Quantum Hardware Outlook 2023 - Long Summary for QCR Paid Subscribers

QCR by GQI
Jan 9, 2023
11 min read

The full report is available for purchase on the GQI website at this link CLICK HERE

The performance of today’s quantum hardware remains modest, and despite the hard work of innovators that is not going to change overnight. Each qubit platform faces its own significant challenges as vendors seek to scale up. But if we look at the details we can see that underlying progress remains strong across the hardware stack. Progress on error correcting codes must not be neglected. Innovation there has the potential to disrupt our standard assumptions about what it takes to build a quantum computer.

The strategic hardware playing field

As the quantum computing sector comes off the peak of its first hype cycle, questions on progress are inevitably being asked. This outlook asks specifically how the hardware itself is stepping up to the challenges it faces. To understand this, we need to appreciate the interplay between the performance specs that current hardware is able to achieve, and those required to deliver against different scenarios that players are targeting.

The NISQ era

Some look for early noisy intermediate scale quantum (NISQ) devices to step-up and support commercially useful applications in the next 2-5 years.

Opinions differ on requirements. Some believe quantum annealing is already starting to do this. GQI believe gate-model protagonists are ideally looking for:

100-200+ physical qubits
99.99%+ fidelities, especially 2Q gate fidelity
High qubit connectivity

Probably this would be delivered in a single module, so interconnects may not be an issue.

It's important to realize that there is no guarantee that quantum algorithms can deliver on their part of this bargain. Classical algorithms are a fierce competitor. Niche applications certainly look possible but broad quantum adoption looks much less likely without a further breakthrough.

A low depth algorithm breakthrough could still yield broad NISQ quantum advantage. However without further progress on algorithms the alternative could be a hard quantum winter. Making hardware developers financial plans robust to both these scenarios is a challenge.

Early FTQC

Others focus on the potential that will be realized by larger machines to deliver fault tolerant quantum computation (FTQC). Early machines will still be limited.

Much current discussion assumes a ‘standard model’ of how this will be achieved:

99.9%+ fidelities, especially 2Q gate fidelities
Beyond a certain module size, interconnects are essential for scaling
Sufficiently fast classical control logic
The 2D surface code (or similar) used to encode physical qubits into logical qubits

With standard error correction techniques overheads are very high. A system with 1 million physical qubits might offer only 200-300 truly high performance logical qubits, though better fidelities or better connectivity could radically improve this.

If we believe that several of today’s vendors can deliver on their announced roadmaps we will see broad early FTQC in 5-10 years. Many don’t see things playing out so smoothly. Scaling up is hard and some believe only their approach has what it takes to deliver eye of the needle FTQC.

The longer term

The longer term landscape will be shaped by developments that are even more uncertain.

Goliath FTQC - the direct future path would simply be brute force scaling of current hardware architectures. In the end we want 10,000+ logical qubits. Large device footprints could lead to truly immense machines, putting pressure on affordability.

Distributed FTQC - the emergence of quantum enabled networks provides new niche opportunities. In the end, an entanglement based quantum internet could potentially allow a literal exponential multiplication of quantum resources.

Turbo FTQC - to allow a fuller range of quantum speedups to become viable we would ideally like a new architecture with faster logical clock speeds, low overheads and capabilities such as a hardware efficient implementation of QRAM. No one is yet building such an architecture, but it is reasonable to believe that someday someone will.

FQQC opportunities

Not all applications require lots of qubits. ‘Few qubit’ quantum computing (FQQC) devices may find a variety of use, particularly in network related applications.

Cryptographic applications may provide an early opportunity in quantum enabled networks. Some see helping to join up a future entanglement based quantum internet as the biggest opportunity of all.

Not all technology platforms will turn out to be suited to scale-up for large scale computational tasks. FQQC might be an excellent plan B.

For players with the right technology, and the strategy to finance the journey, there is no reason why NISQ, FTQC and FQQC opportunities cannot be joined up within one overall roadmap.

Jerry Chow (IBM Fellow) comments “The true opportunity is that there is one model of quantum computing: quantum circuits. We will make stepwise advances that allow us to execute quantum circuits more accurately. This will involve a mix of error suppression, error mitigation, and eventually error correction. Different hardware will be able to leverage these various techniques in different ways. Being optimistic about near term applications doesn’t preclude a desire to move on to FTQC. But current theoretical approaches to FTQC are speculative. We need to develop more hardware-aware higher efficiency error correction codes.”

Key challenges

The reality is that today’s devices are not yet anywhere near as good as we need them to be: the smaller devices are not large enough to move beyond the reach of classical simulation; the larger devices are not able to sustain the quality we need. The potential roadblocks we face in scaling up current designs vary by qubit platform, but are significant.

Looking across progress in 2022, GQI believes several overall themes emerge:

Current real multi-qubit devices struggle in particular on 2Q gate fidelity. However, 2nd generation designs are emerging that do promise 99.9% or even the chance of 99.99%.
NISQ-optimists may seek to get over the line using only single modules, but for most FTQC approaches interconnects will be required to scale further by linking modules together. Making these work at the required rate and fidelity looks more challenging than many realize. Startup activity is targeting this segment.
Photonics remains an important tool across the wider quantum tech sector. Both as a potential platform for quantum computing in its own right, but also in the control and interconnection of multiple other technologies. But the most developed conventional photonics platforms (SOI and SN) don’t necessarily work at the wavelengths we want or provide the active components we need. Innovation to expand our photonics toolset is an increasing area of activity.
Fabrication cycle times are an increasing R&D challenge, both due to growing technical complexity and for global supply chain reasons. The need for new and more exotic components (e.g. 3He, 28Si) and fabrication capabilities (e.g. thin film LNOI) is taking us further into the domain where geopolitics will be important in shaping practical R&D possibilities.
Progress on the theory of error correcting codes has been rapid and has proceeded in parallel to much other work in the sector. Platforms that are able to offer higher levels of connectivity could be well placed to exploit new emerging opportunities. This could be highly disruptive to the assumptions many have for what it takes to build a quantum computer.

This review looks at progress in the lower hardware layers of the quantum stack: the quantum plane, the control plane, control logic and device architecture. As we shall see, current designs face challenges, but in many cases there is tangible progress pointing to new solutions to the known issues.

Quantum Plane & Control Plane

Trapped Ions

Systems with the scale and performance desired by NISQ-optimists remain some way off. The current leading players such as Quantinuum and IonQ use Raman lasers to drive gates. Further innovation is needed here to drive fidelity over the line.

Many point to the challenge in scaling the laser optics systems for gate control. An increasing number of new trapped ion startups are proposing radical new schemes that seek to mitigate or remove this challenge.

The ability to leverage photonic interconnects is a relative strength for ions. But at current levels of performance it also threatens to be a bottleneck. Startups are working on solutions to this challenge.

Superconducting Circuits

Early processors have highlighted the practical challenge of combining larger qubit counts with sufficient qubit fidelity. IBM’s 100x100 challenge stands out as an ambitious attempt to bring together the enabling technologies required to target NISQ applications. The research market itself and co-design opportunities are themselves important early drivers of this segment.

Quantum annealing and quantum-inspired ‘digital annealing’ represent a separate and complementary approach for those seeking early benefits in optimization applications.

We see existing fidelities as inadequate. An increasing number of new entrants and startups are pointing to new coupler or qubit designs that they believe will deliver a step-change in performance.

Interconnects remain a challenge for superconducting circuit based devices. It's increasingly acknowledged that these will ultimately be required, but work to deliver this technology remains at a relatively early stage.

Managing the cooling budget within the dilution fridge remains a challenge for those serious about scaling. Leading players are increasingly emphasizing their initiatives in this area.

Photonics

Photonics based systems have made great strides, USTC’s demonstration of ‘beyond classical’ gaussian boson sampling with Jiuzhang has been followed by an even more convincing demonstration by Xanadu’s Borealis.

The roadmaps of current leaders capture the benefits of conventional photonic platforms (such as SOI or SN). However it is also becoming increasingly clear how working within the constraints of these platforms puts limits on full system integration.

New startups are exploring the potential of new photonic platforms to enhance functionality and integration. Thin film lithium niobate (TFLN) stands out as a material with much potential for wide application in photonic quantum computing and across quantum technology.

Neutral Atoms

Analog mode for NISQ era applications continues to be an area of leadership for neutral atom platforms. Practical gate-model schemes for neutral atom systems also took much more practical shape in 2022. Scaling the required laser control optics is a recognised challenge, as are interconnects. Startups are targeting both these areas.

Silicon Spin

Silicon spin qubits have made great progress, but remain at an earlier stage than other approaches. As pioneers start to move into true multi-qubit devices, innovation is still focussed on questions of practical component layouts. Contrary to the image sometimes presented, this segment encompasses a wide variety of manufacturing techniques, from an emphasis on standard mainstream silicon fab, to the ultra cutting edge.

NV Diamond

NV centers in diamond offer unique characteristics for quantum applications. Commercial groups are working to exploit these in both cryogenic and room temperature settings. A clear match of performance to application remains to be demonstrated.

Further qubit platform Innovation

There remains plenty of space in the quantum sector for new qubit platforms to enter the race. These typically promise properties that could be very disruptive to conventional assumptions, but have not yet demonstrated in practice all the operations necessary for qubit control. CAT qubits are an exciting proposal that seek to simplify the error correction challenge, with potential to sharply reduce overheads. New host platforms for spin qubits continue to show promise. Topological qubits remain a tantalizing long term prospect.

Control Logic

The needs of the early market have created a vibrant market in R&D control hardware. The question remains which of these early players will manage to evolve their business models to embed their products or IP in production systems.

For the future, many envisage that at least front-end control logic will have to run at cryogenic temperatures. Both tech majors and startups are developing cryo-CMOS solutions. Others look to novel solutions such as SFQ, or photonic technology.

The large, highly parallel, classical processing challenge likely required to support the operation of quantum error correcting codes remains underappreciated. This could be a challenge particularly for those platforms targeting fast code cycles.

Architecture

2022 was a great year for ground breaking practical demonstrations of quantum error correcting codes and basic fault tolerant gate operations. On the other hand the most ambitious of these, Google’s 2D surface code demonstration, serve as a reminder that overheads do not look workable unless physical qubit fidelities are first further improved.

Practically focussed work increasingly points to the benefits possible if we are able to move beyond simple 2D nearest neighbor connectivity. Photonic, trapped ion and neutral atom platforms currently look best placed to take advantage of this, though other platforms are working to bridge this gap.

Progress on ‘good’ quantum LDPC codes points to an entirely new paradigm of quantum error correction. A qubit platform with the high connectivity able to make practical use of these techniques would be highly disruptive to current sector assumptions regarding what it takes to achieve FTQC.

Vendor roadmaps

IBM roadmap – 27Q (Falcon) 2019, 65Q (Hummingbird) 2020,127Q (Eagle) 2021, 433Q (Osprey) 2022, 1121Q (Condor) & 133Q x p (Heron) 2023; 1386Q l-couplers (Flamingo) & 408Q m-couplers (Crossbill) 2024, 4158Q l&m couplers (Kookaburra) 2025; Scaling to 10k-100k Q 2026+.Preferred metrics: qubit count, QV and CLOPS.

Google roadmap – 100Q (logical qubit prototype), 1000Q (logical qubit), 10kQ (tileable logical modules), 100kQ (engineering scale-up), 1MQ (error-corrected quantum computer) by 2029. Error correction via surface code protocol. Preferred metric: 2Q gate fidelity in simultaneous operation.

Rigetti roadmap - 40Q (Aspen) 2021; 80Q (Aspen-M) 2022, 84Q (Ankaa) 2023, 336Q (Lyra) 2023/24, 1000Q 2025, 4000Q 2027

Quantinuum roadmap – H1 (linear trap), H2 (racetrack layout), H3 (grid layout), H4 (integrated optics), H5 (large scale via tiling); by 2030.

IonQ roadmap – 22AQ 2021, 29AQ 2023, 64AQ 2025, 256AQ 2026, 384AQ 2027, 1024AQ 2028. Error correction – 16:1 2025, 32:1 2027

D-Wave annealing roadmap – 5000Q Advantage 15X connectivity, Advantage performance updates 2022; 7000Q Advantage 2 20X connectivity 2023/24, improved coherence enables improved connectivity 2025.

D-Wave gate-model roadmap – Phase 1: validate qubits in multilayer stack; phase 2 validate error correction; phase 3 demo logical qubit manipulation; phase 4 Design scalable-task specific components; phase 5 First integrated general purpose processor.

To watch in 2023

IBM - can ‘little bird’ fidelity can indeed be combined with ‘big bird’ scale?
Google - can improved fidelities make a 2D surface code logical qubit realistic?
D-Wave - will we see more details of its gate-model design and SFQ control logic?
Couplers & transmons - can a new design deliver a leap in fidelity?
Writing up the fridge - will we see states for IBM’s flex wiring and System Two fridge?
Quantinuum - will we see firm dates for the promising H3 QCCD grid architecture?
IonQ - what performance specs can barium ions deliver?
Trapped ion gates - will one of the new gate schemes break the fidelity ceiling?
Faster interconnects - will we see progress with optical cavities for ions or atoms?
Xanadu - how will the Borealis technology integrate into its roadmap?
PsiQuantum - will we finally see some stats on its resource state generators?
New photonic platforms - will we see thin film lithium niobate start to make waves?
Neutral atom analog mode - will metrics emerge for these interesting platforms?
Neutral atom gate mode - will we see a practical readout scheme demonstrated?
Silicon spin - will we see a practical 2D device layout demonstrated?
NV Diamond - what future roadmap will emerge for this platform?
CAT qubits - will we see a demonstration of qubit control?
Topological qubits - what more will we learn about Majorana modes in InAs-Al?
Control hardware - will we see evidence of what SFQ can offer over cryo-CMOS?
Active-volume architecture - what other platforms might exploit PsiQuantum’s ideas?
Q LDPC gate schemes - what practical, efficient gate schemes will we see proposed?
Disruption - will we see an architecture exploit high connectivity Q LDPC codes?

This report is a Must Read for providers, users, investors or anyone interested in quantum computing and what’s in store for it in the future.

The report totals 59 pages in length and covers the quantum hardware landscape in greater depth than any other source. Our analysis is product and technically driven, often with much scientific detail, while also offering a good overview for the quantum novice on major approaches and differentiations.

You can find a free summary of the report at https://quantumcomputingreport.com
QCR members can access an extended summary at https://quantumcomputingreport.com/

If you decide to purchase the report you can pay via PayPal or credit/bank card by selecting the appropriate option on the PayPal form. You can do this at the GQI website at this link CLICK HERE

The Table of Contents is shown below.

January 9, 2023