Pod Insights a16z Podcast Quantum Computing: A Primer About this episode One of the key insights that legendary physicist and Nobel Prize laureate Richard Feynman had was that quantum mechanics (the branch of physics that deals with subatomic particles, uncertainty principle, and many other concepts beyond classic physics) is just way too complicated to simulate using traditional computers.
Nature, of course, can handle these complex calculations -- computers however can't do those same calculations (or would take a prohibitively long time and amount of resources to do so). But this isn't just about being able to do more with computers in a faster (or smaller) way: It's about solving problems that we couldn't solve with traditional computers; it's about a difference of kind not just degree.
So what is a quantum computer and "qubits" -- especially as compared to a traditional computer and bits? What is Grover's Algorithm? And besides speed of processing, what are some of the new applications that wouldn't have been possible before? From how traditional computers work and quantum computers will work to why this all matters, a16z Deal and Research team head Frank Chen walks us through the basics of quantum computing in this slide presentation. And even though may feel like you finally understand after watching this, just remember what Feynman once said: "If you think you understand quantum mechanics then you don't understand quantum mechanics."
[for sources, recommended readings, and image credits, please visit a16z.com]
Transcript 0:01 Speaker 0 Hello, this is Frank Chen from the A16C deal and research team. One of the ways that I think about my job is that my team and I are off scouting the future. We're trying to understand what the big breakthroughs right around the corner are that will turn into great startups, great products for you and me. And one of the things that we've found is around quantum computing. So today I'd like to give the scouts report on what's happening in this fascinating field of computer science, material science, mathematics, physics, and it's mind-blowing. So let's get started. So you know what I love? I love it when politicians teach computer science. It's a real revenge of the nerds moment for me to see world leaders talking about the big impacts that computer science is going to have on our society.
0:55 Speaker 0 And so let's watch the prime minister of Canada, Justin Trudeau work some magic. This was from a speech that he gave in April 2016 at the Perimeter Institute for Theoretical Physics in Waterloo. And it's safe to say he won the internet that day.
1:12 Speaker 1 Very simply, normal computers work by. No, no, no. Don't interrupt me. When you walk out of here, you will know more. No, some of you will know far less about quantum computing, but most of you, normal computers work either there's power going through a wire or not. It's one or a zero. They're binary systems. Uh what quantum states allow for is much more complex information to be encoded into a single bit. Regular computer bit is either one or zero on or off. A quantum state can be much more complex than that because as we know, things can be both particle and wave at the same time and the uncertainty around quantum uh states uh allows us to encode more information into a much uh smaller computer.
1:57 Speaker 1 So uh that's what's exciting about quantum computing, and that's
2:01 Speaker 0 what's that great. That speech reminded me a little of when President Obama visited Google in January of 2008, and Eric Schmidt asked him what the most efficient way to sort a million 32-bit integers was. So if you miss that, go look it up on YouTube. Another fun moment when political leaders are teaching computer science. Now, if you listen to the prime minister, he implied that what we were trying to do with quantum computers was make smaller computers. And while he didn't say anything technically inaccurate, I think that misses the point a little. So we're not really trying to make smaller computers. So what are we trying to do with quantum computers? So understand that, let's go back to the beginnings. So I'm gonna pick up the story with this fine-looking fellow.
2:48 Speaker 0 His name happens to be Richard Feynman. Most physicists would consider him one of the top 10 most influential physicists of all time. He helped design the atomic bomb at the Manhattan Project. He laid down standard notation for how quantum interactions work. And turned out to be a great bongo drummer, if you want to look that up on YouTube. And oh, by the way, he was on the commission that figured out why the space shuttle challenger exploded. So Richard had a bigger brain than most people. And he was fascinated by the physics of very, very small objects. So think of these objects as atoms or electrons or photons. And what happens at this very small, small, tiny scale is that nature gets weird. And so because nature gets weird, what does every quantum physicist want to try to understand the nature of nature at this scale?
3:39 Speaker 0 Well, what every quantum physicist wants is a particle accelerator. What a particle accelerator does is lets the quantum physicist run experiments about subatomic and atomic scale objects. It turns out we have one right in our backyard. So our offices at A16Z are in Menlo Park, right on the tip of the Stanford campus. And in our backyard, literally, you will see the rooftop of the Stanford Linear Accelerator, which is one of these ginormous devices designed to accelerate atomic and subatomic particles and collide them into each other. And this is the way that we run physics experiments at this scale. Here you see a picture of an even bigger accelerator. This one's actually round and not linear. This is the large Hadron Collider operated by CERN just outside of Geneva, Switzerland.
4:27 Speaker 0 The large Hadron Collider is 27 kilometers around, and that makes it the world's biggest machine. There is no machine bigger than this. So what is mankind doing with the biggest machine that it's ever made? It's doing quantum physics experiments. So obviously we can't build many of these machines. They're incredibly expensive to build and to operate. They are the ultimate, not in my backyard project. In fact, it took a while to find the site for the LHC as the community calls it. And so if you can't build a lot of these machines, what's your consolation prize as a physicist? So if you don't get a particle accelerator for your birthday, what do you get? Well, what you get is simulation time on a supercomputer. And so much of the experimentation that happens today on quantum physics happens inside supercomputers.
5:16 Speaker 0 So if you don't get to actually run an experiment on the large Hadron Collider or at the Stanford Linear Accelerator, and you can imagine the queue is very long for the experiments that want to be run. What you do instead is you simulate the interactions of these particles in a supercomputer. In fact, if you go out east from our office, eventually you will run into the Lawrence Livermore Berkeley labs. And this is the bulk of what they do. They are simulating the effects of what would happen if we detonated nuclear weapons to see if our nuclear weapons will continue to work and to improve their efficacy. So the United States doesn't actually explode nuclear devices anymore. We signed the comprehensive test ban treaty. And so all of the work that we do now is in simulation.
6:04 Speaker 0 But there's a problem with simulating quantum mechanics. And let me try to give you the intuition behind the problem. The key insight that Richard Feynman had was that quantum mechanics is just way too complicated to simulate using traditional computers. So I'll give you a concrete example. Let's say we wanted to simulate the state of a thousand hydrogen atoms. And in particular, we wanted to simulate the state of a thousand of the electrons in a hydrogen atom. Now the hydrogen atom is one of the simplest we've got. We've got one proton, we've got an electron. And if we wanted to simulate the up or down state of the electron, and we wanted to do that for a thousand of these electrons, we would need a computer that could represent two to the thousandth states.
6:51 Speaker 0 So you could have the possibility that all of the electrons were spinning down. You had the possibility that all thousand were spinning up, and you have all the possible combinations, which works out to two to one thousandths. So even with this very relatively simple example, we would need a computer that could represent more states, two to the one thousandth, then there are atoms in the known universe. So think about that for a while, which is no computer can actually store that state. But the weird thing is somehow nature seems to be doing these calculations by itself. Somehow it's keeping the state represented in a way that leads to consistent quantum mechanical behavior. And so Richard Feynman's insight was hey, nature seems to be doing these calculations.
7:38 Speaker 0 Can we somehow hitch a ride on the quantum mechanical properties of these atoms and get nature to do calculations that we want to do as opposed to calculations nature seems to be doing to keep nature working? So that's the fundamental insight behind the quantum computer. So to boil it down, quantum computers rely on properties or quirks in quantum physics. And the goal of building a quantum computer is to be able to do calculations of a unique sort, and I'll describe that in a little more detail as we go along. So we're trying to hitch a ride on nature to perform computations that using a traditional computer we just couldn't do or would take an incredibly long time. So to go back to Justin Trudeau, where he kind of implied we were trying to build smaller computers.
8:25 Speaker 0 That's not what we're really trying to do. What we're really trying to do is build quantum computers that can solve a class of mathematical problems that we couldn't solve with traditional computers or that we could only solve with way big amounts of memory and would take a long, long, long, long time. So how do quantum computers work? So to understand how quantum computers work, first we need to understand how traditional computers work. So traditional computers work basically, as Justin Trudeau explained correctly, using transistors. And these transistors can either be an on or off state, just like your light switch. Originally we made these transistors out of vacuum tubes, but eventually the geniuses at Bell Labs figured out how to create transistors out of semiconductors.
9:13 Speaker 0 So we have this physical property of a semiconductor, which it can be conducting electricity or not, and we're going to declare that the on or off state. And then we marry that up with logical gates that can perform operations. And in the case of modern computers, we're performing Boolean algebra operations. So not and or. The knot of one is zero. The and of one And zero is zero. And basically, if you extrapolate from these two building blocks, the transistor, which encodes on or off state, these logical operations that are performing Boolean algebra applications, that deles directly to the smartphone that's sitting in your parking. So all of traditional computing basically is on a linear path straight down these two fundamental building blocks.
9:59 Speaker 0 So the fundamental physical property again that we're taking advantage of is a transistor that is either conducting electricity or not. And the logical operations we're performing are Boolean algebra operations. By contrast with quantum computers, we're using a different part of nature. And in particular, we're either using atoms or photons or electrons. Let's lay aside the implementation detail of what we're using for now. And let's sort of try to understand the what of what we're doing. So in contrast to traditional computing bits where a transistor represents an on or off state, what we have with quantum computers is a quantum bit. So when observed, the quantum bit behaves exactly like a traditional bit, which is to say it can be zero or one.
10:45 Speaker 0 So it can be an electron in up or down spade. It can be an electron that's negatively charged or not. But a crazy quantum mechanics property is that when it's not observed, this qubit is actually representing the probability that it's zero or one. So for those of you who remember Schrdinger's cat, your physics teacher told you that inside this box is a cat, and it is either alive or dead. And when you pop open the lid of the box, the cat will definitely be alive or definitely be dead. But if you close the lid, the cat is in some indeterminate state. It is a probability somewhere between alive and dead. And this was one of the weird quantum mechanical properties of various, very small protocols like atoms or electrons. We're going to take advantage of this by having the qubit and the probability that it's either zero or one to encode information.
11:37 Speaker 0 And so the way to think of it without getting into the math is think of the qubit as representing an arrow or a vector. So this vector has a direction and it has a magnitude, right? So picture the arrow is pointing in a direction, that's its direction, and it has a length. That's the magnitude of the arrow. And so fundamentally, the qubit represents this arrow rather than an on or off state. And then the analog to the operations we're performing. So instead of performing Boolean algebra operations, in quantum computers, what we're using is a set of quantum gates and they implement quantum operations. And the way to think about this is as linear algebra operations. So for those of you that remember your linear algebra, you have a vector, think of it as an arrow with a direction and a length.
12:26 Speaker 0 We apply a linear algebra operator on it. And that fundamentally changes the arrow. It might change its direction, it might change its magnitude. But basically think of it as an arrow goes in to a quantum operation and an arrow comes out. Just a different arrow, but mathematically related to the arrow that came in. So the fundamental physical property that we're going to take advantage of is a quantum mechanical property rather than an electrical conductivity property. And then the operations we're performing aren't Boolean algebra operations, they are linear algebra applications. So that's what's happening at the core of quantum computers. Let me put this together by describing how a quantum computer would solve a problem and how that's different than a traditional computer.
13:13 Speaker 0 And let's use the so-called phone book problem. So the phone book problem is this. You're asked to find a particular phone number, but you're given a phone book where the phone numbers are stored by, let's say, last name. And so if you're trying to find a specific phone number with a traditional computer, what you would do is look at the first entry and say, hey, is this the phone number I'm looking for? If not, go on to the second entry. Is this the phone number I'm looking for? If not, go on to the third entry, and so on and so forth. And if you wanted to guarantee that the phone number existed or didn't exist in this big phone book, you would have to iterate through all of the entries. There is no mathematical shortcut that you can take.
13:59 Speaker 0 There's no algorithm, in other words, that would get you more efficiency than iterating through all of the entries in the phone book. And if you had a very large phone book, this could end up taking a long time. In contrast with a quantum algorithm solving this problem, and in this particular case, the quantum algorithm was designed by Love Grover, working at Bell Labs at the time. It's going to take advantage of quantum mechanical properties to solve this problem much, much, much faster than iterating through all the entries. So let's see how that works. So remember how I said that qubits can be thought of as vectors, which is uh they're representing information that's uh in the shape of an arrow with a direction and a magnitude.
14:44 Speaker 0 What quantum computers are doing is basically passing a lot of these vectors through these quantum gates, performing linear algebra operations on them, which change the direction or the size of the arrow. And essentially the intuition behind Grover's algorithm is there's a vector representing the correct phone number. And what we want to do through these quantum operations is to lengthen that vector to get the right answers vector to be bigger and bigger and bigger, and to get the vectors representing the wrong answers to be smaller and smaller and smaller until we get to the point where we can reliably detect the difference between the big arrow representing the correct answer and the smaller arrows representing all of the incorrect answers.
15:30 Speaker 0 And so basically the magic is as you sort of sweep the arrows through these quantum gates, that's exactly what these quantum operations are doing is they're lengthening the length of the right arrow. They're shrinking the length of the wrong arrow until we get to the point where we can reliably detect the difference between the right answer and the wrong answer. That is the arrow with big magnitude versus the arrow with small magnitudes. The visualization you're seeing shows you the quantum operations as they're performed on these arrows, and it comes from this fabulous design studio called Twisted Oak Studios. And I encourage you to go visit their website for more insight behind how this actually works. So here is the big payoff by doing it the quantum way as opposed to the traditional way.
16:19 Speaker 0 The traditional way we'd have to iterate through all of the entries in the phone book, and depending on the size of the phone book and how long it takes us to iterate through one particular entry, this could take a very, very long time, and there's no mathematical shortcut. In contrast with the quantum computers, we only have to perform a small number of operations, and it's proportional to the square root of the size of the entries. And so this table shows you the number of operations the quantum computer would have to do in contrast to the traditional computer. And the advantage grows with the number of numbers to search, right? So if we have a million numbers that we're trying to search, the traditional computer would have to do a million iterations, whereas the quantum computer would only have to do a little over 30,000.
17:10 Speaker 0 And so here is the fundamental thing that we're trying to do with quantum computers. We're not trying to build smaller computers. We're trying to build computers that can solve a specific class of problems much, much, much faster
17:24 Speaker 2 than traditional computers.
17:26 Speaker 0 So if you don't quite understand all of that intuition and the vectors and the size of the vectors, don't worry about it. Quantum mechanics is so counterintuitive that Richard Feynman famously said, if you think you understand quantum mechanics, then you don't really understand quantum mechanics. So don't beat yourself up. I'll recommend some reading for you if you are interested in following up and understanding more of this, and we'll put that with the blog post. Now that sounds great. So you're probably thinking to yourself, wow, if there really are a class of problems that can be solved using this class of computers, then why aren't we all running quantum computers? Well, it turns out that quantum computers are incredibly hard to build.
18:12 Speaker 0 And let me give you some fascinating stats that give you the intuition behind why these are so hard to build. So it turns out that most quantum computers are made out of superconducting materials. And the reason for that is you need to minimize the interaction between the quantum computer and the outside world. It's very easy for the outside word to disturb the results that you're getting from quantum computers. And so one way to minimize it is to run them with superconducting material. So most quantum computers are cooled to 0.1 degrees Kelvin to minimize their interactions with the outside world. So that is a very, very powerful refrigerator. Just to give you the intuition, interstellar space is 2.7 Kelvin. So we need computers that are colder than interstellar space in order to minimize the interactions with the outside world.
19:02 Speaker 0 Another thing that makes these hard to build is this thing called the coherence time. So because quantum computers interact so readily with the outside world, we only have a small amount of time in which we can perform all of our calculations, and that's called the coherence time. Researchers have made a lot of progress on raising the length of the coherence time. So we're up to state of the art these days is 100 microseconds. But even 100 microseconds is not a lot of time to perform operations in before the quantum coherence time makes it impossible for us to get a result out of the quantum computer. Another thing that makes these hard is I showed you the quantum operations that are performed, and each of those operations takes 50 nanoseconds.
19:48 Speaker 0 So we only have a limited amount of time if each operation takes 50 nanoseconds and our quantum coherence time is 100 microseconds, we only have a certain amount of time. We can only perform so many operations before we can't get a reliable result. And then one last fun stat, I talked about how what we were trying to do with Grove's algorithm was lengthen the arrow enough so that we could tell it from all of the incorrect answers. The length of the correct answer was bigger. So if you took the energy difference between the correct answer and the wrong answer, the energy difference is only a difference of 10 to the negative 24th joules. So it's incredibly difficult to detect the answers. And so hopefully this give you some intuition for why quantum computers are so hard to build.
20:35 Speaker 0 The good news is that researchers are making a lot of progress on getting the quantum coherence time up, inventing more algorithms. In particular, the quantum coherence time seems to be progressing very, very quickly these days. And in fact, faster than Moore's law. And so we're pretty optimistic that at least on the quantum coherence time, we're on a path to longer and longer coherence times. So, wow, hard to build a quantum computer. What could we do if we had quantum computers? So, one of the things that you could do with quantum computers is that you could train so-called deep learning networks much faster. In another podcast, I talked about the history of AI and machine learning and deep learning and highlighted that deep learning is one of the most profound techniques that researchers are using to create artificially intelligence programs.
21:27 Speaker 0 And if we could train these deep networks faster, we could get much better deep learning results. So that's one thing that you could get from a quantum computer. Another thing that you can get from a quantum computer is finding the prime factors of very, very large numbers. Now, everybody likes to say that you've got a supercomputer in your pocket, so why don't you just take your supercomputer out of your pocket right now, take that number that's on this slide and quickly give me the prime factors of that number. No takers. Okay. That still is a very, very hard mathematical problem, despite the fact that you have a supercomputer in your pocket. And it's good news that it is a hard problem, even with that supercomputer because all modern encryption depends on this fact that while it's relatively straightforward for a computer to multiply two prime factors to get a very large number.
22:20 Speaker 0 It is very difficult for a computer to divide that number to recover its original prime factors. In fact, this mathematical property is what's behind all of modern day encryption. So if we had a quantum computer, we could dramatically reduce the time it took to recover the prime factors, and all of modern crypto that we're using today would instantly be susceptible to attack because it wouldn't take a long long time to decrypt somebody's messages. You could do it very, very quickly. This is why the NSA and other state researchers are actively investigating how to use quantum computers and new algorithms to solve this particular mathematical problem. So what else can we do with quantum computers? Well, one intriguing area of research is called quantum chemistry.
23:07 Speaker 0 And the basic idea here is we're going to try to use quantum computers to perform calculations that would help us find new chemical catalysts. So let me give you a prime example of a chemical reaction that would benefit from a new catalyst. And that's called the Haber Bosch process. So the Haber Bosch process today creates ammonia from natural gas. And it turns out we need ammonia for fertilizer. So scientists regard the Haber Bosch process as the literal detonator of the population explosion. So we had 1.6 billion people living on the planet in 1900, and we have over 7 billion today. And a lot of that is credited with finding enough food for 7 billion people. And a lot of that is credited to fertilizer. Now, the process that creates ammonia from natural gas happens at very high temperature, four to 600 degrees Celsius, and at many times the pressure of the atmosphere.
24:00 Speaker 0 And as a result, it's very energy intensive. Scientists estimate that between one and 2% of all of the energy generated in the world today goes to making ammonia from natural gas so we can make fertilizer. And this process is so important that 80% of the nitrogen in your very tissues comes from this process. And so if we were able to find a catalyst that would allow us to either lower the atmospheric pressure or lower the temperature at which this chemical reaction takes place, we could produce ammonia much, much more cheaply and feed even more people. Let me give you another example of quantum chemistry. So the picture you see here is a picture of the Tesla gigafactory, which will start producing lithium ion batteries in 2017 to put into the fabulous Tesla electric cars that everybody wants.
24:48 Speaker 0 When it's in full production in 2020, this gigafactory will produce more batteries each year than were produced in the entire world cumulatively through 2013. So very, very important factory for the future of electric cars and for Tesla in particular. Now, the chemistry that these batteries depend on, which involve lithium ions was discovered in 1980 by an American physicist named Professor John Goodenough at the University of Texas in Austin. And since that time, the energy density, in other words, the amount of energy we can cram into a given volume of batteries has been progressing at a rate of about 3% a year. Now, that seems like good progress, but of course, in the computer industry, we're used to things progressing much, much faster.
25:33 Speaker 0 We are used to things like Moore's Law, where we're doubling the transistor count every 18 to 24 months in our chips. And so 3% a year is pretty slow progress, really. And so we're hoping for a breakthrough. So research scientists around the world are racing to discover new battery chemistry, a breakthrough in batteries that would allow us to store either much more energy so that we can drive our electric cars much further. So that's another very important application that we could apply quantum computers to. So why pay attention now? Well, the reason that we are paying attention now is that a lot of the pieces to making a quantum computer are really coming together. I mentioned how difficult it was, but research scientists are making progress on all of the key ingredients: the hardware, the software and the developer tools that enable people to write software and the algorithms that we're going to use.
26:28 Speaker 0 And the rate of progress is astounding. One of the things that I wished I could have done, but I wasn't born in the right decade for this, is to be a fly on the wall inside Bell Labs when all of the key ingredients for computing as we know today came into being. So the transistor itself, semiconductors, information theory, the C programming language, the Unix operating system. You also had Grace Hopper outside of Bell Labs with compilers. So all of this intellectual ferment that has led to us being able to use the supercomputers in our pockets, it would have just been so fun to listen in on the conversations as they were happening. And basically that same ferment, that same intellectual ferment is happening right now around quantum computing.
27:14 Speaker 0 So in the university labs, in the corporate RD centers, and at startups, there's just so much excitement that right now we should be able to put all the pieces together to learn from the lessons of the past and build these breakthrough computers that can perform these calculations that are mysteriously hitching a ride on the calculations that nature seems to be doing itself. So with all the key ingredients coming together, it's no surprise that some of the biggest university labs, corporate RD centers, and even state-sponsored research is being directed at solving the quantum computing problem. So on this slide, you see the budgets, the team sizes and the investments that well-known organizations are making. And this is really to us a sign that now is the time to start looking for startups that can participate in this ecosystem and help bring quantum computing to life.
28:07 Speaker 0 And in fact, keep your eyes on this space, and we'll have announcements to come about companies that we have found that do exactly that. In the meantime, just marvel at the fact that we're going to try to build these classic computers that take advantage of quantum mechanical properties of atoms and subatomic particles, somehow getting them to perform calculations for us that they're doing by themselves in nature already. Be well.
Previous a16z Podcast: Lobbying Tech Next The Promise of AI Page updated August 25, 2026 · transcript and summary are machine-generated