnings that you had which seem obvious in hindsight, but were kind of, yeah, challenging and formative at the time?
39:30
>> I You know, I mean, a million things, right?
39:32
You know, I think one of those things, now that there is the uh infrastructure era is now rewarding Twilio again, you kind of look back at it and, you know, in many ways we were trying to be something we weren't when we did delve into applications and trying to be SaaS, you know, some of our solutions we were trying to bring to market.
39:50
And I think you know, we kind of knew it, but we also kind of thought, "Okay, well, this is a transformation we have to undergo."
39:56
And great companies always have an act two and an act three.
39:58
Like, great companies don't just sit still with the product they launched with and have that be the forever product.
40:03
And um and so, we always were willing to try new things and evolve, which is the right instinct.
40:10
Um but, the transition from an infrastructure company to a a software or a company was much harder than we thought it was going to be for a number of reasons.
40:21
Uh and I think I would have now with the benefit of hindsight and not even necessarily without the >> [snorts] >> uh AI era that we're now in, I would have said, "Yeah, but you need to be true to who you are."
40:32
And so maybe there's a way that is more Twilio to actually address some of these problems our customers want us to do.
40:40
But do it in a way that feels more natural to us.
40:41
It's almost like being a teenager and like, you know, like going through your goth era.
40:45
Like, "Okay, like, you know, I'm all goth now."
40:47
It turns out no, you aren't.
40:51
>> [laughter] >> And I think that's a little bit what it felt like at Twilio. >> Fantastic.
40:54
Well, Jeff, I know you call yourself a technologist, but you have a an awesome grasp of technological historical trends and some really amazing insights you just shared with us there.
41:01
We ask this question of every guest who comes on the show.
41:05
What is it about Jeff Lawson that has enabled you to have so much success?
41:10
Uh I think so a lot Well, a lot of people don't know about me.
41:13
The people assume I have a computer science degree, which I do, and I also have a film degree.
41:19
And film is about storytelling.
41:23
And I think that I've always really enjoyed the art of trying to figure out how do I tell a story to bring an audience along with what I am trying to say, whether that's when I'm fundraising, whether it's when I'm trying to hire employees, when I'm trying to close a customer, um or being on a podcast, right?
41:38
Storytelling is this fundamentally human thing that we do to try to build shared understanding. >> I love that. I love that.
41:44
As someone who's uh a musician uh and spend my life speaking to entrepreneurs, I think the collision of of skills and knowledge and uh is is really where a lot of innovation happens.
41:54
Jeff, this has been awesome.
41:56
I feel like we could keep going on, but unfortunately we have to wrap it.
41:59
But thank you so much for joining us on Giant Ideas, and I'm very, very excited for Fusion uh and for you to bring this vision to life.
42:07
>> Well, thank you very much.
42:07
It is uh great to be on the show, and I appreciate the opportunity.