The world’s first quantum computer to exceed 1000 qubits has more than double that of the previous record holder, IBM’s Osprey machine, which has 433 qubits. Though having more qubits doesn’t necessarily mean better performance, large numbers of them will be needed for future error-free quantum computers that are useful, unlike today’s noise-filled research machines.
The largest quantum computers, such as those from IBM and Google, use superconducting wires cooled to extremely low temperatures for their quantum bits, or qubits. But the record-breaking machine from California-based start-up Atom Computing, which has 1180 qubits, uses neutral atoms trapped by lasers in a 2-dimensional grid.
One advantage of this design is that it is easy to scale up the system and add many more qubits into the grid, says Rob Hays, CEO of Atom Computing. Any useful quantum computer in the future that is free of errors, a feature called fault tolerance, will need at least tens of thousands of dedicated error-correcting qubits working alongside the programmable qubits, he says.
“If we’re only going to scale by dozens of qubits, like most of the trapped ion and superconducting systems have been scaling up until now, it’s going to take a very long time to get to the fault tolerant era,” says Hays. “With the neutral atom approach and the speed of scaling that we have, we will be able to get there much more quickly.” Hays says the team aims to multiply the amount of qubits in the machine by around 10 every couple of years or so.
Unlike conventional computing bits, which can have a value of 1 or 0 and are largely interchangeable, qubits are more varied, having a range of different properties depending on how they are made.
Neutral atom qubits lend themselves better to quantum entanglement, a strange quantum effect where qubits are linked so that measuring a property of one qubit reveals that of the other. They are also more stable, with qubits in Atom Computing’s machine keeping their quantum state from collapsing – a feature called fault tolerance, which is essential for error correction – for almost a minute. IBM’s Osprey, for example, has coherence times of around 70 to 80 microseconds.
These long coherence times are due to the ytterbium atoms that Hays and his team use as qubits. Most neutral atom machines use an atom’s electron as the quantum element with which to do computing, but this can be easily affected by the powerful lasers used to hold it in place. With ytterbium, a quantum property of the atom’s nucleus called spin can be used, which is much less sensitive to disturbances. “The nucleus just doesn’t interact with the outside environment as strongly as the electron does,” says Ben Bloom at Atom Computing.
Because qubits have so many different features, it can be difficult to compare across different machines, but Bloom says Atom Computing’s machine is comparable in processing ability to IBM’s, though the company has yet to release figures on this.
The team hopes to offer the machine to customers next year for cloud computing applications, similar to what companies like IBM do today. “Atom Computing’s machine can’t currently perform computing operations on all the qubits at the same time, which will be required for fully error corrected machines,” says Bloom.
“There’s multiple groups now building systems that will have 1000, and even several thousand, atomic qubits,” says Mark Saffman at the University of Wisconsin-Madison. “This is really where the frontier of the field is now, with this 1000-plus scale that people are developing.”
However, more details of how the machine works will need to be released by Atom Computing before it can be properly assessed, says Saffman, such as how many of its qubits can be used and have logical operations performed on them.