In 2020 everyone will want to persuade you to join their quantum software camp. Not everyone will come away happy. Huddling round the community fire may be a good way to keep the chills of quantum winter at bay.
Everyone knows how lucrative controlling the software playing field became in the original Digital Revolution. Everyone knows that building a community into a mutually beneficial ecosystem is a key value driver. These insights remain true, but might the seductive similarity of the jargon blind us to important differences buried in the quantum software stack?
Increasingly we need to ask who is running a real service with real users, versus who is building an aspiration.
2019 saw an explosion in the number of quantum platforms, each typically offering: tutorial materials, a software development environment, a simulator for testing code and (in some cases) execution on real quantum hardware. Some have users, some don’t. Some aren’t even really launched yet. All want you to be part of their community.
Early existing proprietary offerings such as IBM Q, Rigetti QCS and D-Wave Leap have been joined by launch announcements from big tech players offering semi-hardware agnostic solutions, notably Microsoft Quantum Azure and Amazon Braket. IBM Q has already responded by adding access to an early trapped ion device from AQT. Atos may have spotted a niche for those needing a secure on-site classical simulator to avoid having to expose their ideas to the cloud.
Quantum startups have also entered the market, typically with hardware agnostic visions, such as QC Ware Forge (open beta) and Zapata Orqestra (private beta). More distinct again is Strangeworks with QuantumComputing.com (open beta), a vision focused on gluing the community together.
Other commercial players have early initiatives that probably prepare us for more to come: Xanadu Strawberry Fields prepares the community for its planned photonic processor; Alibaba QC-QDP registers their intent; European startup JoS Quantum is also set to enter this arena. China based Origin Quantum has initially offered a simulator based cloud service.
Google Cirq and the Google team’s activity across other open source initiatives has helped prepare us for a presumably much bigger cloud launch to come.
2019 saw the first quantum-on-quantum acquisition. Hardware-led Rigetti acquired software startup QxBranch, who are now operating as the Rigetti Applications Team. This reinforces the QCS ecosystem and enhances Rigetti’s client facing capability.
To date a key target user segment has been university and college establishments that want to introduce their researchers and graduates to this new world. This is set to expand as early business adopters also start seeking to build teams and experience. Others will focus on the more advanced tools and libraries to support the actual development of early applications.
Following Q2B 2019, analysts VentureBeat commented “The devil is in the details. How much are these cloud services for quantum going to cost? Amazon and Microsoft haven’t said. When exactly will they be available in preview or in beta?.”
Solutions, Applications & Libraries
Long ago the digital software marketplace saw a transition from custom solutions to packaged applications and then SaaS. 2019 found early quantum software startups wrestling with the parallel question of balancing vertical market insight with horizontal problem class expertise.
In the conventional software market, vertical industry expertise is often crucial in successfully selling to enterprise customers. In quantum software it’s set to be doubly vital, as often a core challenge will be working out which business processes or computing task hides a problem whose computational complexity is susceptible to solution via a quantum algorithm. However, in the quantum sector genuine horizontal problem class expertise is already a scarce resource and so a key differentiator. We can already see IBM carefully integrating these dimensions of their software strategy.
Quantum software startups are typically doing all of the above at once. However some differences of emphasis are emerging.
Some are focusing on key vertical opportunities by building focused tools and libraries. For example in drug discovery and materials design: 1QBit QEMIST, CQC EUMEN and Riverlane Anian. ProteinQure have an even more specific focus on protein therapeutics, also including conventional machine learning toolsets.
Other players are notable for early engagement with the Financial Services sector. JP Morgan Chase, Goldman Sachs and Barclays are all part of the IBM Q Network. Standard Chartered Bank and Nomura Securities are among D-Wave users. Multiverse is doubly notable as a quantum software startup in southern Europe and one that focuses on the Financial Sector. Multiverse emphasise their expertise in both early quantum and advanced classical algorithms.
Others emphasise particular problem classes such as quantum machine learning: Xanadu (Penny Lane) and Rahko (Hyrax); or hard optimisation: Beit, also notable as a quantum software startup in central Europe.
Startups with expansive visions such as Zapata, QC Ware and Strangeworks are offering platforms that aim to span all of these areas and are happy to take a hardware agnostic view. Still there are differences of emphasis. Zapata talk first about industry aligned solutions, QC Ware first about problem class excellence. Strangeworks is notable for the emphasis it puts on enabling collaboration across the wider quantum community to drive innovation and discovery.
As an engaged potential end-user Airbus has itself seized the initiative. It has defined five key ‘flight physics’ problems (the vertical insight) and has launched the Airbus Quantum Computing Challenge to encourage the community (horizontal problem experts) to find ways quantum computers can help. Interestingly the respondents to the challenge have been evenly split across academic institutions and startups, illustrating the rapid commercial development of the field. An important aspect of the challenge is also the vehicle it provides to build awareness and focus thinking internally within Airbus.
Compilers, Errors & Control
The lower levels of the quantum software stack are even further removed from their conventional equivalents: optimising quantum compilers are significantly more conceptually daunting than conventional compilers; digital bits don’t require error correction; implementing optimal control over quantum hardware is a much more important and challenging task than it is for conventional hardware.
In 2019, early movers have been active in all of these proto-layers of the quantum software stack.
CQC have emphasised the superior performance of their compiler t|ket⟩ over rival platform-native compilers. Quantum Benchmark point to the ability of the True-Q compiler to leverage results from their respected cycle benchmarking technology .
Q-CTRL point to the performance of their BOULDER OPAL control optimisation suite and see strong opportunities for future gains by combining the control and error correction layers . Intel’s launch of Horse Ridge gives us a definite model of where such firmware could one day reside.
It’s still not clear how the compilation, error correction and control layers will in the end interact or possibly merge. This field remains in active development. Opportunities to optimise compilation will vary widely between FTQC and NISQ applications.
To watch in 2020
IBM Q – By far the leading platform for quantum first contact. The wider user base now stands at 193,000 cloud users, 115 billion circuits executed, 240,000 development kit downloads, 200+ external papers, 1500+ colleges and universities, 300+ schools, 300+ private institutions. The IBM Q Experience offers a free easy-to-use introductory experience. The more powerful QISKit, including the OpenQASM intermediate level language is well established in research and expert educational use. Combined with leading processors and a stable development path IBM will be a tough nut to displace from this segment of the market in the short term.
D-Wave Leap – Supported by their Ocean software development kit and focused specifically on quantum annealing. Its growing user community has developed 200+ pilot applications. While most quantum computing startups are still talking about how much investor funding they have raised, D-Wave is able to point to $60m of customer contracts. In 2019 D-Wave expanded access to their Leap quantum cloud service to Europe and Japan. Watch out for D-Waves’ partnership with NEC, including a focus on hybrid solutions and enhanced applications development capability, significantly strengthening this ecosystem.
Rigetti Quantum Cloud Services – Supported by the Forest development environment featuring the Quil instruction set and pyQuil library. Rigetti point to the unique features of their setup that specifically facilitate near-term hybrid quantum-classical applications: tight coupling of CPU & QPU directly within QCS, support for (iterative-processing friendly) parametric compilation and (latency-friendly) active qubit reset. Activity in this ecosystem has been bolstered by the acquisition and integration of QxBranch and the lure of the Rigetti $1m Quantum Advantage prize. Rigetti currently offer an Aspen 32-qubit processor and are focusing on increasing the scale and performance of their devices. How many expert users will it attract?
Amazon Braket – 2019 saw the dramatic announcement that Amazon Web Services, a leader in conventional cloud computing are finally entering the quantum sector. Braket promises a hardware agnostic development environment with access to a variety of classical simulators and quantum processors from a variety of providers (currently IonQ, Rigetti and for quantum annealing D-Wave). There is an emphasis on training and introductory materials and easy to access virtual environments. This platform seems well placed to appeal to corporate teams wanting to get a first foot in quantum, but not wanting to commit to one hardware provider. When will this service move beyond the current ‘preview’ phase? How many users will it add?
Microsoft Quantum Azure – Announced but not yet publically launched. This service looks to be a big enhancement of Microsoft’s previous offer, promising access to quantum hardware from a variety of partners (currently IonQ, Honeywell and QCI are named). Featuring Q#, its purpose-designed quantum high-level programming language, this could be a popular choice for corporate software teams familiar with Microsoft’s .Net and Visual Studio environments and wanting to build initial experience with quantum. Watch out for the public launch. How many users will it add?
Google – Currently offering the Cirq development framework, and notable participants in the ProjectQ and OpenFermion open source initiatives. Google has made no secret of its plans to launch a cloud service. Access to its early supremacy era processors will give it a strong cachet. Cirq is aimed squarely at NISQ applications and is supported by strong internal work on early applications. Expect this to make a big splash when launched.
Atos Quantum Learning Machine – The QLM is a powerful classical simulator rather than a quantum computer. But by offering a relatively affordable standalone on-premise package Atos provides a solution ideal for corporates worried about the security of sensitive research passing onto a cloud environment. Partnering with Zapata to bundle the Orquestra software platform with QLM promises a uniquely vertical industry specific focus. Who needs a real quantum processor that hasn’t even shown quantum advantage yet?
Origin Quantum – This startup based in Hefei, China is currently offering a simulator based quantum cloud service. Will we see it add access to one of its early quantum processors?
Hardware agnostic startups – Software led players such as Zapata, QC Ware and Strangeworks will want to work with the widest possible selection of hardware options. Each will work hard to build-up a community. QC Ware’s development of the successful Q2B conference series and Strangeworks’ activity on Stack Overflow are not accidents. Watch out for which hardware majors are willing to support a genuinely level playing field.
Compiler supremacy – Platforms aimed at serious research and development use are likely to increasingly compete in terms of the efficiency of the software environment they provide. Will we see a software provider able to claim that they delivered a quantum supremacy calculation based on compiled rather than handcrafted code?
Airbus Quantum Computing Challenge –1000+ respondents from 70+ countries registered an interest in the challenge. 160 initial proposals have become 36 final submissions. An impressive jury of eight international quantum experts has been assembled. Watch out for awards in Q1/Q2.