Chad Rigetti at Startup School SV 2016

Everybody, our next speaker is Chad Retti from Retti Quantum Computing. Retti Quantum Computing went through YC in the summer of 2014. Um, at that point they had nothing. Uh, they are now one of the leading Quantum Computing companies in the world. And next year, well, I don't know exactly when, they are getting close to a Quantum Supremacy machine.

I have a particular love for the startups where they're trying to do something it's not clear if it's technically possible. But if they do, uh, it changes the world. And it just goes, nothing, nothing, nothing, nothing, and then everything is different. Uh, these companies are super fun to work with, uh, and I think they're super fun to work on, which is why I'm so excited Chad's here to tell us about Retti and also hard technology startups in general.

Um, and why it's okay if you don't have a growth graph every week. Uh, sometimes you just work and work and work, and then everything comes together. Um, Chad, come out here because I have a question for you. So Chad and I met on the streets of San Francisco, uh, walking around the financial district in 2014. Chad had not yet started his company but was thinking about it. Chad and I have actually never spoken about this again, but I think, I hope you remember this.

Um, I think Chad thought I was really crazy because we got into a long conversation about this simulation hypothesis and if building a quantum computer was actually really bad because it would use a lot of resources, uh, to emulate and shut down the simulation. Did you think I was really crazy?

It was a very interesting conversation. Is that like a really nice? You can say yes. I didn't think you were crazy, but I had just moved to San Francisco into Silicon Valley from New York City, and that is not a conversation you can have in New York City without people thinking you're crazy. Uh, but I'm an open person, and I, you know, I like to learn, so I was very interested in this hypothesis, and it was an amazing conversation. It was very fun.

All right, well, I hope the company works, and I hope the simulation does not shut down. Um, great. Thank you, that will help. All right, so Chad is going to talk to us about Retti. Thank you for coming, Chad.

Awesome! Thank you so much, Sam. All right, I'm incredibly excited to be here today and to talk with you guys. Um, I'm also incredibly honored, so thank you very much, Sam and Cat and YC for having me. Uh, giving me the opportunity to talk with you guys.

So this, uh, as Sam mentioned, when we joined Y Combinator in summer 2014, we had nothing. Uh, we had not built a single qubit, the fundamental, like, building block of a computer. We did not know how we were going to build qubits.

Uh, I knew what they needed to look like, and I knew, uh, what the fundamental requirements were. In fact, I'd been working in the field for already about 10 or 12 years in Quantum Computing. Um, and, uh, this is a 5-qubit quantum computer that we have built at Retti Computer.

Uh, it's less than 2 years later, and, uh, we're about 35 people in a warehouse in Berkeley. Uh, about 20 some PhDs, and we're building core computers. And we're in a race, uh, to define this technology for the next 20, 30, 40 years, and it's incredibly fun.

It is a kind of thing that gets you up in the morning and is worth spending your life on. Humans are building. Humans have been building tools and technology to store and process information for millennia. Uh, I like to think of the sundial as the original computer.

It takes an input and it does something useful with it and tells you the time. Uh, from the sundial to the abacus, uh, and then things like the Babbage Difference Engine, which you can go see at the Computer History Museum. It's amazing, and you should do it.

Um, to punch cards. Those are all systems based on Newtonian mechanics, and every time our understanding of nature evolves and develops further, we have a deeper level of understanding. Our technology progresses as well. So, uh, many great companies were born of this first transition that I'm talking about on the slide.

Uh, IBM started as a Computing Tabulating and Recording Company. They made punch cards and time clocks. Uh, they were kind of born out of this first transition into the microchip era. And they, you know, they blossomed and evolved and survived through this time to grow into a very, very large company. Other companies that drove that first transition, uh, who was it? Fairchild, Intel. And then what's been born in the era of the microchip? Well, almost everything we see around us.

It's driven the global economy for the past 50 or 60 years. That's incredible. The amount of leverage you can get from a computing technology is massive. And you know what? We've known since 1905 that beneath, uh, there's Newton's laws on one hand, and then beneath that there's kind of Maxwell's equations.

And we've known since 1905 that there's a fundamental description of nature, and that's quantum mechanics. Uh, at Retti Computing, we're driving this next transition from microchips to those systems based on individual atoms or artificial atoms that we build in the lab. Uh, and that that's a very exciting opportunity for us.

So why is it worth doing this? This sounds incredibly hard. Why is it worth, uh, trying to build quantum computers in the first place? Well, the kind of problems that these machines will solve are incredibly impactful. They fall into two broad categories.

We think of them in, in kind of two broad categories. On the one hand, you have a class of applications that we think of as quantum chemistry. So what you're doing in this case is using the quantum mechanical, the quantum computer as a tool to investigate other quantum mechanical systems.

It turns out that that's almost everything, uh, in science, uh, medicine, material science. This is stuff that's going to lead to better drugs, uh, better materials, better nuclear reactors—highly impactful technology. Why are we doing this? Because if you can, you build a quantum computer, you can do simulation-driven design of new catalysts, catalysts to capture carbon or nitrogen from the atmosphere.

It can help solve global warming. This is an incredible, incredibly broad technology. Uh, the other bucket of applications, uh, is machine learning and artificial intelligence. So, in this case, what you do is learn to embed a learning problem on a classically intractable physical model that you can simulate on a quantum computer.

That's a mouthful. What it means is quantum computers are going to lead to fundamentally more powerful forms of artificial intelligence. So when I say more powerful, what do I mean?

Uh, does anyone recognize this H? It's on the slide. So this is Tianhe-2. This, until about three months ago, was the most powerful computer on the planet. It cost about $400 million; it burns about 20 megawatts of electricity. You guys know how much electricity that is? It's enough to power about 20,000 households. It's about half the size of a football field.

It's based on 3.2 million Intel cores. So there's two problems with this. Ultimately, our approach to building large scale computers is starting to break, and there's two problems. The first is that Moore's law is ending. That's been happening for a long time. Lateral transistor scaling, the fundamental measure of Moore's law, has sort of leveled off for the past eight years.

The other thing is something that I hope you've all heard about in your computer science class is called Amdahl's law. Amdahl's law talks about the limiting benefits or the diminishing returns of parallelization. These are massively parallel machines, 3.2 million cores running in parallel. Only problems that can be parallelized run fast on these kinds of machines.

Uh, there's a fundamentally better way. Now, President Obama has shown some significant vision in, uh, sort of driving, uh, a return of American leadership in high-performance computing. And he has said that America is going to build an exascale computer, something about 30 times more powerful than Tianhe-2 by 2020.

That machine will cost about a billion dollars with current technology, and it would require a nuclear power plant to run it. We're going to do it, and we need to. But there's another path. Ultimately, when you can compute with quantum physics, you have a faster and cheaper path to that level of computing power.

How is that possible? Well, these are two developmental systems in our lab in Berkeley. These are what you see in the big white cans; they are cooling systems. Every high-performance computer has a cooling system. These ones are very powerful; these run at low temperature.

Uh, inside each of those cooling systems is a single chip. When this picture was taken, we had a 5-qubit processor in that machine. A single chip with about 60 or 70 qubits on it would be more powerful than that entire half a football field-sized machine. That's what quantum computing unlocks.

Uh, so this is a true hard tech startup, and one of the challenges that hard tech companies—I'll talk about this more in a moment—brings is this challenge of operating in an arena that is not well-defined. That does not have a well-developed supply chain. You know what you want to do, but the capabilities to do it do not exist or are not commercially or easily accessible.

We've had to develop all of the building blocks, the entire supply chain for this technology. What does that mean for us? We have developed new simulation-driven design methods to actually design these quantum chips. Uh, we had to develop our own fab, our own microfabrication capability.

Uh, we've developed advanced electronics that allow you to control these chips. You can think of a quantum computer as something like a nuclear reactor, where you have the core, the, you know, where the reaction is happening that generates all the power. But then there's this really complex traditional engineering system around it that stabilizes it and extracts a computational capability.

And that thing is very expensive, and it requires very advanced control electronics. And then ultimately, uh, we serve access to these machines over the cloud, and so we have to have cloud software and applications.

And that's a lot of work. There's a lot of different things you got to tie together in one organization to do this. Uh, this slide is frankly a synthesis of the past 15 years of my life. I've been working on this problem my entire adult life.

Uh, I was a junior in college in Saskatchewan, Canada. And really, that's amazing. Where are y'all from? Oh wow, Moja here. Mo! Um, I was a junior at the University of Regina in Saskatchewan, Canada, and a physics major. I was incredibly frustrated. I was very, very frustrated because I didn't understand two basic things that I thought every human on the planet should understand: what's quantum physics and how do computers actually work?

How do they actually store, process, represent information? About the same time, I read about a field called quantum computing. I like to synthesize things. I thought this was amazing. There's one field that I can learn instead of having to learn two. This makes life so much easier. I've been working on that ever since, and that was in 2001.

Uh, I heard about these people at Yale University that had an idea for building quantum computers. Uh, you can build these things out of real individual atoms or ions. That's really hard because individual atoms or ions are extremely tiny and very, very hard to control. They said, why don't we build these things out of special electrical circuits based on superconductors that have no dissipation and build them in such a manner that they mimic real atoms, build an artificial atom out of an electrical circuit.

I like that, that's amazing. That's such a great idea. Because what's going to happen is you're going to be able to leverage all of traditional semiconductor manufacturing capabilities. You're going to be able to build a scalable chip technology someday based on those superconducting qubits.

I spent about seven years with that group at Yale. I was a PhD student at POSTECH. Uh, I spent about three years doing further research at IBM in Quantum Computing, and in 2013, I started Retti Computing to develop quantum integrated circuits, and that's what we've done.

So, this is the jump on my second slide. This is the leap from Newtonian plus Maxwell's equations to Newtonian plus Maxwell's equations plus Schrödinger's equations: quantum physics.

And, uh, ultimately, I want to offer you a working definition. If there's any physics majors in the audience, I want to give you a working definition of a quantum computer. Quantum computers store and process information in individual photons. That's it.

Um, your iPhone, the iPhone 7, uses about 100 billion photons per bit processed. It may be as much as 10 to the 17 and 100 billion photons per bit transmitted. This is far more efficient technology.

Okay, so I want to talk for a moment about the distinction between a hard tech company and a tech company. First of all, when I say hard tech, I don't mean it's harder, although it feels harder. What we mean is every company—and look at the company on this slide—these are incredible organizations, incredible products, incredible founders.

These are all incredibly hard things. There's one distinction a hard tech company has to deal with, in addition to all of the tech execution market risk, all the standard things, all the gauntlet that you have to steer your organization through to survive. Fundamental questions of possibility. You have to deal with that when you're a hard tech company. That is the signature of a hard tech organization.

Now, there's two things I want you to take away from this. The first is that, well, with all of that risk, we're saying it might not even be possible. Why the hell are you going to do that? Why the hell is anyone going to invest in your company if it might not be possible to even do it? Why is it worth it if you're not sure it's possible?

Well, there's two things you get from my perspective. The first is, if and when you are successful, you create monumental leverage and defensibility. I think of the Manhattan Project as a canonical hard tech organization. Look at the leverage they created; that's incredible.

The Apollo missions, when President Kennedy said in 1961, as a nation, we're going to put an American on the surface of the Moon and return him safely to Earth by the close of this decade, that was an epochal moment for mankind. That's the kind of story that isn't written in quarterly reports and spreadsheets; that's written in hieroglyphs on cave walls.

That's the kind of stuff that stirs the heart. That's what you get when you do a hard tech company, and that is so powerful. I want you to notice something on the slide. Look at these amazing companies. At some point, the defensibility and leverage that hard tech provides and the incredible passion that you engender, uh, by working on problems of that scope and impact leads organizations on the left to pushing into hard tech in order to access those things.

Think of Uber moving into self-driving cars. Think of Google with now 100 seedlings hoping to turn something into a long-term defensibility. Microsoft built software and has started working on quantum computing. That's what you get with a hard tech company.

Now, there's also significant challenges, and I want to talk about three of them for a moment: team communication and integration. So first of all, building an organization that has world experts in XYZ, all of the things that you need to master as an organization, is incredibly hard. It comes with its own set of challenges. You're going to have to interface with the best scientists and engineers on the planet in those various fields.

You're going to hire them. You're going to have to bring them all into one company, and you're going to have to find a way to impedance match them to allow them to talk to each other under the roof of one organization. And that leads into this integration challenge: integrating an organization that does all these disparate things.

We have people on the team. Uh, we have a Rhodes Scholar, uh, who's been doing research in quantum computing for 10 years and is 27 years old. Uh, we have tenured physics faculty. We have someone teaching a course at Caltech who built the communication systems on the Mars rover missions.

We have an incredible organization, and one of the problems we've solved is that we have people who can all have a conversation together because they all have developed a shared language. That level of integration becomes a long-term weapon for your organization.

A huge amount of value can be created by it. Now, communication. We're not doing this because quantum computing is interesting; we're doing this to cure cancer and to solve global warming. How you talk about what you do as an organization is incredibly impactful.

I encourage you to spend a lot of time thinking about it. I did this a few years ago, and I decided that our mission as an organization when there were three people in the company was, "We're on a mission to build the world's most powerful computer." It gives you something to sink your teeth into; it's a little more tangible than photons.

We did this exercise again in the past month, uh, with the team fully built out, well partially built out at this stage, and, uh, we got to the same thing. It's really amazing. Why are we doing it? We're building the most powerful computers in the world to solve humanity's most important and pressing problems. That's a rallying cry that can be very impactful for your organization.

Now the other thing you get with integration is the opportunity to create all this defensibility. So the chip, the quantum integrated circuit, is one part of it that gives you an Intel kind of business for the quantum computing era.

We're not stopping at that because no one is competing at all these different levels today, and we have an opportunity to build moats around the entire thing. The next layer up is to build the system and become a master system integration. That's an IBM-style business.

And then ultimately, to own the software and platform, add in the Microsoft-style business. Ultimately, I want to go full stack and also include the Google-style business of applications and designing new drugs to save people's lives.

All right, so before I move on to this, I want you all to do something for me. I want you to do this one thing for me. I want you to take 10 minutes today after this, at the break, when you go home today, during your meditation, whatever you want to do, I want you to take 10 minutes and look in your heart.

And I want you to ask yourself the question, "What kind of company do I want to join or found? What kind of organization resonates with me?" And when you do that, spend that 10 minutes. When you do that, some small fraction of you—it won't be many—some small fraction of you will say, "You know what? I want to do the thing that stirs my heart. I want to do the thing that calls to me."

That is worth spending your life on. One last thing. So one of the special challenges that a hard tech organization faces is the capabilities to do what you need to do don't exist. If they existed, it wouldn't be a hard tech organization.

So there's this tension that exists in all companies, but is especially accentuated in hard tech. A tension between developing the product, taking your product from concept to market, getting the product built, shipping the product, with the things you need to do to enable that at each successive stage of evolution of company development, building the capabilities and creating organizational clarity.

That's your job as a founder and a leader. What should you work on? Well, I found that there's a useful framework. I thought about this problem for a really long time because I was spending every day trying to balance these two competing tensions.

Ultimately, these are both processes of pumping entropy out of the system. You have an idea for the product. I know what this quantum computer wants to look like. It's all, you know, we, we, but it doesn't exist. It doesn't exist yet. There's all these questions we have to answer.

Think of when Elon started SpaceX. He had an idea for what this rocket was going to look like. There's a gazillion micro decisions that have to be made to actually manifest that thing in the real world. That is a process of pumping entropy out of the vision. Same with company development, creating this organizational clarity, building the capabilities, answering the question: are we going to do our own fab? Are we going to outsource fab? Are we going to partner with IBM to do the fab?

Are we going to—how are we going to do the fab? We have to answer that question. Ultimately, that is a process of pumping entropy out of the system. There's a lot of things that this leads you to. One is it tells you who you should hire. Some people create order and clarity in their wake. They create systems; they execute systems; they reinforce systems; they train other people how to use those systems.

Other people generate entropy. You know what you're looking for? Hire people who pump entropy out of your vision for your organization. It's incredibly powerful.

So this is one of my favorite pictures I've ever seen. This is a picture of the Control Data Corporation 6600 machine. What I love about it is look at how it is. There's wires hanging out of this thing everywhere. You can see the pumps down in the corner. This machine was the—is widely considered the first supercomputer.

The United States blocked the export of one of these things to our allies in France. It was incredibly impactful at a geopolitical level, and who built it? A group of 34 folks in the woods of Wisconsin. 34 people built the world's most powerful computer. I can't even read the memo; it's too powerful.

Ultimately, 34 people out competing a giant behemoth. That is what happens in high-performance computing; that is what happens with a lot of hard tech organizations. And it's an incredible opportunity for you if this pulls at your heartstrings. Thank you very much.