Imagine a systems architect at 3 AM, staring at a CAD drawing of a data center that looks less like a server farm and more like a particle accelerator. He isn’t thinking about Kubernetes or latency; he’s wondering how to keep a thousand lasers perfectly synced across a room that costs more than some small towns. He knows that one stray vibration—a heavy truck driving by outside or a poorly timed sneeze—could turn his entire compute cluster into a very expensive light show.
This is the reality of the photonic approach to quantum computing. While the rest of the industry has spent a decade obsessing over superconducting qubits that need to be kept colder than deep space, PsiQuantum is betting on light. The idea is to use photons to carry information, which theoretically avoids the thermal noise that plagues the IBMs and Googles of the world. According to MIT Tech Review, the goal is a machine with a million qubits.
The sheer scale here is the point. We’ve reached a plateau with the “small” quantum machines. Having 50 or 100 qubits is a neat academic exercise, but it doesn’t actually do anything that a beefy NVIDIA H100 cluster can’t simulate or approximate. To get to actual utility—the kind that breaks RSA or simulates new catalysts—you need a million qubits. (Or so the marketing deck claims).
The problem is that building a million-qubit machine isn’t just about adding more hardware; it’s about the physics of the interconnects. Superconducting qubits are bulky and temperamental. Photonics, however, allows for the use of standard semiconductor manufacturing. You can etch your quantum circuits onto silicon wafers using the same fabs that make the chips in your phone.
But don’t mistake “standard manufacturing” for “easy.” The precision required is absurd. We are talking about controlling light at a level where the slightest misalignment renders the qubit useless. It’s like trying to balance a thousand spinning plates while walking a tightrope. Who actually believes a room full of lasers is going to behave consistently over a production cycle?
Here is where we take a stand: the photonic path is the only one that doesn’t end in a dead end. The superconducting route is a luxury project for companies with infinite budgets and a fondness for liquid helium. It’s a hardware cul-de-sac. If you want a machine that actually fits in a data center (even a very large one), you have to move toward light.
Of course, the friction is immense. The cost of the laser arrays and the power required to maintain coherence across a massive photonic chip is a nightmare. We aren’t talking about a rack-mount server; we are talking about a specialized facility that requires its own power substation.
Still, the logic holds. By moving the computation into the photonic domain, you bypass the most grueling part of the quantum puzzle: the cooling. You still need some cooling, but you aren’t fighting a losing battle against the laws of thermodynamics every time you add a qubit.
It’s a massive gamble.
If this fails, it fails because of the engineering, not the theory. The theory is sound, but the gap between a wafer and a working computer is a canyon. However, if the manufacturing holds up, the jump in capability will be vertical.
We will see a public demonstration of a photonic cluster that maintains coherence across more than 100 logical qubits by Q4 2027. If they can’t hit that mark, the million-qubit dream is just a way to keep VCs interested while the physicists figure out why the lasers keep drifting. Until then, it’s just a very expensive way to move light around a room.
