Kipu Quantum: Shaping the Future of Quantum Algorithm Compression
Quantum computers promise to perform some computations exponentially faster: Computations that may take a classical computer 10,000 years may run within hours on a quantum computer. Yet, building universal quantum computers turned out to be notoriously difficult.
Current noisy intermediate-scale quantum computers are limited by the number of processing units, so-called qubits, and by noise—every step in a quantum computation adds a little error that accumulates until no calculation results can be obtained anymore. This severely limits the number of steps one can perform.
The startup Kipu Quantum found a way to compress quantum algorithms, i.e. to arrive at the same result with fewer steps. This makes quantum computing useful earlier compared to other approaches, as it requires machines with fewer qubits to run these quantum algorithms and solve problems of interest to industry users. Founded in 2021 by Enrique Solano, Daniel Volz, and Tobias Grab in 2021, it has raised a €3 M seed round from Quantonation, Entrada Ventures, and First Momentum Ventures.
Learn more about the future of quantum algorithm compression from our interview with the CEO, Daniel Volz:
Why Did You Start Kipu Quantum?
I am a chemist, and I really like this discipline—it’s one of the most worthy disciplines to study as it’s shaping the material world around us. Yet, many workflows in chemistry are purely empiric, meaning a lot of trial and error, which is holding back progress. And this goes not only for product development but also later on for upscaling and mass production: You change a minor thing in the process, and the product’s properties may change completely. Best if you never change a running system.
However, modern challenges around the green energy transition, replacement of fossil fuels, and sustainability, in general, will trigger the chemical industry to change profoundly, establish new processes, and be more agile. One part of that is digitization and using computers to simulate chemical processes upfront, which saves a lot of trial and error in the lab.
While established computational chemistry tools only go so far, quantum computing offers the promise to shake things up from the ground—even if just a fraction of quantum computing’s promises were to materialize, it would still have a profound positive impact.
How Does Quantum Algorithm Compression Work?
Since the 1980s and 90s, science has shown on paper that quantum computers could vastly outperform classical computers for certain computations. Yet, no one knows when this vision of a universal quantum computer will come true, if ever. As of today, quantum computers are small-scale devices, mostly good for fancy tricks but not quite for solving real-world problems.
At Kipu Quantum, we’re striving to make quantum computers useful earlier, not at the hardware level, but by providing shorter quantum algorithms. We’re applying the GPU playbook to quantum computing by tailoring quantum algorithms to the specific quantum computing hardware:
First, this involves a hardware-independent compression of the quantum algorithms, i.e., reaching the same goal with fewer computational steps. Quantum computing in 2022 is all about getting somewhere with a limited number of steps, as every step accumulates a small error. Eventually, the quantum computer falls out of coherence, losing its quantum properties that underpin quantum computing in the first place. We found a universal compression algorithm to make quantum algorithms shorter.
The second step is hardware-specific, tailoring the compressed quantum algorithm to the respective hardware platform. While quantum gates elegantly describe quantum algorithms in a universal way, they can’t be executed straightforwardly. Nowadays, performing an actual quantum computation still involves tweaking and twisting ions, lasers, and their excitations, which requires a decent understanding of quantum physics. Yet, it makes a quantum computation even shorter and more efficient.
Given IBM’s roadmap for scaling quantum technology, as well as similar announcements by other players, we expect quantum computers involving a couple of thousands of qubits to be around in 2025. By compressing quantum algorithms to involve fewer steps and thus fewer qubits, we’ll make useful quantum algorithms run earlier on, even if quantum computers comprise just a couple of hundreds or thousands of qubits. There we are following a portfolio approach, working with different quantum hardware manufacturers, prospecting which one will become large and useful enough first.
How Did You Evaluate Your Startup Idea?
I have been in quantum computing since 2018, advising corporates at McKinsey about the need to prepare for quantum computing in advance. Most still think quantum computing is a long way out; most haven’t even caught up on using machine learning for advanced data analysis. I then worked at BASF devising its quantum strategy internally, figuring that quantum computing has been promising the world but achieving very little over the past decades.
There I realized that there has to be clear evidence for specific applications that quantum can make a difference and eventually change the world—but where to start? We have explored different use cases at Kipu Quantum around molecular modeling, determining chemical reactivity, catalysis, corrosion research, and charge transport in batteries.
Another promising area is quantum machine learning, not necessarily running machine learning algorithms on a quantum computer but rather using quantum neurons as part of a neural network e.g. for better image recognition, especially when data sets are sparse and costly to obtain, such as for agriculture. There’s no one silver bullet, so we’re exploring a portfolio of potential use cases.
What Advice Would You Give Fellow Deep Tech Founders?
Keep in mind that explaining your technology is a crucial part of the game. Academics, especially in Europe, seem to think that giving talks with lots of jargon and making others feel stupid has some merit in itself, but this is not the case.
During my studies, I worked as a journalist in science communication for a local newspaper in Karlsruhe, financing my studies by explaining complex topics to laypeople. If I wanted to pay my rent, I had to have empathy for the audience and be very understandable—the same holds if you want to raise a funding round from investors. Most investors and customers are well-intentioned and would love to support you, but you need to meet them halfway. Even if they don’t care in the end how you solve a particular problem, as so often in industry, you need to help them understand your technology easily—people get comfortable around what they understand.