Alice & Bob: Shaping the Future of Fault-Tolerant Quantum Computing

Quantum computing attracted millions of investments due to its potential to exponentially speed up certain computations—computations that no classical computer could complete even in a thousand years.

While these extreme speedups are supported by mathematical proofs, they require quantum computers resistant to errors that inevitably come up in the delicate quantum computation process. However, despite decades of research, the race is still on among dozens of quantum hardware startups to solve this challenge. Most focus on making error-prone, noisy quantum computers useful in the short term. This “brute-force” approach works by adding lots of noisy qubits, making error correction an engineering nightmare. But what if one could design inherently fault-tolerant qubits? 

Enter Alice & Bob. Founded in 2020 by Théau Peronnin and Raphaël Lescanne, the startup explores a special kind of qubit called “cat qubit”, named after Schrödinger’s cat, that combines two “mirrored” quantum states to build a qubit that is protected from errors by design—making error correction an easier task and a fault-tolerant quantum computer an attainable goal. The company raised a €3M seed pre-incorporation and another €27M in series A funding in 2022, led by Elaia, BpiFrance Digital Venture Fund, and Supernova Invest and joined by Breega.

Learn more about the future of fault-tolerant quantum computers from our interview with the co-founder and CEO, Théau Peronnin: 

Why Did You Start Alice & Bob?

For a long time, I knew that I was meant to do physics. I studied physics at the École Polytechnique and then joined the group of Nobel Prize winner Serge Haroche to study the control of single photons, the particles that make up light. My main question was how the classical world could emerge from something as weird and fascinating as quantum physics. 

At the same time, I was also fascinated by electronics. But the microchip revolution had already occurred, so I turned my attention to quantum computing. This field was at the intersection of the two, and is still very early, on the same level as the invention of the first transistor for a microchip.

So, in 2015, I joined a group of brilliant physicists at École Normale Supérieure who were doing quite exotic research in quantum computing. Many research groups globally were trying to build logical qubits, the fundamental processing unit of a quantum computer, but it was, and still is, tough since none has succeeded so far. Our approach was to consider random access memory (RAM): DRAM had been dynamically stabilized, and static RAM was stable by design, so what if we could design qubits that were stabilized by design? 

Cat qubits remained an academic curiosity until we built the first quantum device from this special type of qubits that showed an exponential suppression of one type of quantum error, bit-flips. Researchers at the National Institute for Research in Digital Science and Technology, Inria, had demonstrated that it could be used to build a quantum computer. So in 2020, we realized that we had a talented team and a promising qubit technology, and we had to do something about it! So my co-founder and I left academia to found Alice & Bob, named after the fictional characters typically used as placeholders when discussing cryptography. 

How Do Cat Qubits Work?

First, our general philosophy is that you cannot cheat nature. If you want to have exponential speedups, you need a quantum device that is capable of correcting errors. Current quantum computers with a few dozen qubits and lots of errors won’t get us anywhere close. There is no mathematical proof of a computational advantage for noisy-intermediate scale quantum (NISQ) computers, the type of quantum computer currently being developed. But there is for fault-tolerant quantum computers (FTQC). 

Quantum error correction is tough for two reasons. First, it only works if each additional qubit you add for quantum error correction prevents more errors than it introduces. If your qubits are bad to begin with, they’ll add more noise, and thus errors, than they can correct for. So you need to meet a basic threshold, which is about a 1% chance of making an error per quantum operation, so that when you add more qubits, things get better. Only a few players, like Google, have demonstrated that they can reach this threshold. 

The second challenge for quantum computing is that there are two possible types of errors. The first is the bit flip, which also exists in classical computing, where a bit flips from 0 to 1 and vice versa. The second—a purely quantum error that has no classical equivalent—is the phase flip error, where a superposition of qubit states flips from 0 + 1 to 0 – 1 and vice versa.

In classical computing, you correct bit flips by keeping copies of the bits to check if a minority of these copies has a different value than the rest and promptly correct them. A similar strategy for qubits requires a 2D array of copies since you need to account for the two possible errors, as implemented by the surface code, the most established quantum error-correction code everyone uses. 

Even if you solve the first challenge and meet the error threshold, you still need to add quadratically more qubits to produce an additional error-corrected logical qubit. For example, if you want to use Shor’s algorithm to break RSA encryption, you’ll need about 20 million qubits, where only 10,000 qubits are used for computing, and most are needed for error correction. That’s why many physicists are so skeptical of quantum computing: quantum error correction is just tough.

Our realization was that we can’t brute-force our way through this by just adding more qubits. We need a fundamentally better qubit, one that has a feedback loop and automatically regulates itself. So we explored cat qubits! 

Cat qubits are built from a superposition of two coherent states with opposite phases and represent a paradigm shift. While others are trying to isolate qubits to avoid interference from the environment that might disturb a qubit’s fragile quantum state, we want our qubits to interact with their environment in a specific way that dissipates energy to mitigate noise, reducing entropy without losing quantum information. 

Cat qubits are stable by design, and as we demonstrated in a Nature Physics publication in February 2020, the probability of bit-flip errors gets exponentially suppressed.  Bit flips take minutes before occurring—an enormously long time in the quantum world. Of course, there is no free lunch, as you get linearly more phase-flip errors. But it’s an exponential improvement for a linear cost. Remember the 2D surface code? Now you don’t have to care for bit flips anymore; you effectively just need a linear array of copies for quantum error correction. The raison d’être of cat qubits is to provide a new generation of bosonic qubits that, by design, only have one type of error and a more forgiving error threshold. 

Our processors have demonstrated to be resistant to bit flips for 100 seconds in 2020 and this metric today has improved to several minutes. We have a great chance of becoming the first to develop the first logical qubit by next year. This would be a huge proof point for our approach, ultimately allowing us to have a lot fewer qubits for error correction. In the example above, running Shor’s algorithm would require 300,000 cat qubits instead of 20 million regular qubits. 

We’re confident that with our qubit architecture, we’ll be able to address several initial use cases. If you get the error rate low and a few 1000s logical qubits, you can already do some nice chemistry. Eventually, you will address more applications, such as Monte Carlo simulations, once you get to tens of thousands of logical qubits and error rates of one in 100 million or even one in a billion. 

How Did You Evaluate Your Startup Idea?

We knew the big picture and came up with a strategy to get there. We go after proven use cases, built on one of the most established hardware platforms—superconducting circuits— and use standard control electronics, fridges and equipment as most other quantum computers. This will allow our platform to be highly compatible and readily adoptable. Our core focus is on developing the best possible qubit design, and having in-house control over development, fabrication, and characterization for short feedback loops enables us to do that efficiently. 

We measure progress by very concrete milestones, such as fabricating the first logical qubit. It will be a Sputnik moment, and we took quite some time to make this clear to our investors. When a qubit has escaped decoherence, it will no longer have a causal link to the noisy outside world and truly be quantum. This will be a similar milestone to Sputnik escaping gravity into space. 

Quantum computing is a tough bet for investors. Many generalist investors venture as tourists into quantum computing and waste founder’s time. They don’t have the right macroeconomic point of view or the ability to assess the technology, talent, or IP of an early-stage quantum startup.

Growing as a quantum startup in Europe is challenging, as very few funds can lead a 100M series B for a pre-revenue deep tech startup. But with only €30M in funding, we already showed that we could make B2B sales and get to first revenue and contend on an even round with hyperscalers like AWS. We also provide a consulting service to connect with clients and help them navigate the hype, put numbers on a use case, and understand what is proven, what has very good reasons to function, and what is just quantum snake oil.

What Advice Would You Give Fellow Deep Tech Founders?

I was a good student and listened to a lot of advice, but at the end of the day, you build your own business, not someone else’s. Being a physicist helps to get an intuition for science, but it’s also important to build intuition for entrepreneurship, for example, by reading about entrepreneurial success stories of the past. Most challenges come down to hiring the right people, so it’s up to you to build a strong and ambitious team culture, as we did.

Work on the Future of Computing at Alice & Bob

Check out opportunities on Alice & Bob’s career page*