Quantum has a lot of hype, as all new technologies with large promises do. Some of this hype is because the idea of quantum mechanics is so romantic. It moves us a little bit closer to a more accurate model of reality. Rather than the binary state of bits in classical computing (similar to heads and tails of a coin), quantum computing rests upon qubits (comparable to an ever-spinning coin, as the qubit exists in a ‘superposition’ of both states at once). Quantum superposition represents the complexity of the universe in a way that binaries never really have. But it’s also incredibly hard to know, across many different areas of research, what approaches will succeed.
Quantum mechanics has been pitched as holding promise for a lot of industries especially hardware and software for computing, as well as sensing and metrology. I decided to try and work through what is valuable here between now and the next five years, and hopefully this will be a good baseline for people to understand what we at Compound are digging into across quantum hardware, software, and sensing fields and what we find promising in each.
“While the state of a classical computer is determined by the binary values of a collection of bits, at any single point in time the state of a quantum computer with the same number of quantum bits can span all possible states of the corresponding classical computer, and thus works in an exponentially larger problem space.” – Quantum Computing: Prospects & Progress, AAAS
Figuring out the Right Problems
Guided by a lot of theoretical work and the foundations of quantum information science, we know some problem spaces are much better handled by classical computing, and some are best handled by quantum computing. But in between there’s a big grey area, and some companies are systematically working through those in-between problems to try and better understand the limitation of current quantum hardware/architecture.
The complication here is that the companies we’re interested in are asking the question, “based on the hardware we have right now, on what problems is quantum hardware actually the best tool to use over classical?” This is necessarily tenuous, because surely there are quantum algorithms out there we just haven’t thought of yet.
“Nature isn’t classical … and if you want to make a simulation of nature, you’d better make it quantum mechanical, and by golly it’s a wonderful problem, because it doesn’t look so easy.” – Richard Feynman.
Figuring out the Foundations
In terms of physical hardware progress, it can be hard to differentiate the publicity from what’s actually going on behind the scenes. The cascading real-life benefits that come from being perceived as a leader in the capital- intensive hardware race also make this development all less interpretable from the outside (ie positive perception leads to additional funding leads to ability to continue research). Additionally, the nature of competition in this space means companies are often guarded about their projects, for fear of leading competitors towards their unique approach. This Twitter account, for example, has gained fame by calling out bogus or at best dubious press releases related to quantum. Therefore, I’ll be careful to frame companies’ claims as such. I also should note that my background is not in quantum computing.
Setting the Stage: the State of Quantum Computing Hardware
Today the progress in hardware that underlies quantum computing is roughly comparable to the vacuum tube or transistor stage of classical computing, as if we were currently trying to invent the integrated circuit and scale up. For comparison, the integrated circuit largely represented scalability in computing – something quantum is still grasping for.
There is massive hype around quantum computing, quantum supremacy, and even venture capital placed into quantum. The past few years have been a whirlwind for the industry; the White House launched The National Quantum Coordination Office in 2018 as a hub for research and education, the UK is running a national initiative, India is making it a national mission, the EU has a quantum technology flagship, and China is wrapping quantum into their next five year plan.
Two of the main quantum algorithms, Shor’s algorithm (for factoring) and grover’s algorithm (for search) were researched decades ago, with the assumption that you had a perfect quantum computer and fault tolerance (i.e. millions of error corrected qubits). But they’re prefaced on being far past the NISQ era of quantum computing – we need algorithms for today. Alphabet’s X, or moonshot lab, notably has a second internal team dedicated solely to this, just focused on building quantum algorithms and software libraries.
Today, researchers are hard at work testing new quantum algorithms to prove that quantum computers are well suited to tackling the theoretical problem spaces they’ve been hypothesized to be a good fit for. Their journey has been bumpy. There’s been a repeated pattern of someone finding an extremely fast quantum algorithm for a practical problem, and an equally fast classical algorithm is found shortly after. Formally, researchers use quantum complexity theory to try and rank which problems quantum computers will be uniquely effective at. However, the press releases can be almost whiplash inducing in terms of the amount of times a supposedly fantastic application has been found, only to be toppled a few months later by a classical algorithm equally or more effective.
NISQ Era Hardware
If there’s one more important thing to take away from this as an investor, it’s that we are currently in the NISQ era. John Preskill describes our current stage as the era of Noisy Intermediate-Scale Quantum (NISQ) computers. According to Preskill, this era is primarily defined by its limitations; in particular, contemporary quantum computers of only 50-100 qubits.
It can be hard to cut through the noise when navigating the hardware progress towards building a quantum computer, but the massive simplification is that hardware companies are all in a race to fault tolerance – simply a computer that can scale up qubits while cutting down on noise and error rate, and run indefinitely.
Currently there’s a race between a number of different approaches, and as I mentioned previously we’re still in the earliest stages. Current competitive companies are often categorized by the particles researchers are using to create the qubit themselves, and the cascading constraints imposed by said particles. Different companies are using electrons (superconducting qubit), ions (trapped ions), atoms, photons, and anyons. Of course, this is a massive simplification of even that one aspect of quantum computing development, let alone the necessary steps that will follow to create a quantum computer that is programmable at scale.
This matters because it means that, if we can make meaningful strides toward fault tolerance, we can compute meaningfully difficult problems with negligible errors. For example, researchers expect that finding the prime factor of big numbers will be exponentially faster with quantum computers. As it stands, many classical computers’ cryptographic systems rely on these calculations not being possible since they would take so long.
There are a few people approaching this by creating businesses that choose problems that have great likelihood of being more efficient on quantum computers then running research to test relevant algorithms out on all quantum computing hardware available to date, as well as classical – to then compare and effectively enter consulting agreements with large corporations interested in the respective problem space.
Notable Quantum Hardware Startups
- Anyon Systems (superconducting quantum computer)
- Atom Computing (neutral atom quantum computers)
- D-Wave Systems (building a quantum annealer)
- IonQ (trapped ion quantum computers – less need for low temperature but harder to scale)
- Quantum Circuits (smaller quantum computers that can be linked and might ease the difficulty of error correction)
- Rigetti (superconducting hybrid quantum classical computers)
- PsiQuantum (photonic quantum computer)
- Xanadu Quantum Technologies (photonic quantum computers)
State of the Market
Each of these hardware approaches have tradeoffs. Photon-based computing for example, like that of PsiQuantum, promises the ability to scale to 1 million qubits right away with necessary levels of error correction that holds other hardware approaches back, but – this is only after the photon-based computing companies efficiently producing small entangled states, which is a challenge in its own right.
Quantum hardware companies that are farther along also notably offer quantum compute as a service. D-wave, which was the first to the market with quantum annealers, purchased initially by Google and NASA, currently sells cloud compute, but they actually spin up CPU and GPU instances with AWS, only then to submit the quantum piece back to their own quantum processor.
Quantum Hardware Go-To-Market
This go to market is fascinating largely because although they’re first and foremost a hardware company, their proprietary approach of using the upside of quantum annealers and scaling the result on classical computers means that they’re able to offer the compute itself to customers rather than just the foundational hardware. The secret sauce here is the ability to utilize the upside of quantum annealing (greater speed on combinatorial optimization problems), while marrying this effectively with scaled classical compute power that is currently unavailable for quantum hardware.
Alibaba is also now offering quantum computing, and there’s news of other smaller companies in China that have notable progress, but it’s difficult to find information about their funding or other details of their systems.
Few of these companies, however, have a fully functional or large enough scale working computer. And even for those that do they’re only functional on “NISQ era” scale tasks, which we haven’t found large scale & effective use cases for. So where does this leave us? Hybrid quantum classical algorithms are a common medium term solution that utilizes aspects of both quantum (state preparation) and classical computing (optimization).
Quantum Software & Simulation
Quantum software is naturally the next area investors look to when discovering that the state of most quantum hardware companies is not as far as one might be led to believe. It’s important in that it will be the access layer of industry professionals to the fundamentals of quantum computing. Between now and fully fault tolerant computers we need to build all the layers of abstraction that exist within classical that make it easier for programmers to build.
Thus far one could classify this area of the market as two categories: hypothesizing what the software interface for hardware systems that largely don’t exist will be, and companies that are mostly investing in research on algorithms that work with quantum hardware.
Although at this stage quantum software still largely depends on hardware that is much more capable than we currently have, we see potential for highly vertical companies doing computational discovery for very specific industries.
Quantum Simulation for Molecule Discovery
Because of quantum’s ability to predict the properties of complex materials more accurately than the most advanced current methods can do at present, we’re especially excited about its ability to speed up molecule discovery. This includes materials, chemistry, and drug discovery, with the hope that these software layers create a seamless user experience when introducing less familiar uses to quantum. IonQ’s investor presentation which accompanied them becoming the first publicly traded quantum computing company also notably lists materials, chemical, and drug discovery as primary use cases of their quantum computers.
Scott Aaronson, a well regarded theoretical computer scientist known for his quantum computing research, summed up the state of the field well (76:00):
“By the way the biggest practical application of quantum computing that we know about by far… is simply the simulation of quantum mechanics itself in order to learn about chemical reactions, design, new chemical processes, new materials, new drugs, new solar cells, new superconductors, all kinds of things like that.” – Quantum Sensing
Notable Quantum Software Startups
Some companies to keep in mind in this space are:
- 1QBit (researching quantum algorithms esp for finance, materials science, and chemistry)
- Classiq (software platform for algorithm development)
- QC Ware (quantum computing as a service, working on ML especially)
- Q-Ctrl (quantum control engineering – for quantum computing and sensing)
- QSimulate (using quantum simulations for pharmaceutical and chemistry spaces)
- Strangeworks (development and education environment for quantum)
- Zapata – (task workflow composer for classical and quantum algorithms)
This space is difficult because the applications are generally relying on several hardware building blocks that don’t exist yet, as well as on simulation of quantum mechanics. Most importantly, many of these algorithms are reliant upon QRAM (quantum RAM) which is still purely theoretical. Another interesting indicator is Alphabet’s Sandbox Team working on using tensorflow processing units to simulate quantum computing workloads.
“Can we find a promising real-world application of quantum mechanics that exploit its most counterintuitive properties? Today, quantum computers and quantum cryptography are widely believed to be the most promising ones… Interestingly, however, this belief might turn out to be incomplete. “Quantum sensors” capitalize on the central weakness of quantum systems – their strong sensitivity to external disturbances. This trend in quantum technology is curiously reminiscent of the history of semiconductors: here, too, sensors – for instance light meters based on selenium photocells – have found commercial applications decades before computers.” – Quantum Sensing
Lastly, quantum sensing is an area that applies the same fundamentals of quantum mechanics, occasionally overlaps with quantum computing, and represents an area of future interest for us. Outside of quantum computing, quantum metrology/sensing most commonly uses quantum coherence or entanglement to improve the measurement that can be done with classical sensors. This area of study is certainly new, as atomic clocks have harnessed quantum mechanics for over a decade, the pace of research in quantum sensing for new applications has seen a massive uptick in the past few years.
Nitrogen Vacancy (NV) Centers in Diamonds as Sensors
Although there are many types of quantum sensors, NV sensors have seen a lot of use over the past 30 years. Since one of the most difficult parts of quantum in the real world is keeping a stable quantum state, diamonds present another uniquely strong and stable opportunity for holding an atom in a quantum state without interference, especially because of the diamonds’ photoluminescence. This means that by shining a laser into the diamond they can read the internal spin-state.
NV Magnetometry Startups
NV sensors exemplify the potential of quantum sensing today. Currently, startups Qnami and QZabre are building NV magnetometers, and exploring a wide variety of use cases. Their unique value is tied especially to their size, which allows them to be brought unusually close to a sample which improves the magnitude of magnetic field detection. Some include: brain imagining, improved GPS (by way of removed solar flare noise read out by magnetometers), measuring current within integrated circuits, general non-invasive detection of bio-magnetic or bio-electric signals, and studying magnetism within different material samples. The CSO at Qnami dives a bit more into the space here. We believe the precision sensing space represents potential for exponentially improved capabilities for medical and research capacity, and would love to invest in the space.
Conclusions & Open Questions
Somehow it seems like the further I go into quantum, the more uncertainty there is. There are many open questions, and although I typically like to write about rather specific things that seem promising, in the case of quantum, we are looking for:
- Companies focused on fundamental hardware adjustment and automating it as an initial wedge for long term ownership of the hardware market such as pulse shaping, which Q-Ctrl and Quantum Machines are notably working on.
- Companies with a novel approach to quantum algorithm research, especially when their expertise overlaps with an area of niche quantum simulation research that they’re focused on commercializing within.
- Companies using quantum to better predict the properties of complex materials, especially those with a plan to integrate vertically with those they’re selling to.
- Acquiring quantum talent is notoriously hard. What’s the future of exciting more people about quantum computing in a way that encourages them to enter the workforce – especially playful, educational tools and community spaces? Qbraid is one company that’s currently pioneering this space, and Michael Nielsen & Andy Matuschak’s Quantum.Country provides a foundational reader that incorporates spaced repetition for optimal retention.
- Companies with a novel approach to quantum sensing. I’m really interested in people tackling sensing quanta without the same sampling energy expenditure in terms of how we approach mining for rare earth materials on lunar surfaces.
As always — feel free to tweet or message me questions, thoughts, disagreements, or pitches on twitter or at firstname.lastname@example.org
Appendix & Educational Resources
Lastly, I want to provide a few educational resources that really helped me along the way, since I think well-made explainers are a true art form and should be shared with as many people as possible!
- Here’s a list of some of the people and organizations on twitter I watch in the space
- Lastly, here’s a well regarded open course from Harvard on Quantum Phases of Matter
- And a small grant program for people working on quantum research
- A collection of all sorts of white-papers, blog posts, and notes I took while researching
Corporate Partnerships & Quantum Brokers
On the larger corporation side, IBM, Google, Intel, and Microsoft are all also in a race to build quantum computers. On top of all of this, there’s a race for the lower-hanging fruit in terms of becoming the educational and strategic partner to those doing hardware or algorithms research – the brokerage race. It’s unclear to me whether big corporations’ race to being brokers of quantum compute is defensible or not – Amazon has a specifically dedicated quantum solutions lab launched to allow companies to work with experts to get ready for quantum, for example. It may make sense for these companies to build relationships with clients willing to pay in the future now since they have room to spend on sales and acquiring talent, but I imagine a large portion of their market dies out as smaller quantum hardware companies progress and find product market fit and are able to better educate their ideal customers.