The trick in developing a high performance quantum computer is to create one that combines a high number of qubits with a high quality level in those qubits. IonQ has been working on this for some time and their newly announced 32 qubit machine is a follow-on to their 11 qubit model which is currently online and represents their initial offering through Amazon Web Services (AWS).
Although detailed technical specs are not yet available for this new machine, IonQ did point to a paper recently posted on arXiv that described fault tolerance experiments on a nearly identical system. That paper mentioned that the physical qubits had a T2 decoherence time of 2.25 seconds, a single qubit gate fidelity of 99.8% and two-qubit gate fidelity of 98.5%-99.3%. These values are an improvement over the 11 qubit design that they disclosed in an arXiv paper last year. (The T1 decoherence times in these designs are so long such that they can be ignored because they have no detrimental impact on the calculations.)
Although IonQ has not apparently performed the experimental tests yet, they believe that the combination of more qubits and better qubits will allow them to achieve a quantum volume metric of over 4,000,000. This would imply they could pass the required test procedure specified by Quantum Volume with an array of 22 qubits and a circuit depth of 22 levels. (222 = 4,194,304 to be precise).
However, there are a few things to note with this announcement. Although IonQ is calling this machine “The World’s Most Powerful Quantum Computer” in their press release, it still is not more powerful than a classical computer. Classical computers can simulate quantum algorithms with 40 or 50 qubits with perfect qubit quality. So this new 32 qubit processor would not be able to demonstrate quantum supremacy or quantum advantage over a classical processor. In addition, ion trap designs have relatively slow gate speeds. Although the gate delays in IonQ’s design are not specified in the technical papers, ion trap technology typically has gate delays that are roughly 1,000 times worse than the delays in Google’s superconducting Sycamore chip. So program run times would be much longer although the number of shots (repetitive runs) required might be fewer due to the improved qubit quality.
On the other hand, the IonQ computer does have all-to-all qubit connectivity which can provide a significant advantage in embedding programs on the physical hardware. IonQ disclosed they plan to at least double the number of qubits every year and are already working on the next two generations of machines. We expect that one or both of these future generations would be able to provide quantum advantage versus a classical technology. IonQ plans to make the 32 qubit machine available soon in private beta for their pre-existing customers and partners and then make it available later with their existing cloud partners Amazon Braket and Microsoft Azure Quantum.