Analysis

March 29, 2023

On the quest for qubits: Different types of quantum startups, explained

A run down of some of the technologies quantum startups are using to build


Steph Bailey

9 min read

With hundreds of European startups working on quantum and plenty of fresh cash being injected into the sector, the race to build a fully fledged large-scale quantum computer is on. 

Unlike classical computers that operate on binary bits (0 or 1), quantum computers use qubits — quantum bits — which can exist simultaneously in multiple states, allowing for parallel computations. This can allow quantum computers to potentially calculate problems that even supercomputers can’t handle. 

“A conventional computer is closer to an abacus than to a quantum computer,” says Chris Ballance, cofounder of quantum computing startup Oxford Ionics. 

Advertisement

To build a quantum computer, you can pick anything in principle that obeys quantum mechanics — atoms, protons, electrons. “This is why you see this zoo of different approaches to building the quantum computers,” says John Morton, professor at UCL and CTO of London-based Quantum Motion. 

We’re currently in the noisy intermediate-scale quantum (NISQ) era — which means we have some quantum computing tech, but it’s not yet advanced enough to solve a problem without errors or perform better than a classical computer. Once we go beyond this era, humanity could unlock applications from finance and drug discovery to finding new materials to stall climate change.   

And there are lots of ways to get there. Here are some of the technologies European quantum computing startups are using, and how they work. 

Superconducting qubits

The most mature approach — used by the likes of IBM and Google — is superconducting qubits. At the basic level, a superconducting qubit is a circuit loop made up of metals that become superconducting (ie. able to conduct current when cooled down) with an electric current travelling around it. They use electric currents flowing through them to store and process information. When Google claimed quantum supremacy in 2019, it used a 53-qubit superconducting device, and in 2022, IBM unveiled "Osprey", a 433-qubit superconducting processor. 

In the UK, Oxford Quantum Circuits (OQC) has built an eight-qubit superconducting quantum computer named Lucy. Brian Vlastakis, its quantum R&D lead, says because all of its quantum information is encoded into electrical signals, it can use a lot of the same circuits that are used for other electronics. 

A quantum computer developed by Oxford Quantum Circuits
A quantum computer developed by Oxford Quantum Circuits

The startup has been providing quantum-as-a-service since 2019. Lucy, for example, is available on the cloud (Amazon Braket) for customers “to try out and learn more about how quantum computers could be useful for the problems that they’re trying to solve”.  

Vlastakis says one of the reasons he’s excited about OQC’s technology “is that our architecture is incredibly flexible. We can essentially design many different quantum processor variations to function in a way that will work better for customers.”

Pros:

  • One of the most mature approaches, so lots of R&D has been done already 
  • Uses similar tech and materials to other industries 
  • The qubits have fast operating times, meaning they can calculate problems quicker than other qubits 
  • Fairly straightforward approach (if that’s possible with quantum) 

Cons:

  • Require ultra-low temperatures (near absolute zero) to operate, which is very expensive 
  • Each qubit is slightly different, so needs to be calibrated 
  • That requires a lot of error-correcting qubits or error-correction techniques 

Trapped ion qubits

Another method is using trapped ion technology, which consists of “trapping” single atoms in place using an electromagnetic field. Unlike superconducting qubits, trapped ion qubits are identical to each other.  

Ilyas Khan, cofounder and chief product officer of UK-headquartered Quantinuum, says trapped ion devices offer two advantages — stability and circuit depth — which provide relatively low error rates. However, it’s not clear how scalable the tech will be and the method is slower than superconducting. 

“At the moment there’s no point to being fast if you can’t do anything,” says Khan. 

While Quantinuum (and others such as IonQ and Alpine Quantum Technologies) rely on complex laser systems to control the trapped ions, Oxford Ionics uses a technology that can be integrated into a standard silicon chip. 

Advertisement

Ballance says Oxford Ionics’ focus is on optimising a lower number of qubits with very low error rates, rather than scaling the number of qubits massively.

“Most quantum computers on the market have far more qubits than they can use in useful computation because of the error rate," he says. “So for example, IBM have their 433-qubit devices they’ve launched, but when you benchmark them they perform less good than a perfect nine-qubit system,” he says. “Our focus is getting to those few 100 qubit devices as fast as possible.”  

Pros:

  • More stable than other qubits, offering lower error rates 
  • Can operate at room temperature (or at a cool temperature much easier to maintain than absolute zero)

Cons:

  • Slower than superconducting qubits 
  • Requires a vacuum to “trap” the ions 
  • Unclear how scalable the tech will be  

Silicon based qubits

While superconducting and trapped ion qubits were originally physics experiments in labs, Morton says Quantum Motion has a different approach: silicon-based qubits. 

“We’re ultimately saying that for quantum computers to be useful you’re going to need a lot of qubits. What does a lot mean? Well, hundreds of thousands or millions of qubits,” he says. “There aren’t many technologies that make millions of anything — one example of something that has is the silicon transistor. 

“If you don’t try to correct for errors, then it’s true maybe you can do something useful with just a hundred or a few hundred qubits, but the problem is you still are going to want to be able to run lots of problems, and run them many times, and so you still, in the end, want lots of qubits.”

Quantum Motion's silicon chip
Quantum Motion's silicon chip

The startup hopes the silicon approach will be more scalable and cost-efficient, as it can build quantum processors with far less specialist technology, such as lasers or a high vacuum. Quantum Motion’s approach offers qubit densities that are highly miniaturised and its silicon-based quantum chips are typically a few millimetres across. Morton expects the cooling system required to operate the chip to be similar to a standard 19-inch server rack.

Pros: 

  • Uses less specialist technology so could potentially be more cost-efficient 
  • Can theoretically scale to millions of qubits 

Cons:

  • A relatively new approach to scalability hasn’t been demonstrated 
  • High error rates limit their usefulness 

Photonics

Another approach is photonic qubits, made from particles of light. PsiQuantum, a US company founded in the UK, says photons are the only way to reach a million qubits — and a million qubits is the only way for a quantum computer to be useful. 

“There are many advantages when you decide to use photons, because first of all photons are a quantum particle that have no mass and no charge, so that means that photons are less exposed to disturbance than other kinds of techniques,” says Marine Xech-Gaspa, chief of staff to the CEO of Quandela, a French startup also betting on photonics. “So to be more concrete they can be manipulated at room temperature, because you don’t have to be in a specific environment, also it consumes less energy.”

Quandela's quantum computer MosaiQ
Quandela's quantum computer MosaiQ

Nordic Quantum Computing Group also has the aim of developing a quantum computing platform based on photonic integrated circuits. 

Its focus is two-fold, according to Axel Mustad, its founder and CEO. On the hardware side, it will use quantum dot-based single photon sources, and on the software side it will develop algorithms which can be implemented on photonic hardware — in particular algorithms to solve hard problems in capital markets and financial services, and in energy management and trading.

Pros:

  • Can operate at room temperature and are less sensitive to the environment 
  • Can integrate into existing optical-based infrastructures 
  • Can theoretically scale to millions of qubits 

Cons:

  • A relatively new approach 

The others

Other startups outside of those building hardware are also an important part of the race. 

Steve Brierley is founder and CEO of Cambridge-based Riverlane, which is building an error correction layer (using different qubits types) that different hardware companies can use. 

“We call it an operating system, because operating systems manage complexity for the user,” he says. “This is like an additional fabric that sits on top of the qubits, really removes errors during the computation and it means it can do much longer and ultimately trillions of operations before failure.” 

Bristol-based Phasecraft is working on algorithms to provide to hardware companies. 

“If you want to do something useful you need to have a quantum algorithm to run on that quantum computer, because quantum computers are not just faster computers, you need to think in totally different ways to get the most out of them to do something useful,” says Ashley Montanaro, cofounder of Phasecraft. 

“We’re particularly thinking about near-term quantum computers, so the kind of machines that we have now, or that we might have in the next two to five years” 

Are we close to cracking quantum?

Ultimately at this early stage, it would be impossible to claim one technology is better than another.  

“That would be foolhardy and in fact misleading — we are years away from being able to evidence superiority in any given platform,” says Khan. “If you look at this moment in time and you’re able to magically transport yourself to 2030, it would be a bit like measuring a marathon in its first or second mile.” 

But what he can say with confidence is that “the early signs are that different architectures might lend themselves better to certain tasks in the future”.

“My expectation is that in 10 years, a lot of the dust will have settled, it’ll become very clear and the market structure will have changed from lots of noise and lots of different approaches sprouting up to consolidation and stabilisation of one or two hardware platforms that cut the mustard and a few other hardware platforms that are specialised,” says Ballance. 

And while funding and access to talent stay on quantum founders’ minds, the biggest battle is the sheer scale of the challenge facing quantum computing startups. 

“It’s equivalent to landing on the moon,” says Brierley. ”It’s that kind of scale and ambition and so that’s going to require bringing together lots of different skills and expertise and ideas. I don’t think any one company is going to solve this problem.”

Steph Bailey

Steph Bailey is head of content at Sifted. Follow her on Twitter and LinkedIn