Michele Dallari, co-founder and CEO, and Marco Polini, co-founder, of Planckian, are interviewed by Yuval Boger. Michele and Marco talk about their company’s novel superconducting chip architecture, which uses global control and always-on ZZ interactions to drastically reduce wiring complexity and improve scalability. They discuss the advantages of controlling many qubits with just two or three classical lines, their early-stage experimental progress, and the software adaptations required to compile algorithms for their architecture. The conversation also touches on Italy’s growing quantum ecosystem, Planckian’s collaborative approach to chip manufacturing, their excitement about recent advances in pulse engineering to compensate for fabrication imperfections—and much more.
Transcript
Yuval Boger: Hello Michele. Hello Marco. Thank you very much for joining me today.
Marco Polini: Hi.
Michele Dallari: Hello. Hi, glad to be here.
Yuval:So Michele, who are you and what do you do?
Michele:Yeah, so I am Michele Dallari, I’m a co-founder and CEO at Planckian, helping Marco and Vittorio, two other co-founders, shaping business and helping the company grow in the quantum computing space.
Yuval:What does the company do?
Michele:So as a company, we are a quantum computing company. We are developing a new quantum processor architecture, trying to solve some of the most relevant scalability challenges in superconducting computers such as the wiring problem.
Yuval:Is it focused on a particular modality, superconducting qubits for instance?
Michele:Yeah. So thank you for asking. We have a clear mission at Planckian. We want to build our core technology that enables quantum computers to scale to meaningful size by design. And we want to do that with a distinct approach. So rather than introducing new qubit modalities, we are developing our quantum chip based on the most known superconducting qubit design, which is the transmon. However, our architecture fundamentally redefines how quantum information is processed, enabling a significant reduction in control requirements by having multiple qubits controlled by shared control lines and allowing complex, non-trivial multigate operations. So our vision is that by minimizing control complexity and also maximizing algorithmic efficiency, we can build a cornerstone for future large-scale quantum computing platforms.
Yuval:Marco, I’m curious. Sounds like a great idea. Obviously the number of wires in a superconducting architecture seems to be a limiting factor. Why is not everyone doing that?
Marco:Well, that’s a tough question. Interestingly, in the history of quantum computing that’s the way things started. It was a very nice little paper in Science that was published by Seth Lloyd in 1993. And it’s precisely about the general idea, not yet really, practical and pragmatic, but it’s a general idea of doing quantum computation by using global control. So Seth had this idea back in 1993. But then you know, how humankind proceeds, especially physicists, right? We don’t start by understanding complicated things. We want to start building by understanding very, very well the building blocks.
And that’s where, where we started from. So you know, people began to work on a single quantum bit, a single qubit, fabricating that. And then you know, using a bottom up approach, adding a second qubit, then a third, then a fourth and through these, you know, bottom up approaches, you end up at a dead end in our opinion, in the sense that you face this wiring problem that really dramatically hampers scalability. So there are other people working on this, I think with other platforms, not the superconducting platform, but I think the idea is spreading and it will spread. This global control that Michele was talking about will spread pretty quickly.
Yuval:Could you explain, one of you, how global control works?
Marco:Essentially the idea is the following. Imagine you have an ensemble of qubits. Technically, an arbitrary number of qubits, and you can partition them into let’s say, two families. So the qubits are nominally identical, let’s say. But I subdivide this family of qubits into two subgroups or families. All of the qubits in the first subgroup, family A, let’s call it, are controlled by a single external classical control line. And family B, which is the other half of the qubits which are left in the picture above, are controlled by another control line.
Then there are special qubits in the entire group of qubits that are special qubits where you can do single-qubit and two-qubit gates. And we even invented architectures where you can do multi-qubit gates, for example, like single-shot Toffoli gates. With classical external pulses, you can move quantum information around in these architectures and then perform local unitaries. So single-qubit gates, two-qubit gates, or one-shot Toffoli gates, just by using classical control lines. Two classical control lines. That’s more or less the gist of the idea.
Yuval:The system, I would guess, still needs to be cooled to at least millikelvin temperatures for the superconducting hardware to work in the first place, right?
Marco:Right, Right, Absolutely. Yes, yes. That’s precisely the issue. Because as you know, you’re bringing down many, many control lines. So in an ordinary locally-controlled architecture, you have to bring the temperature of the hardware down from room temperature down to let’s say 10 or 15 millikelvin, and you have to bring down two or three control lines per qubit. And that brings a lot of thermal load. So you have to spend a lot of energy trying to keep the temperature of the dilution fridge sufficiently low.
Yuval:Yes, I know. I work for a neutral atom company. And when we explain the issues with controlling superconducting, I usually say, look, imagine if you had a 4K television and you open it up and saw that there are two or three wires going into every pixel. That would be one heck of a heavy and expensive and large television, right?
Marco:Right, right.
Yuval:What Is the status of the technology, your technology right now?
Michele:Yuval, thank you. So we are in the stage where we are fundamentally experimentally testing the building blocks of our chip architecture. So as we come up with a new chip architecture, we are basically adding specific building blocks. I don’t want to give out too much on that, and disclose too much. This is fundamentally related to how you couple the qubits together and how you can scale them in a full chip.
Our near term goal is to prove that this principle can work in our basic chip design and of course in the long term to scale it to more meaningful sizes. So we are basically, yes, in a very preliminary experimental point of view, we are very humble, very, very technical, experimental oriented. So we just don’t think too much about the product as of today, more about derisking the technology and the risk and the science behind that.
Yuval:One of the questions that comes up in superconducting qubits is connectivity. How many qubits are connected to each individual qubit in your design?
Marco:The design requires that each qubit is connected with pairwise interactions to nearest neighbor qubits. It’s two-body interactions of a special type. That’s a distinctive feature of our current architecture in the sense that typical locally-controlled quantum computers based on superconducting qubits rely on swap-type interactions. We instead work with longitudinal ZZ interactions, which is usually something you don’t want in an ordinary architecture. Actually, it’s something that people try to minimize, but it’s something bad that we try to use to our benefit. We want to work in a regime in which these pairwise ZZ interactions are really strong.
Yuval:Could I think of your design like a chessboard in which the black squares are, say, connected to one control line and the white squares are connected to another control line. And, and this way I could perhaps create with two control lines an interaction between any white square and an adjacent black square. Is that a correct way to think about it?
Marco:Partially. Almost everything is correct. Let me put it this way. So the other good thing about our architecture is that we use always-on interactions. We don’t use external classical controls to engineer and tune interactions between qubits. Those are to be native interactions that you produce at the fabrication Level. Our designs maximize the ZZ interaction.
They are there, but the partitioning of the black family and the white family of qubits, that’s perfectly fine. And you use the controls to move quantum information around and carry out logical operations. But the interactions are always-on, they are static interactions.
Yuval:If you think about your initial chip, can you provide some numbers, how many qubits you expect to have there and how many control lines in total will you need?
Marco:We are currently working with the companies that fabricate the chips for us with up to five qubits. But these, I’m referring to the stuff that Michele was talking about, these are de-risking technology. So we are now working with 3 or 5 qubits in total. And, we are in fact emulating global control by local control. But, eventually we will advance our platform by derisking the building blocks. For example, to track the strength of the nearest neighbor interaction that we were talking about earlier.
Yuval:Once the de-risking is done, and let’s assume you try to build a chip with, say, 100 qubits, how many control lines would you expect that to have? Only two control lines?
Marco:So that’s the principle. I’m sorry because I’m a theoretical physicist, so let me discuss principles, and then we can deep dive and see whether there are technology limitations. But in principle, the beauty about this is that if you scale from five qubits to 10 to 100 to a million, in principle, you need only two control lines. Two or three maximum control lines in our design. So that’s the beauty. The number of lines does not scale thermodynamically with increasing the number of qubits, but it’s always order 2 or 3.
Yuval:Every approach has disadvantages. What would you say, Michele, are the disadvantages of the approach that you are pursuing?
Michele:First of all, as a first mover, I will say of course there are disadvantages and challenges. There are some technological disadvantages which are more related to the challenges that we already envision that we need to face. As Marco indicated, in principle this is an architecture that is scalable with a fixed number of control lines. From a technological development point of view, we know this will come with some challenges we need to face. So probably we will need a little bit more control lines than expected in order to get to where our architecture properly works. A second challenge, which we are already solving, is how we can properly engineer the control of this architecture in order to also maximize the computational efficiency of the platform. Then of course there is a business side. We are a newcomer in a space that is very mature, let’s say at least one of the most mature of qubit platforms with big players out there. So the question arises about what kind of potential business model we want to focus on and how we can fit in the large vision of how this industry is maturing.
And this is very interesting for us because we are coming up from a nascent ecosystem, which is Italian, where new stories are maturing and also a new business model can be developed around this technology. So I think that one of our main features is that we don’t work at the qubit design level, but rather at the chip architecture level. So we can leverage some freedom here in this space, which can be explored from a business perspective.
Yuval:Do you expect gate fidelity single qubit, two qubit gate fidelities to be on par with traditional superconducting designs?
Michele:These are our KPIs. So one of their near-term KPIs is to understand how this architecture impacts in terms of gate fidelity for single-qubit gates and 2-qubit gates, or to rephrase it differently, we want to maximize our chip architecture in order to be on par with the state of the art.
Yuval:Does one need to design the software differently to take advantage of your algorithm? So if I have a Qiskit program, does it need to be compiled completely differently to run on your future computer?
Michele:That’s a very good question, Yuval. As I mentioned, we are working at the chip architectural level. This is a very nascent technology. We recognize that this comes with interdependencies of playing with different levels of the stack. And specifically the closest one is the middleware and the firmware. So yes, probably, and I’m sure about that, because we are working already on that. There needs to be some work that has to be done from a compiling point of view, how we can compile high level algorithms in our chip architecture. We are working on that along with the development of a quantum simulator based on our chip architecture. But the goal here is to close the gap and use as much as we can of the software ecosystem that is already being developed, at least in the superconducting community.
Yuval:Can you share some information about the company? When were you founded? How large are you? Where are you based? How is funding for this company?
Michele:We are a little less than 20 people as of today. We launched the company in 2023 with a pre-seed round from many Italian investors with some international interest. We’re growing pretty fast considering where we are based. From 3 to 20 people in less than a year, 80% of which are PhD, which is pretty common in this space, with four nationalities. And this is something I want to acknowledge because we are an Italian-based company, based in Pisa, which has a longstanding tradition and rich ecosystem in terms of research activity in this space. We are the first joint spinoff of two leading institutions there, which are the University of Pisa and Scuola Normale Superiore. So we are as of today based in Italy. We recruited and brought some talent which were abroad back to Italy. And we’ve also been able to recruit from the international pool of talent. So we are particularly satisfied with where we are, where we’re standing today.
Yuval:The company name is Planckian. Any particular reason?
Marco:Well, you know, H Bar, right? Everything we do, everything we do has H Bar. So I don’t see any other good name for a quantum company. So this is the best. I mean, it’s difficult of course to pronounce maybe a little bit, you know, somebody forgets a K and sort of replaces the K with an H and so on. It’s Max Planck. Come on. I mean, you know, you have to honor tradition.
Yuval:But Max Planck was German, I think.
Marco:Right, right. I mean, but you know, fermions we have already used,and Italians are associated with more fundamental stuff, yand Olivetti had already been used for classical computers, so why not borrow it? We are in Europe at the end of the day. So that’s what we – new generations and ourselves – we have to think in terms of being united, too. So nothing bad in borrowing German scientist names.
Yuval:How is the Italian quantum ecosystem? Is it rich? Do you see many other quantum companies or is it primarily focused on academia? What can you share on that respect?
Michele:Italy has historically benefited from a strong scientific ecosystem, but has often struggled to translate research into technology and successful ventures. There are several factors that contributed to this. Some are cultural, some are related with the chronic underinvestment in applied research, which is essential to scientific advances and to prepare for product commercialization. In the last couple of years there have been some interesting ventures, especially in the quantum space, that have emerged, without naming ourselves, but other players that’re coming out with very interesting activities. And we cheer for them as we hope that they cheer for us because it means that the ecosystem is growing. Plus, Italy has worked intensively in the last couple of months to develop a national quantum strategy which was released in February. And there is an open consultation ongoing.
We contributed to it and we’re pleased that much of our suggestions that we as an industry player made there with the government have been accepted. We are coming later to try to close the gap. It’s something that we really push for as we believe that it was critical for the quantum ecosystem to emerge in Italy.
Yuval:How soon can I use your chip?
Michele:The soonest.
Yuval:Any expectation on when it will be ready for people to try out.
Michele:Let’s say that as of today we are focused on desrisking the building blocks and developing what is a basic chip, but it’s not something you want to put in the hand of a customer. Yes, in the next couple of years, our goal is to develop something that we can have people use and experiment with. And for us to gather data about how our architecture performs with the rest of the technology out there. So yeah, this is the basic timeline we can share as of today.
Yuval:You’ve been doing this, at least in the context of this company for the last two years. What have you learned recently? What do you know today that you didn’t know, say six months ago about this technology, about the industry?
Marco:I can answer technically. One particular thing, when you build a solid state architecture, there are lots of issues with fabrication. It’s not like neutral atoms where they are all identical by nature. When you build qubits, the level spacings are a little different, there’re going to be homogeneities in the coupling, nearest neighbor coupling between the qubits. There’re going to be lots of issues. And so we were pretty worried about this when we were developing our ideal calculations. As you said six months ago, we were a little scared about this. But now we have a digital emulator and we know how to solve these problems or compensate, for example, with pulse engineering for these artifacts that are there in real life.
Those are things that we were really worried about six months ago. But now we are very confident. You were asking Michele earlier about gate fidelities, for example. So the digital twin says that these artifacts of homogeneities can be compensated for with optimal control strategies. And we are very happy about that. So we can sleep now.
Yuval:Anything to add to that, Michele, on things that you’ve learned?
Michele:Yeah. To me I was really surprised, amazed how the quantum industry is open and collaborative compared to other industries. I came from a different experience. I worked in the energy sector and after that I worked for the government. I worked in other deep tech sectors. But I’ve been really surprised about how collaborative this ecosystem is. Also from different players that you can see from one side that look like competitors, but in reality they want to build this amazing technology because they are the first user.
They want to be the first user of this technology. So this is something that I learned and I’m constantly amazed by that actually.
Yuval:Where would the chip be manufactured?
Michele:As a newcomer industry we also want to have a different approach to the market. So we try to be as collaborative as possible with other players. As of today we have different partners that are manufacturing our building blocks. And also we design in house. We want to keep the core IP in house. Then we have selected partners that we are working with from the manufacturing point of view. And then we’re going to test either in house or have a different partner do the testing.
We want to accelerate these R&D phases. And this is what we have done to date. Then in the future there are different options we are assessing and considering. Of course we want to keep as much as possible in house in terms of core value. But at the same time we also want to collaborate as much as possible with partners because we think that is also an opportunity for us to have our technology tested and assessed by industry.
Yuval:And last, I wanted to ask you a hypothetical. If you could have dinner with one of the quantum greats, dead or alive, who would that person be?
Michele:Oh, that’s very super. Yeah, it’s a good question. I think maybe you allow me to name someone who is not related to quantum. To me it’s Thomas Edison as the paragon of an inventor. He’s really someone who I can constantly be inspired by.
Yuval:How about you, Marco?
Marco:But definitely a scientist. I mean, how about Max Planck? Given our name. No, it’s a very good question. I can think of many people of the caliber of, say, Peter Shor.
It would be very nice to have to have a discussion. I would be very happy to do that.
Yuval:Very good. Michele and Marco, thank you so much for joining me today.
Michele:Thank you, Yuval.
Marco:Thank you, Yuval.