Quantum Computers: Roadmaps, Features, and Security
Quantum computer hardware companies actively publish development roadmaps with timelines for upcoming features. This newsletter article begins the new year by reviewing existing roadmaps from major quantum computer hardware companies. The article especially focuses on new upcoming features, and how they may impact the security of quantum computers. Currently, security or security features are not listed in the roadmaps, but upcoming hardware or software features such as chiplet-like designs can give some insights into potential ways the quantum computers will be used, and in turn this gives ideas about their possible security implications.
Scaling of Quantum Computers Continues to Increase
This newsletter article looks in particular at roadmaps from IBM and Rigetti, but many other quantum computer hardware vendors exist. In the roadmaps from these two companies, we can see continued increase in scaling of the sizes of quantum computers.
In the latest IBM roadmap, shows continued commitment to increased size of the quantum computers. Starting with 27 qubits in 2019, IBM has by now announced 433 qubit machine. And the roadmap predicts 1,121 qubit machine in 2023. This is where scaling of an individual quantum computer chip ends, and the plans seem to switch towards design of computers made of many smaller chips. This is very much the chiplet design style from classical computers. There seem to be two design path. First, “processors made from multiple chips”. Second, processor utilizing “quantum communication links” to connect individual processors. The first processor made from multiple chips is planned for 2024. In parallel, first processors made using the quantum communication links are planned for 2023/2024. The ultimate target in IBM’s roadmap is the Kookaburra processor combining the two technologies of multiple chips plus quantum communication links. The target for Kookaburra is 2025. All of these designs are of course for non-error corrected quantum computers. But the goals stated by IBM target scaling to 100,000 qubits and beyond around 2026, which could enable machines that can support multiple logical qubits built from the plethora of physical qubits.
In comparison, the latest Rigetti roadmap shows that the company also aims for 1000 qubit machines in 2025/2026, and 4,000+ in the following year. This may be a little lower target, but in the same range. One interesting point that can be found in Rigetti roadmap, is the “multi-fridge” design. This hints at quantum computer processing unit designs that can span multiple fridges. This will likely be achieved by classical communication between the fridges, but nevertheless could make it easy to scale the size easily by adding more fridges.
Many other quantum computing companies also exist. A useful summary of some of the roadmaps from the different companies can be found on the Quantum Computer Roadmaps page. Many of the companies, regardless of the technology, are focusing on increasing the quantum computer size.
Potential Paths to Multi-tenant Quantum Computers
While not explicitly called out in the roadmaps, one feature to be on the lookout for is whether and when (or if) the upcoming computers will support multi-tenancy. Multi-tenancy can offer ability to run programs or circuits from multiple users in parallel. The obvious disadvantage at the current stage is that the quantum computers today, even with single program or circuit executing at a time, are too small for practical computation. However, with superconducting machines from Google projected at 1,000,000 physical qubits by 2029, or photonics-based PsiQuantum projected at 1,000,000 physical qubits already in 2025, the limitation of today’s machines may soon be gone.
One interesting approach to multi-tenancy with chiplet-like designs in IBM, for example, would be to dedicate one or more multiple chip processors to a program or circuit. The inter-chip communication may be a bottleneck, so while adding quantum communication links for inter-chip communication may help achieve bigger processors, it is not clear how the inter-chip communication will affect the performance and fidelity of the different algorithms. On the other hand, within a single chip, the performance and fidelity should be better; of course at the cost of fewer available chips. Thus for users requiring fewer physical qubits, dedicating a chip per program or circuit seems a natural choice. Similarly, for Rigetti machines, if multi-fridge designs are realized, then having one fridge per program or circuit seems natural choice. Connecting multiple fridges can help realize bigger quantum computer, but limitation of the inter-fridge communication is not clear. Within a fridge, however, the performance and fidelity should be better. In both of the examples, the design of the control logic may be simpler and easier, e.g., each program or circuit would be controlled by classical controllers dedicated to each chip (IBM) or fridge (Rigetti).
Security Implications of the Roadmaps
Considering security, it can be easily observed that none of the roadmaps discuss any security protections or defenses. Today, most of people running programs or circuits on these machines may be benign researchers and academics. But as the computers increase in size and capabilities, they will start to produce novel, valuable results. Once there is valuable information or data processed by these machines, then there will be incentives for attackers to try to compromise the machines. In parallel, already there may be incentives to reverse-engineer the designs or architectures, given the high cost of these machines.
The reverse-engineering threat exists regardless of size, and is likely something that is valuable to prevent already today. Meanwhile, the threat of stealing valuable data will increase as quantum computer sizes increase, and actual, useful computations are performed. Especially, potential multi-tenancy may make the threats worst.
As performance is main target of the companies, developing security solutions early can be valuable to learn their performance overheads. Just as people are using small quantum computers available today to learn about them, to prototype algorithms, etc., small quantum computers could be used to test security protections. If for any threat protection hardware changes are needed, then it is especially important to consider them now. The roadmaps show designs and plans up to 2026 or later. This implies the companies already may have internal plans and designs for even future hardware. If that is case, then hardware to 2026 or beyond may not have any security protections, according to the roadmaps.
Besides performance, if multi-tenancy is enabled, that will open up a new set of threats. Clearly specifying plans for multi-tenancy can give insights into how users should think about protecting their computation. If suddenly their programs or circuits may start to be run in parallel with other users, then users will have to scramble to assess the security and possibly modify their programs. Knowing ahead when (or if) shared quantum computers will be enabled gives time for researchers and users to prepare for potential new threats and develop defenses.
About the author:
Jakub Szefer is an Associate Professor of Electrical Engineering at Yale University where he leads the Computer Architecture and Security Laboratory (CASLAB). His research interests broadly encompass computer architecture and hardware security of computing systems, including security of quantum computers and post-quantum cryptography.