Pete Shadbolt says no useful quantum computer exists today, yet PsiQuantum is building a million-qubit machine by decade's end.
Shadbolt argues AI cannot solve the hardest chemistry problems, and drug and battery chemistry customers will need quantum's first-principles calculations.
PsiQuantum builds its machine like a conventional computer using semiconductor supply chains, with photons networked over telecom fibre.
PsiQuantum bets a million photonic qubits can beat AI
PsiQuantum is building a photonic quantum computer the size of a supercomputer, betting that useful quantum computing starts at about a million qubits with no small-scale shortcut. Photons were chosen because they network over telecom fibre and fit the semiconductor supply chain. Shadbolt's case for patience rests on AI's limits in chemistry and markets finding the uses.
Takeaways
A million qubits is the threshold for usefulness. Shadbolt says the scientific consensus for twenty-five years has been that about a million qubits are needed for impact, and no known applications are worth doing on small machines. PsiQuantum went straight for the large machine because it is scientifically correct.
AI cannot solve the hardest chemistry problems. AI learns approximations from data and falls down where high precision is needed. Drug discovery and battery chemistry customers use AI heavily but say they will ultimately need first-principles calculations, which Shadbolt says only quantum can provide.
Photons scale and fit the chip supply chain. A million qubits cannot fit on one chip, so they have to be networked, and photons link over standard telecom fibre without transduction at 99.7% fidelity. PsiQuantum builds in regular fabs on a 45-nanometre process Shadbolt calls ancient by TSMC standards.
Quantum will never be a mass adoption technology. Only hundreds of people will ever touch it directly, like ASML's EUV tools or a TSMC tape-out, with value flowing downstream into drugs, fuels, catalysts and fertilizers. Australia is investing $1bn of taxpayer money.
Quantum will be geopolitically significant. Shadbolt says power is captured on the frontier, and ten years ago no politician cared about a fab while now they all know TSMC is in Taiwan. Frontier models need billions of dollars of GPUs, and Shadbolt expects quantum to follow the same pattern.
The market will decide quantum's applications. Shadbolt says capitalism will turn quantum overwhelmingly into a good thing, and civilian applications are broader and more commercially lucrative than national security ones. Sustained serious payment is a good signal that a use is useful.
Building the largest molecular beam epitaxy tool
PsiQuantum pitched investors including BlackRock and Temasek to fund a molecular beam epitaxy tool larger than any built before.
The tool had to hold molten titanium at a few thousand degrees Celsius and liquid helium at minus 270 degrees in the same vacuum chamber.
Half the engineering team told leadership the plan was insane, and during COVID the shipping company damaged the tool in the Alameda Bay.
Tap a name to go there · tap a line to see why they are linked
On this episode 3
Show 1
The team still delivered ahead of schedule.
What comes next
PsiQuantum has three cryostats running and three racks installed, and is loading thousands of photon sources into its Alpha system.
Next-generation systems will be built in Chicago and Australia.
Most teams are aligning on roadmaps toward the end of the decade, though Shadbolt says not everyone should be expected to reach useful quantum computing and various efforts will be delayed.