A quantum-computing startup announced Tuesday that its future quantum processor designs will differ significantly from its current process. Rather than building a monolithic processor as everyone else has, Rigetti Computing will build smaller collections of qubits on chips that can be physically linked together into a single functional processor. This isn’t multiprocessing so much as modular chip design.
The move is consequential for both Rigetti processors and quantum computing more generally.
What’s holding things back
Rigetti’s computers rely on a technology called a “transmon,” which is based on a superconducting wire loop linked to a resonator. It’s the same qubit technology used by large competitors like Google and IBM. The state of one transmon can influence that of its neighbors during calculations, an essential feature of quantum computing. To an extent, the topology of connections among transmon qubits is a key contributor to the machine’s computational power.
(This is in contrast to things like Honeywell’s ion-trap computer, in which any qubit can interact with any other, at least at the current qubit count.)
Two other factors that hold back performance are the error rate of individual qubits and the qubit count. Scaling up the qubit count can boost the computational power of a processor—but only if all the added qubits are of sufficiently high quality that the error rate doesn’t limit the ability to perform accurate computations.
Once qubit counts reach the thousands, error correction becomes possible, which changes the process significantly. At the moment, we’re stuck with less than 100 qubits, so this is change is still coming in the indefinite future.
What changes for Rigetti
As a startup, Rigetti doesn’t have access to the sort of resources available to companies like IBM. And while Rigetti has done an impressive job of fabricating its own transmon processors, it has still lagged behind its larger competitors. The qubit count in IBM’s latest processors is over 60, while Rigetti’s latest is 31.
That hasn’t been a major limitation at a time when we’re still unsure if useful calculations can be performed on quantum processors without enough qubits for error correction. At the same time, a clear path for rapid scaling is critical for getting to the point where error correction is possible, and we may find that some algorithms can be run effectively on qubit counts between the ones available and the ones needed for full error correction.
For Rigetti, the ability to merge several smaller processors—which the company has already shown it can produce—into a single larger one should let it run up its qubit count rapidly. In today’s announcement, the company said it expects that an 80-qubit processor will be available within the next few months. (For context, IBM’s roadmap includes plans for a 127-qubit processor sometime this year.)
Another advantage of moving away from a monolithic design is that most chips tend to have one or more qubits that are either defective or have an unacceptably high error rate. By going with a modular design, the company can reduce the consequences of that reality. Rigetti can manufacture a large collection of modules and assemble chips from those with the fewest defects.
Alternately, the company could select for the modules that have qubits with low error rates and build the equivalent of an all-star processor. The reduced error rate could offset the impact of a lower qubit count.
The bigger picture
As things stand, qubit technologies require components that are much larger than anything found in a computer chip. Optical quantum computers and transmons both need optical cabling, while the atoms in ion-trap computers are held in space by an array of electrodes. As a result, scaling the number of qubits isn’t anywhere near as simple as adding computational resources to a traditional silicon chip—the physical constraints are just too different.
As a result, other quantum-computing companies we’ve talked to have already acknowledged that they will have to find ways to integrate qubits on more than one chip. So Rigetti may have solved a problem that other companies are likely to face.
But transmons, which can be linked by wiring, are likely to be one of the easier technologies to work with in this regard. And the lessons learned here won’t apply to competing technologies like trapped ions.