One of the challenges facing the development of quantum computers is how to scale them up in size. While some modalities propose large, monolithic designs, others propose interconnecting and/or networking, smaller modules together. The idea is that these modules will be more immediately achievable, that the concepts of interconnects and networks are already well known, and that any kind of monolithic designs won’t be realized in the near future. The potential benefits of modularity include:
- As we note in our article titled “Top Applications Of Quantum Computing for Enterprises,” solving classically-intractable problems will require large, fault-tolerant quantum computers (FTQC), which will most likely be built from modular designs
- As also noted in that article, the ability to solve larger graphs and more complex optimization problems could be as conceptually simple as adding another module
- The cost of a quantum computer can be based on the number of qubits needed; there won’t have to be any need to either develop or purchase quantum computer excess capacity
- Realizing quantum computing advantages in speed and precision will likely require more qubits than will be available in any one quantum processor
- As noted in IDST’s article titled “Unleashing the Power of Modular Quantum Computers: Towards Large-Scale Programmable Quantum Computing and the Quantum Internet,” modules have value beyond computation, and may serve as nodes in quantum networks
- As also noted in the IDST article, modules may be able to function as memories within quantum networks, as well as in quantum computers
- Thanks to quantum mechanics, the transmission of data between networked modules would be cryptographically secure
While the realization of interconnected and networked modules is still quite some time away, it is important to note that research into both is well underway. Neutral atom quantum computers actually straddle a line between both monolithic and modular designs, with the ability to scale single traps up to 10,000 atoms while also having the potential to interconnect and network processors together.
What is Modular Quantum Computing
A modular quantum computer can be thought of in a similar way to a supercomputer. To increase classical computational power, additional processors and memory are connected to the cluster. If you have a maxed-out 10,000-atom trap and you need more atoms, following the supercomputer model, simply connect another processor. The key attributes of modular designs are:
- The size of any given system, as quantified in quantum computing by the number of qubits that are available, is adjustable
- Computational resources can be distributed geographically, which not only has performance considerations but also disaster risk management and business continuity considerations
- Uptime can be maximized as individual modules may be taken offline for maintenance, not necessarily requiring entire systems to go offline
- Upgrading, at least in regard to system sizes, is conceptually as easy as adding one or more additional modules
- Like the classical computing concept of multithreading, different modules can perform different calculations simultaneously
- Computational loads can be balanced across modules, similar to how classical computer networks balance network traffic
- In the future, modules might be specialized for performing different calculations, similar to how a classical computer might have a GPU with its CPU
Monolithic designs contrast many of these attributes. The only way to upgrade a system might be to entirely replace it. Maintenance probably requires the entire system be taken offline, although there are classical computing solutions for certain issues. And, the entire system has to be, of course, in only one location.
Top Benefits of Modular Quantum Computing for Businesses
One of the top benefits, as noted in our article titled “The Dual-pronged Energy-saving Potential of Quantum Computers,” is that scaling quantum computers requires far less energy than scaling high-performance (HPC) computers. While some supercomputers consume near 30 megawatts of power, a 10,000-atom quantum computer would consume only 10 kilowatts of power. While doubling the power of a supercomputer means doubling the number of processors and doubling the amount of memory available, thus doubling the amount of energy consumed, doubling the power of a quantum computer means adding a single qubit. Clearly, quantum computing is the more sustainable and the much more environmentally friendly strategy.
While the term “fault tolerance,” in quantum computing, usually refers to error correction, the IDST article notes how modularity grants a different kind of fault tolerance. This is an extension of the ability to perform maintenance on individual modules without having to take entire systems offline, in that an entire module could fail and the remainder of the system should continue to operate.
Some additional significant benefits include:
- Lower financial barriers to entry due to greater cost efficiency, whether that applies to building, upgrading, or leasing resources
- Future-proofing infrastructure investments through the ability to add and to replace individual modules
- Possible computational advantages deriving from modules that are particularly well suited for specific tasks, such as one or more neutral atom modules for optimization tasks
- As an extension of this, enabling a great diversity of applications through the inclusion of arrays of specialized modules
It’s important to note how similar these benefits are to the benefits of modularity in classical computing. That’s no accident. Modularity and distributed computing are commonplace for good reasons.
Challenges and Considerations About Modular Quantum Computing
The IDST article adds that control is a challenge. After all, controlling a single quantum processor is still one of the great challenges of today. Therefore, adding interconnects and networks to the mix raises the challenge to another level. Some additional challenges include:
- Managing interconnectivity may require the interoperation of different modalities, perhaps leveraging photons as a universal medium to transmit quantum information or to generate entanglement over vast distances
- While the need for fault-tolerant quantum computation is well understood, the interconnects and networks will require error detection and correction, as well
- These connections need to be unusually robust, as the loss of a single qubit is significant for an entangled system
- As will become the case for quantum networks, in general, standardization among providers will become essential to establishing any kind of interoperability
- Quantum computers won’t operate on their own, and consideration will have to be given to all the essential classical systems involved, including high-performance computing resources
- Software will have to be developed that can not only interoperate with components from multiple vendors, but that can also manage and balance the use of computational resources
- By limiting available resources to only those needed, returns on investments (ROI) might be realizable faster than might otherwise be the case
- The distribution of computational resources around the globe could introduce new government regulations that require strict compliance
- Beyond any generalized regulatory issues, great scrutiny can be expected in regard to industries for which data protection and privacy are particularly serious issues
- The talent for quantum computing, in general, is already perceived to be in short supply even without additional consideration given for quantum communication
- The distribution of data and computational power may result in concerns over intellectual property rights
- As paradigms shift toward quantum computers and high-performance computers operating together, assessments of HPC’s environmental impact may be warranted
In summary, while modularity resolves certain problems, or at least contributes to their resolution, it introduces new challenges of its own. Fortunately, these mostly appear to be engineering and policy challenges; none are fundamental physics restrictions. Therefore, it ought to be merely a matter of time before the benefits mentioned herein can be realized.