Quantum will push us beyond classical limits
Few could have imagined that the room-sized classical computers of the 1950s would one day fit in our hands. But the impressive development and expansion of computing technology wasn’t only driven by consumer markets. On the contrary, many of the deepest computing revolutions were first adopted and pioneered by high performance computing (HPC) integrators and researchers. The HPC community has long adopted new waves of technology, from vector processors to GPUs, parallel computing and TPUs, to build ever larger and more capable supercomputers. Today’s systems are roughly 16 orders of magnitude more powerful than the first ones, fueling discovery across science and industry through extensive modeling capabilities.
Computational performance is increasingly constrained by fundamental physical limits. Moore’s law, the observation that the number of transistors on a chip doubles roughly every 18 months at constant cost, and Dennard scaling, the principle that power consumption decreases as transistors shrink, are both slowing down. As a result, the traditional growth in the performance and cost-efficiency of classical supercomputers is reaching its limits.
Now, quantum computing is poised to be the next disruptor. It’s not a distant dream. It’s coming fast. This is why, together with Hyperion Research, we published a report to explore what’s next: early fault-tolerant quantum computers (eFTQC). These early quantum computers with 100–1,000 logical qubits and logical error rates between 10⁻⁶ and 10⁻¹⁰ will form a new class of accelerators, enabling HPC to move beyond the limits of any classical supercomputer we could ever conceive.
Building on recent breakthroughs in quantum error correction, Alice & Bob aims to deliver first eFTQC quantum processing units (QPUs) – with 100 logical qubits and logical error rates of 10⁻⁶ – by the end of the decade.
In this blog post, we share three reasons that inspired the report, highlight how it connects to Alice & Bob and future directions.
1) Where early quantum computing will bring value
First, we were motivated by the lack of clarity on the role of eFTQC QPUs in the HPC ecosystem.
The real promise of eFTQC lies in targeted acceleration. For instance, certain quantum algorithms, such as quantum phase estimation, can deliver algorithmic speedups for problems like calculating the energy levels of molecular systems, which are key for understanding drug metabolism or designing cleaner chemical processes, like in fertilizer production. Here, quantum and classical computational methods naturally complement each other: classical HPC systems handle large-scale processing, while quantum processors take care of the most computationally expensive subproblems, such as modeling strong electron correlations in quantum chemistry. The future of scientific computing is not about choosing between classical and quantum, but about blending them in hybrid HPC workflows.
But quantum computing will not impact every HPC workload. eFTQC QPUs will not make large-scale data analysis or near-term AI training faster, and algorithms like Grover’s search, which provide only quadratic speedups, are unlikely to offset the slower operation speeds of quantum hardware. eFTQC will be a powerful tool for the HPC community, but its strengths are domain-specific and need to be well understood, which involves co-designing applications with HPC specialists, domain experts and quantum hardware vendors.
2) What will be the scale of eFTQC’s impact?
Second, at the 2025 Spring HPC User Forum in Santa Fe, New Mexico – the first edition to dedicate a panel to quantum computing – we noted that typical quantum performance metrics – such as qubit counts, gate fidelities, and connectivity – remain mostly complex “Q2Q” (quantum-to-quantum) jargon and don’t provide the visibility HPC practitioners need. By mapping HPC workloads against the potential of quantum acceleration, we identified areas where eFTQC QPUs could deliver the most value.
The result? Materials science is expected to be the first field where eFTQC becomes practical, with chemistry and other many-body problems likely to follow. This is far from a niche opportunity for the HPC community. At major U.S. DOE facilities such as Los Alamos, NERSC, and other leadership computing centers, these fields represent between 13 and 54 percent of current computational workloads.
The report also points out that the eFTQC application landscape is evolving rapidly. Algorithmic advances in quantum chemistry, for example, have reduced quantum resource requirements by orders of magnitude in just the past five years. This trend suggests that more applications could soon fall within reach of eFTQC QPUs. Importantly, new approaches may only be discovered once hardware becomes available and researchers begin experimenting with it.
3) Integrating quantum: the path forward
Our third motivation for writing this report is the need to act now to be ready for the integration of eFTQC QPUs into HPC centers. These systems will not be plug-and-play: making them operationally viable will mean co-designing between HPC and quantum makers. It is not the first time that forward-looking HPC centers have co-designed for the integration of early technologies.
Yet, integrating eFTQC will differ fundamentally from past classical accelerators. Unlike GPUs or other devices, eFTQC QPUs demand specialized infrastructure, from cryogenics to vibration isolation, and introduce a completely new programming model. Some centers have already experimented with noisy intermediate-scale quantum (NISQ) processors, gaining some valuable early operational learnings. But eFTQC QPUs add an error-correction layer, leading to larger physical footprints, new software stacks, and different algorithms. User access will also evolve, with greater abstraction built into the system.
This complexity means its important to start integration efforts early. Centers that move first will not only be better prepared operationally, they will also secure scarce eFTQC QPU capacity. In fact, over the next decade, demand for eFTQC QPUs is expected to far exceed supply. Strategic alliances and first-mover advantages will therefore be critical.
Moreover, early deployments of prototype QPUs create opportunities beyond access alone: they allow HPC centers’ user communities to shape the field itself – through the development of advanced benchmarking techniques and by supporting the maturation of promising hardware modalities, such as cat qubits, that could define the era of utility-scale FTQC.
In this report, we provide practical steps HPC centers can take to prepare for eFTQC and offer a framework to help them position themselves relative to peers.
And Alice & Bob in that?
Mainstream transmon approaches betting on the superconducting platform anticipate systems that will be of the size of a full building, with dozens of thousands to millions of physical qubits.
At Alice & Bob, we change this equation. The cat qubit design suppresses bit-flip errors – one of the two main errors affecting qubits – almost entirely, simplifying error correction codes. Crucially, there is already a published architecture for universal FTQC with cat qubits showing that only thousands of physical cat qubits are needed for useful applications [1,2], rather than millions of transmon qubits.
Importantly, there is no engineering penalty for cat qubits over transmons. They use the same control hardware, cryogenic environment, and reliable fabrication processes. Cat qubits also operate at the same speed as transmons, which is orders of magnitude faster than neutral atoms, a difference that brings FTQC workloads from months or years to hours. What changes is the circuit design, yet that design makes FTQC more than 90% more efficient.
For HPC centers, the implications are direct: fewer qubits means less control electronics, less cryogenics, smaller footprint, and lower power demand. That makes it realistic to envision FTQC accelerators co-located with supercomputers, rather than housed in separate dedicated facilities that need to be purpose-built and equipped with new power infrastructure.
Alice & Bob’s Graphene QPU will be well suited for HPC integration. This first generation of eFTQC QPU will fit in a single cryostat for seamless installation within infrastructure that exists already today. The journey starts now. Prototypes of cat qubit-based eFTQC QPUs can already be installed to prepare the arrival of Graphene in 2030.
Future directions
Going forward, we identify four areas for particular attention:
- Deeper analysis of the first eFTQC QPU applications. In the report a few examples are presented. But, it’s still unclear what can be achieved in an advantage regime with the lower bound of the eFTQC regime i.e. ≈ 100 logical qubits at error rates of 10⁻⁶. In his Megaquop paper at Q2B 2024, Preskill noted that the concrete end-user value of such machines needs to be better defined. Demonstrating relevance will require new algorithms, compilation optimizations, resource estimates, and co-design with domain experts in end-to-end workflows.
- Rethinking the software and algorithmic stack. Fault-tolerant systems are not just better NISQ devices: they introduce logical qubits, discrete gate sets, new compilers, and different programming models. Open questions include where classical HPC systems can best contribute to the stack and how much HPC center and user control over future eFTQC QPUs should remain.
- More QC / HPC exchanges on operational lessons from eFTQC QPU prototypes deployments. These machines with a few logical qubits may offer limited end-user value today, but they are crucial learning tools for shaping full-scale integration. Sharing lessons on access models, infrastructure challenges, and environmental sensitivities will be critical to accelerating deployment when the time comes.
- Estimating the ROI of eFTQC QPUs. This report approximates eFTQC QPUs’ impact by mapping their applications’ domains to HPC’s. Building on top of point 1, scoping what specific classical HPC workloads will be offloaded to eFTQC QPUs in an advantage regime (either from a cost, time or accuracy perspective), along with more visibility on the system cost, will enable clearer value estimations for the HPC community.
Finally, unlocking the full potential of eFTQC QPUs in HPC workflows will take continued collaboration between the quantum and HPC communities. Industry events like the HPC User Forum from Hyperion Research and major supercomputing conferences (SC, ISC) have proven to be invaluable platforms for sharing ideas and learning from each other. We’re excited to keep the momentum going!
In the meantime, dive into our report: https://alice-bob.com/hpc-report/
Our inbox is open—get in touch:
juliette.peyronnet@alice-bob.com
References
[1] Gouzien et al., 2023, arXiv:2302.06639
