Has Photonic Quantum Computing Solved the Scalability Problem

[Glenn]

Xanadu's Aurora system got a lot less attention than Microsoft's Majorana announcement but the technical case for it being significant is arguably stronger, since the results are published in Nature. Aurora is 12 physical qubits spread across 35 photonic chips in four server racks connected by 13 kilometres of fibre optic cable, all operating at room temperature. The key claim is that Xanadu has solved the scalability architecture problem. You can in principle add more racks using the same interconnect design and scale toward millions of qubits without redesigning the core hardware. Xanadu's CEO Christian Weedbrook said bluntly that photonics is the best and most natural way to both compute and network, and that the remaining challenges are performance rather than scalability. The company went public on Nasdaq and the Toronto Stock Exchange in March 2026 under XNDU and has a market cap around 5 billion dollars.

The catch that Weedbrook also acknowledges is optical loss. Every fibre connection, switch and connector loses photons and lost photons are lost information. Xanadu has reported a 60 percent reduction in optical loss in recent work and has supply partnerships with Thorlabs to chase further reductions, but this is an engineering challenge that is far from solved. The 100,000 physical qubits needed to achieve 1,000 logical qubits by 2029 is a very different proposition from demonstrating that four networked racks can share quantum state reliably. Still, a room-temperature system with a published Nature paper and a credible scaling architecture is a different proposition from a lab result that nobody can reproduce.

Photonics versus superconducting versus trapped-ion versus topological: which architecture do you think actually gets to fault-tolerant quantum computing first, and why?

[KeyboardWarrior47]

The Nature publication is the key differentiator here. Aurora's results have been through peer review. Microsoft's Majorana 2 20-second lifetime claim has not. That asymmetry matters

[Pixel Mark]

Room temperature operation is massively underrated as an engineering advantage. The cooling infrastructure for superconducting qubits is expensive, complex and a scaling bottleneck in itself

[Anthony]

60 percent reduction in optical loss sounds good until you realise how much loss remains. Photonics has always promised scalability and always run into the loss wall at scale

[Hannah56]

Four racks and 12 qubits published in Nature versus Microsoft's 12 qubits and extraordinary stability claims. In 2026 they are genuinely at similar scales despite taking completely different approaches

[Coastal Otter]

Xanadu targeting 1,000 logical qubits by 2029 at a 100-to-1 physical-to-logical ratio means they need 100,000 physical qubits. That is an enormous jump from where they are today

[AntMan]

The fact that Xanadu is publicly listed is interesting. You can actually look at their financials now. 188 percent revenue growth to 4.6 million dollars with a 70 million dollar loss tells you this is still deep research spend

[RandyOrton]

PennyLane growing 161 percent in adoption is the less glamorous but possibly more commercially significant number. Software ecosystem lock-in matters as much as hardware progress in the long run

[Danny47]

I would not write off superconducting just because photonics has a scalability story. Google and IBM have far more qubits operational today and the error correction work on surface codes is very mature compared to photonic approaches

[NeutrinoX]

The fact that none of these four architectures has a clear lead in 2026 suggests the race is more open than the coverage implies. A dark horse could still emerge from a completely different direction

[WhatUQuant]

DARPA selecting Xanadu for Stage B of the Quantum Benchmarking Initiative is a meaningful signal. DARPA does not advance companies on press releases, they require demonstrated results against defined metrics