By: Thomas Stahura
Last December, Google unveiled Willow, its new quantum chip. The media dubbed it mind boggling when, with only 105 qubits, it was able to solve a problem in five minutes that would take a classical computer ten septillion years to complete. That problem is called Random Circuit Averaging (or RCA) which I’ll explain in a bit.
On the news, Google’s stock jumped, as did its competitors: Microsoft, Rigetti, D-Wave, and IONQ. For a moment, it seemed, quantum hype dethroned AI to become the talk of the town. Two months later, Microsoft responded by announcing its own quantum chip called Majorana 1, causing another stock bump. However, at only 8 qubit, it's still early stage, and the tech giant has yet to publish its RCA results.
RCA is a benchmark, not a problem, as such is designed to gauge quantum computer performance. The whole test is basically "can you sample from this crazy quantum distribution faster than classical computers can even calculate what that distribution should be?"
To do this, researchers must:
Pick a number of qubits (like 105 for Google's chip)
Generate a random sequence of 20+ random quantum gates (like Hadamard or Pauli-X mentioned last week)
Run information through the more than 20 gate layers a million times or so
Collect each runs generated bitstring output (something like "01101001...")
Use classical computers to simulate what the "perfect" quantum computer would output
Measure how close your actual results are to the ideal
Each additional gate creates more quantum entanglement between qubits. More layers = more complex quantum correlations = harder for classical computers to track. More than 20 layers is where classical simulation becomes practically impossible. If a quantum computer finishes in minutes but classical takes years → quantum advantage. At least, that's how the thinking goes.
RCA is cool but not practical. It's like saying your AI passed the MENSA iq test. So where are the real world quantum applications?
Enter the wonderful world of optimization and quantum annealing!
Classically, annealing is an algorithm inspired by metallurgy: you heat up a material and then cool it slowly so atoms settle into a low-energy (optimal) state. In the math world of optimization, you randomly explore solutions, occasionally accepting worse ones to escape local minima, and gradually “cool” to settle into the best solution.
Imagine you’re standing in a vast, foggy landscape of rolling hills and valleys. Each point in this landscape represents a possible solution to your optimization problem. The height at each point is the “energy” of that solution — the lower the energy, the better. Classical annealing is like wandering this landscape with a lantern. At first, you’re allowed to take big, random steps, even uphill, so you don’t get stuck in a small valley (local minimum). As time goes on, you “cool down,” and your steps get smaller and more cautious, focusing on moving downhill. The hope is that, by the end, you’ve found the deepest valley, the global minimum. The catch? Sometimes, no matter how clever you are at wandering, you can still get stuck in a valley that isn’t the lowest one (not optimal). The fog is thick, and you can’t see the whole landscape at once.
Quantum annealing replaces random steps with quantum tunneling, allowing the system to “tunnel” through energy barriers rather than climb over them. In our example, instead of just walking over the hills, you can tunnel through them to a lower valley on the other side, even if it looks impossible from a classical perspective. Essentially, thanks to quantum mechanics, quantum tunneling can help escape local minima that would trap a classical algorithm.
Without getting too technical, quantum annealing does not use any quantum logic gates! Instead, an optimization problem is encoded as a Hamiltonian (fancy math representing the system's total energy). This sets up the energy landscape so that the lowest energy state (the ground state) represents the best solution to the problem. Then (thanks to quantum physics), the system naturally wants to stay in the lowest energy state and naturally “relaxes” into the answer.
Companies like D-Wave, founded in 1999, are leading the charge in quantum annealing. D-Wave’s Advantage system, accessible via its Leap cloud platform, has been used by the likes of Volkswagen to optimize traffic flow and by SavantX to streamline port operations, reducing costs and improving efficiency. D-Wave charges a subscription for cloud access and consulting services. In 2024, D-Wave reported contracts with major firms, contributing to its growing commercial traction.
Similarly, IonQ, which runs a quantum computing manufacturing facility up in Bothell, operates primarily as a Quantum-as-a-Service model, providing access to its quantum computers via major cloud platforms like AWS, Azure, and GCP. The company was founded in 2015 and became the first quantum company to IPO back in 2021.
Beyond optimization and the cloud, quantum computing is making inroads in drug discovery and materials science. For example, Algorithmiq’s collaboration with IBM’s Quantum Network focuses on quantum chemistry simulations to identify promising drug candidates, potentially shaving years off development timelines. Generating revenue through partnerships and licensing their software platforms. Algorithmiq secured €13.7 million in funding to scale its offerings.
Quantinuum is also working with firms like Samsung to apply quantum algorithms in materials design, optimizing material properties for semiconductors and batteries. These early applications, still in the prototyping phase, are driving real revenue through research contracts and pilot projects.
Quantum applications are hitting the market and making money. We now have enough qubits to do cool things! It feels like the bottleneck is shifting from hardware to software. The industry needs more quantum developers to build the next generation of algorithms and apps. Or maybe develop an AI that can program in Q# or the other quantum languages. On the hardware side, things are starting to get crowded: Xanadu, Alice & Bob, Atom Computing, PsiQuantum, Rigetti, NVIDIA, QuEra Computing, and Intel, just to name a few, are all developing their own quantum computers.
I think we'll see much more change in the quantum industry in the next 30 years than the last 30. Again with most of that change coming from innovative software.
Stay tuned next week for the final installment of our quantum series!
P.S. If you have any questions or just want to talk about AI, email me! thomas @ ascend dot vc