China’s Jiuzhang Computer is Flashy, but PsiQuantum’s Paperwork Shows It Wants to Be More Than a Fad

In early December 2020, a group of researchers based in Hefei, China announced that their quantum computer, called Jiuzhang, managed to perform in 200 seconds what the fastest supercomputer in the world, a Chinese supercomputer called TaihuLight, would require 2.5 billion years to perform, demonstrating quantum supremacy more than 1012 times over. The catch, however, is that the only operation capable of being performed by Jiuzhang is Gaussian boson sampling, which involves detecting a number of photons emitted from a linear interferometer. While counting particles sounds like a relatively simple task, actually solving this problem requires calculating on the order of 1016 probability amplitudes. Classical computers calculate these probability amplitudes -by no means simple to evaluate – one at a time, so it’s easy to see how this could add up.

The Hefei team utilized a linear optical quantum computer, which utilizes photons as qubits, rather than the nanoscale superconductors used by Google or electrons used by other researchers. Essentially, the team took advantage of the fact that boson sampling is uncharacteristically simple to perform using linear optical quantum computers but essentially impossible to evaluate with classical computers, making it an ideal showcase. Additionally, due to the relative simplicity of the problem, the set didn’t require any of the complications required for linear optical quantum computers that seek to perform a wider range of tasks, such as ancillas (essentially extra bits to assist in computation) or complex methods to achieve and maintain entanglement between the qubits. Don’t expect Jiuzhang to be able to optimize traffic management schemes or model the shifts in electron structure during a chemical reaction.

If Jiuzhang isn’t reprogrammable, why care? While it’s true that solving boson sampling isn’t particularly useful and the Hefei team was beat to the punch on quantum supremacy by Google over a year ago, the demonstration represents a proof of concept for optical computing methods and will likely signal a renewed interest in the field. For investors, the thing to watch for is acquisition of photonic supercomputing technology by one of the major players. Google’s superconductor technology appeared to work great, but there is a major scalability issue. The superconductors require extremely low temperatures in order to operate. As quantum computers grow, both the difficulty and cost of maintaining near absolute zero temperatures is going to skyrocket, potentially putting a cap on development until this issue can be solved. Compared to superconductors, photons are extremely insensitive to temperature fluctuations, meaning that optical computers don’t face similar problems.

Enter PsiQuantum. PsiQuantum may not have the volume of patent documents of companies like D-Wave Systems and Google, but they dominate the optical computing space, with 11 issued patents and another 13 applications still pending. In a previous insight, the Patent Forecast® highlighted PsiQuantum’s ambition in building a one million qubit system, which it believes is the minimum number in order to create a genuinely useful quantum computer. In April of 2020, PsiQuantum raised a massive $150 million from Atomico, bringing their total funding to $230 million, so they are well positioned to capitalize on this critical moment for optical computing. Right now, none of the major players in the sector – Google, Microsoft, Intel, or IBM – have shown any interest in developing optical computing patents, but it’s just a matter of time until one of them comes knocking on PsiQuantum’s door. The questions are who gets there first, how much are they willing to spend, and whether PsiQuantum will sell at all. 

For more on trends on the cutting edge of computing technology, subscribe to the Quantum Computing Patent Forecast®!