Quinn Bolton: Yeah, no. It does. Thank you, Subodh. And I guess just sort of a technology question for you. Obviously, I think your decision to prioritize improving gate fidelity makes a lot of sense, rather than just trying to build a device with the greatest number of qubits. Just my question is as you focus on improving qubit fidelity for Ankaa to 98% this year and 99% in 2024, do you think you’ll be able to maintain that kind of two-qubit gate fidelity, as you start to tile these Ankaa devices together? I know you’ve shown the ability to take two tiles and place them together in Aspen, but that was at lower fidelity levels. And I’m just wondering if you’ve done any internal work today trying to link Ankaa to maintain that higher 98%, 99% fidelity as you begin to tile these devices together, because I think that would be obviously the next key to scaling the technology.
Subodh Kulkarni : Terrific question, Quinn. Certainly, we know as you said, how to tile 240-qubit chips at lower fidelity without losing any performance. We have some basic work done, which shows that we should be able to do the same with Ankaa with high fidelity, but that’s mostly at on a much smaller scale and simulation kind of work, because we don’t have a Ankaa type chip at 98% or 99% fidelity today. All indications are that the tiling approach that we did for Apsen will work for Ankaa and we will not see any deterioration in performance. But we absolutely have to prove that next year once we start tiling to our four Ankaa chips at 99% 2-Q fidelity. So there is some risk here. But we feel that risk is manageable, and we feel that we have a good handle on that risk.
Quinn Bolton: Understood. Thank you.
Subodh Kulkarni : Thank you, Quinn.
Operator: Thank you. One moment, please. Our next question comes from the line of Krish Sankar of Cowen. Your line is open.
Unidentified Analyst: Hey, guys. This is Eddie for Chris. Can you hear me?
Subodh Kulkarni : Yes, we can.
Unidentified Analyst: Yeah. Hey, congrats on the new role. Nvidia at GTC Conference last week talked about the GPU for error correction and mentioned that it’s working with you guys, among other quantum computer companies. Can you walk us through how the error correction process works? And what innings are we, and how GPUs can accelerate error correction mechanism for quantum computers?
Subodh Kulkarni : Certainly, at a high level, both error mitigation and error correction are extremely important in the final computational performance. And we are looking at both those areas, albeit not as much as the fundamental chip performance and mix layer in the stack, which is a control software and application software. Having said that, there’s a lot of good work going on within the superconducting quantum computing companies in error corrections from other companies. And they are doing some pretty good work to show what exactly needs to happen with error correction. I will say that all of us are critically dependent on high fidelity to begin with. So error correction today, when the median fidelities of our chips are sub-98% is good work.
Nothing wrong with it. But it doesn’t really pay any dividends to be doing error corrections, when you are less than 98%, median 2-Q fidelities. To make error corrections work, you really need to be north of 99% median 2-Q fidelity, and ideally north of 99.5% or even higher before error correction makes sense. David is here with me, our CTO. David, do you have any thoughts about error corrections more than what I just said?
David Rivas: The only thing I would add is that for error correction to take place, you need your underlying control systems to be capable of being programmed both to take measurements, our circuits are executing, but also to do what are fairly high performance computations, in tandem with them to evaluate the syndromes that are being produced on the system. So one of the things that Rigetti has invested in early on is a highly capable control system capable of being augmented in such a way. This is something that while it’s not unique in the industry, it is — there’s only a couple of us that are actually doing it.
Unidentified Analyst: Interesting, interesting. Thanks for that. And another one, when you stack more than four tiles together, does it get exponentially difficult to increase fidelity rates? Or is it like the most difficult part, is the early stages. And once you establish that the technology work, it gets a bit easier, or maybe it doesn’t get as hard as during the initial stages.
