Welinq: Shaping the Future of Modular Quantum Computing

Quantum computing is getting a lot of attention because of its revolutionary approach to processing information. However, the challenge in realizing its potential lies not just in computation but in scaling. 

Integrating more qubits into a single quantum processor gets increasingly difficult beyond a few thousand qubits. But what if you could connect several quantum processors into one large-scale, multicore quantum computer with enough qubits to solve practical problems? 

Welinq was founded in 2022 by Tom Darras, Julien Laurat, and Eleni Diamanti to build highly efficient quantum links that will allow connecting quantum processors end-to-end, paving the way to scalable quantum computing. Using laser-cooled neutral-atom technology, they’ve set a world record for storing and retrieving quantum states—a key milestone for building highly efficient quantum links. In 2023, Welinq raised a €5M round led by Quantonation and joined by Runa Capital.

Learn more about the future of modular quantum computing from our interview with the co-founder and CEO, Tom Darras: 

Why Did You Start Welinq?

I am personally very much excited by challenges. If I set myself an ambitious goal, I will do everything I can to achieve it. 

After my Master’s in quantum physics, I began my PhD researching quantum teleportation and networking at the Laboratoire Kaster-Brossel. Despite the COVID-19 lockdown, Julien’s research group conducted complex experiments, achieving a world record for the performance of quantum memory in 2020, while I focused on hybrid quantum networking as part of my PhD.

We quickly figured that this had the potential not just to be an interesting research project but also to solve one of the great challenges the quantum industry faces today: the scaling of quantum computers. Despite the short period of my PhD and on top of the COVID lockdown, we managed to achieve outstanding results. This gave us confidence that we could deal with all the challenges that would come with running a startup. 

As an academic, I had not worked in startups before, so this was an entirely new environment. But I talked to many other founders who had started the likes of Alice and Bob, Quandela, or C12, and I attended an entrepreneurial training program.

Finally, three years ago, we began building Welinq. My PhD supervisor, Julien, and our close partner from LIP6 in Sorbonne Université, Eleni, two exceptional scientists with outstanding track records, and I joined forces as co-founders. Together, we set out to tackle what we see as the key obstacle toward developing useful quantum computers: building efficient quantum interconnects to scale quantum computers. 

Why Is Scaling Quantum Computers So Difficult? 

When you talk to people in the quantum industry or look at the roadmap of QPU providers, you will figure that we can’t put arbitrarily many qubits in a single quantum processor, and we are going to hit a ceiling of around a few thousand qubits.

For example, for superconducting or silicon spin qubits, microwave pulses can be used to implement gate operations, but scaling beyond a certain number of qubits poses significant challenges, requiring photonic interconnects. For quantum error correction and solving commercially relevant problems, millions of qubits are needed, involving photonic interconnects across various platforms—including superconducting, silicon spin, photonic, ion, and neutral atom qubits. That’s where Welinq comes into play. 

By taking several smaller-scale quantum processors and networking them with our highly efficient photonic quantum links, we’ll be able to build modular, multi-core quantum computers, paving the way for quantum data centers that will be powerful enough to solve commercially relevant problems. 

It’s similar to how classical computing developed from individual transistors to processors and, eventually, multicore processors for high-performance computing. Yet, while it’s now relatively straightforward to connect classical computers with copper wires and fiber optics, quantum networking is hard and requires specific technologies to be developed—such as quantum memories.

How Do You Build Quantum Links from Quantum Memories?

The problem of interconnecting quantum processors comes down to quantum processors having to emit a photon, which serves as a qubit itself and is entangled with the processor’s quantum state. The next step is to make the photons emitted by two different quantum processors interfere to create a quantum link that entangles the quantum states of the two. 

The first challenge is to address a wide range of QPU architectures and platforms (neutral atoms, trapped ions, photonics, superconductors, spin qubits).

The second challenge is that photons emitted from different quantum processors need to interfere with each other. After the interference, they are detected, which destroys them.  However, it is also this measurement process, using quantum teleportation, that establishes entanglement between the quantum processors’ qubits. Yet, photons don’t interact with each other, so it’s very unlikely to happen by chance. It is a bit like trying to have two rifle bullets collide in midair. 

But instead of bullets traveling at the speed of sound, we have photons traveling at the speed of light, and this interference process needs to happen not just once or twice but eventually tens or hundreds of times. 

That’s why we need to facilitate the process with our quantum memories. Photons can be lost or scattered during transmission and must meet at the right moment. Quantum memories act as a temporary storage device, preserving their quantum states and releasing them just at the right moment so they can interfere.

Quantum memories positioned at intermediate stations form quantum links, creating entanglement between quantum states without a specific notion of sender or receiver. By interspacing quantum networks with quantum links, stopping photons, and remitting them just at the right time, we can improve the entanglement rate and the network’s performance by several orders of magnitude. 

How Do Quantum Memories Work?

A quantum state, e.g., a photon’s polarization, carries quantum information, but it is fragile, so a quantum memory must preserve it without measurement or collapse. It must capture and freeze the quantum information by transferring it to a physical medium, i.e., its quantum properties, like spin states of electrons or nuclear states, need to adjust to match the incoming signal and thereby encode the quantum state. 

The encoded quantum state remains frozen in the medium, isolated from external disturbances, until the stored quantum state is read out by reversing the process: A retrieval signal interacts with the medium, releasing the stored quantum information, ideally as a photon with the original quantum state. 

There are many ways to build a quantum memory, depending on the physical medium. For example, one can use hot vapors of Rubidium atoms, ions in doped crystals, or color centers in diamonds. Yet the issue with most of these is that the probability of storing and retrieving the original quantum state of the photon, and thus the efficiency of the quantum memory, is very low, like 5-10% for hot atom vapors and 20-30% for color centers in diamonds, for on-demand storage. 

We also use rubidium atoms but cool them down with a magnetic field and laser. Using these laser-cooled neutral atoms, we built the world’s most performant quantum memory. We have set a world record by achieving a 90% recovery probability. Given that the efficiency of quantum memories scales exponentially, this improvement represents orders of magnitude in potential entanglement rates. 

What’s The Opportunity for Welinq?

Several quantum hardware technologies will emerge simultaneously, and each might be suited for certain use cases. We believe in a future where several qubit modalities are deployed together in a high-performance computing center, and we’re positioning Welinq to be the interconnect between all these different types of qubits. 

Each qubit type has its own physics and engineering challenges. And for each qubit type, we’re building an interface so they can be entangled with the photons in our quantum links. Where we can, we’re working with partners who provide best-in-class solutions, e.g., to interface with superconducting qubits. But we need to build everything else from the ground up as there are no ethernet plugs yet for quantum computers. 

Five years ago, quantum interconnects were like the elephant in the room. People assumed they would have to happen at some point to make quantum computing viable, but they didn’t know exactly how and didn’t say it out loud. Now, everyone is mentioning it in their roadmaps, and we’re excited to be part of this trend, as our interconnects will enable an entire ecosystem to be built on top of it.

Quantum links can bridge both short distances within a quantum computer for multi-core quantum computing but also enable long-distance quantum networks between quantum computers or quantum data centers and eventually even build a quantum internet. 

They are two complementary markets, and while we see quantum memories as an enabler of quantum networking in the short term, we will then be able to address the bigger opportunity of multicore quantum computers and solve commercially relevant problems. 

What’s On Your Roadmap?

We have developed our core invention, the efficient quantum memory, into a product that can fit into a 19-inch rack with an integrated vacuum system and electronics. So, it’s not an experiment anymore; it’s a product and will be ready by the end of this year. 

We’ll be able to generate the first revenue by selling these quantum memories, as pairs of them can be used to build quantum links and enable quantum communications between quantum computers. We have ongoing partnerships with Pasqal for neutral atom qubits, with Quandela for photonic qubits, and another one for superconducting qubits. 

We’re now working on the interfaces of this quantum memory with different types of qubits, like photonic and superconducting qubits, with the goal of building a product that works end-to-end. In less than two years, we want to build such a full-fledged interconnect product.

We’re also working on algorithm development. As we create interconnects, understanding how to run quantum algorithms on a multicore architecture will become crucial. That’s why we’re focusing on both software and hardware. Our goal is to provide a complete solution, enabling users to run quantum algorithms seamlessly on multicore quantum computers.

Finally, we have started exploring use cases through the AQADOC project with EDF, Pasqal, and Quandela to advance energy transition-related use cases. Mapping out these use cases helps us understand the requirements for software and hardware to be ultimately able to address them.

We’ll have a team of 25 full-time employees by the end of this year, and we’ll scale it to 50 full-time employees until the end of 2026 and open an office in North America to expand to the US market. 

What Advice Would You Give Fellow Deep Tech Founders?

Building a startup is life-changing, and you will learn many things along the way about developing a technology into a product, scaling a team, and building a company. 

There are three things I would say to other founders in particular: 

First, go for it! You will learn so many things, so there will be no failure—only opportunities to learn something new. 

Second, there’s a pretty good environment to help you build a company—many people will be willing to help you and like to support entrepreneurs. I had initially worried that we would have to fight a lot, but many people were really helpful—don’t hesitate to ask for help!  

Finally, you’re building your own company in the end, so there’s no guidebook—it’s your unique project. Build it based on who you are and your own value, which is important so it lasts for the long term and founders stay motivated to keep going. Make sure it’s a project that looks like yours.