Quantum Computing Is Moving From Better Qubits to Better Systems

In a conversation at Quantum World Congress, Saikat Guha and Valentino Jadriško explore why the next challenge in quantum computing is increasingly about integrating different technologies into scalable systems, not simply building better individual qubits.

For years, progress in quantum computing has been measured largely at the level of the qubit. Superconducting circuits, trapped ions, neutral atoms, photonics and solid-state approaches are compared by fidelity, coherence, error rates and their prospects for scaling.

Those comparisons still matter. But as individual components improve, a different set of constraints comes into focus. A useful quantum computer requires much more than high-performing qubits. Entanglement has to be generated and distributed. Errors have to be corrected. Components need to communicate, operate together reliably and eventually be manufactured at meaningful scale.

At Quantum World Congress in College Park, Maryland, Valentino Jadriško of Constructor Capital discussed this transition with Saikat Guha, Clark Distinguished Chair Professor of Electrical and Computer Engineering at the University of Maryland and Director of the NSF Center for Quantum Networks. Guha’s work spans photonic quantum information processing, communications, sensing and computing.

The shift is subtle but important. Quantum computing is becoming as much a problem of system architecture as a problem of qubit performance.

The best qubit depends on the job

No major quantum platform is strongest across every dimension.

Trapped ions can support high-quality quantum logic, while their efficiency in interfacing with optical photons remains limited. Neutral atoms offer both strong optical interfaces and quantum logic, but moving atoms into the regions where operations take place introduces another engineering challenge. Guha summarises the situation simply: “There’s all these pros and cons right now between different qubit forms.”

Those trade-offs matter more as the industry moves beyond isolated demonstrations.

A processor can have excellent qubits and still face difficult problems elsewhere in the architecture. Large entangled states remain challenging to prepare. Optical systems have to manage loss and probabilistic processes. Error correction can require substantial physical overhead. Interfaces between components can become bottlenecks of their own.

This changes the meaning of scale. Adding more qubits is only one part of it. The surrounding system has to scale with them.

Scaling is becoming an engineering problem

Quantum computing still depends on fundamental scientific progress, but many of the problems now coming into focus are problems of integration.

“We are at the point in the journey where the next step is an engineering challenge.”

Guha makes that assessment after discussing progress in qubit quality, spin-photon interfaces, quantum logic and scalable linear optics. The engineering challenge sits in how those capabilities are assembled.

How efficiently can entanglement be generated? How is quantum information moved between processors? How much physical hardware is required to create useful logical qubits? Which components can rely on established semiconductor or telecom manufacturing processes, and which will require entirely new infrastructure?

These questions can determine whether an architecture that works in a laboratory can become a practical system.

Integration can matter as much as invention

One route to scale is to improve every component of a single hardware platform. Another is to make better use of technologies that already solve different parts of the problem well.

Photonics and matter-based qubits illustrate the logic.

Photons are well suited to transmitting quantum information. Atomic systems, meanwhile, can provide interactions and quantum logic that light alone cannot.

“The atomic qubits that talk to photons, they bring in very useful nonlinear interactions in the optical domain that light does not possess on its own.”

That distinction is not merely theoretical. Guha gives an example from laser communications, where adding atomic qubits to an optical processor can expand what the system can do beyond linear optical processing and photon detection alone.

The same systems perspective extends to infrastructure.

Technologies compatible with mature industrial platforms can inherit decades of engineering development. Silicon brings established fabrication processes and a large semiconductor ecosystem. Optical quantum systems can draw on fibres, modulators and other components developed for telecommunications. Guha specifically points to the value of building on “existing well-researched, well-developed infrastructure.”

For quantum companies, that makes integration strategy part of the technical roadmap. The relevant question is not only whether a component performs well in the lab, but whether it can fit into an architecture that can realistically be built and scaled.

The quantum stack may become heterogeneous

Different strengths across hardware platforms also leave room for systems in which technologies take on specialised roles.

A processor with strong optical properties could handle interaction with a signal or environment, while another architecture performs computation more efficiently. In a quantum network, different nodes could be chosen for scalability, communication or error correction.

Guha gives a concrete example: a quantum repeater could use scalable color-center nodes across part of a network and later shift the burden of error correction to trapped ions with stronger error-correction capabilities. He concludes that “there’s all sorts of reasons for hybrid processors,” while stressing that it remains too early to know where different processors will ultimately find their place.

That uncertainty is important. The argument is not that hybrid architectures have already won. It is that a winner-takes-all view of quantum hardware may be too narrow.

Classical computing offers a useful comparison. CPUs, GPUs, specialised accelerators, different types of memory and networking hardware coexist because no single architecture is optimal for every task. Guha makes the same analogy when discussing the future of quantum infrastructure: mature computing systems are already diverse, and quantum technologies may eventually settle into similarly specialised roles.

For investors, this shifts the question from which qubit modality will win to which companies control the bottlenecks that every scalable system will need to solve. That may include connectivity, photonics, control, error correction, fabrication and interfaces between hardware platforms. In a market where the final architecture is still uncertain, owning a critical enabling layer can be as strategically important as building the processor itself.

This also changes how quantum companies can be evaluated. Technical performance remains essential, but so do manufacturability, dependence on specialised infrastructure, compatibility with existing supply chains and the ability to solve a bottleneck that persists across more than one hardware architecture.

The companies that matter most may not all be building complete quantum computers.

Much will still depend on which architectures prove useful outside the laboratory. As Guha puts it:

“The most important, most valuable things will find their way. And then, obviously, the industry and the use cases will drive the maturation at scale.”

The race to improve qubits is far from over. But the next phase of quantum computing will increasingly be decided by what can be built around them.