# Can All-Photonic Repeaters Deliver a 1,000km Quantum Network?

A new architecture from University of Massachusetts Amherst demonstrates that 1,000km quantum communication is achievable using all-photonic repeater stations spaced just 9km apart, requiring only a few thousand GKP qubits per node — compared to prior third-generation designs that demanded up to 42,000 GKP qubits or 16,400 single photons for equivalent performance. That resource reduction is the headline result: it moves all-photonic quantum networking from a theoretical curiosity to something that can be seriously costed and planned.

Ryosuke Shiina and colleagues at UMass Amherst, in collaboration with Manning College of Information and Computer Sciences and Photon Queue Inc, published the architecture on June 28, 2026. The system combines continuous-variable and discrete-variable quantum error correction — specifically the bosonic Gottesman-Kitaev-Preskill (GKP) code and the Steane code — to fight photon loss simultaneously at two levels of the encoding hierarchy. The team achieved a squeezing level of 15.4 dB, which directly governs how well GKP states hold their fidelity against loss-induced noise.

The 9km repeater spacing is presented as the maximum demonstrated with this specific architecture. Future work will target realistic fibre conditions: imperfections, temperature fluctuations, and the operational overhead of the discard windows used to filter corrupted GKP states.

---

## Why Photon Loss Has Blocked 1,000km Quantum Links

[Entanglement](https://quantumintel.tech/glossary/entanglement) distributed over optical fibre degrades exponentially with distance. Absorption and scattering in the fibre continuously bleed photons from the channel, and because quantum states cannot be amplified classically — the no-cloning theorem forbids it — signal recovery requires genuine quantum error correction at intermediate nodes, not classical signal boosting.

Third-generation all-photonic repeater proposals addressed this by encoding quantum information redundantly across many [photonic qubits](https://quantumintel.tech/glossary/photonic-qubit), then performing error correction at each node without storing quantum states in matter-based memories. The trade-off has always been resource cost: earlier designs required up to 16,400 single photons or 42,000 GKP qubits per station to hit comparable performance targets. Those numbers make real-world deployment prohibitively expensive.

The UMass Amherst architecture attacks that cost directly. GKP qubits encode quantum information in the quantum state of a harmonic oscillator — a continuous-variable approach that provides inherent robustness against small displacement errors caused by photon loss. The Steane code, operating at the discrete-variable layer, protects against bit-flip and phase-flip errors. Stacking these two complementary QEC strategies creates a synergistic effect: each code corrects error types the other handles less efficiently, reducing the total qubit overhead needed to clear the [error threshold](https://quantumintel.tech/glossary/error-threshold) for 1,000km transmission.

The discard window mechanism is an important practical detail. The architecture selectively accepts only GKP states that fall within a defined parameter range — effectively rejecting qubits too noisy to correct reliably. This trades some throughput for substantially higher fidelity, a worthwhile exchange when the alternative is propagating errors that corrupt the entire entanglement chain.

---

## The 15.4 dB Squeezing Figure and What It Means

Squeezing level is the operational parameter that most directly connects this architecture to what hardware must actually deliver. At 15.4 dB, the design reduces quantum noise in one electromagnetic field quadrature substantially enough to maintain qubit fidelity across 9km segments. Current state-of-the-art laboratory squeezers can approach and in some cases exceed this level under controlled conditions, but sustaining it within a deployed fibre network — subject to thermal noise, mechanical vibration, and connector loss — is a different engineering challenge.

This is the gap the team acknowledges: the 9km figure is the maximum demonstrated with this architecture under the modeled conditions, and realistic network deployment will require further optimisation. That caveat should be front of mind for any infrastructure planner reading this paper as a deployment blueprint rather than a research milestone.

---

## Resource Comparison: What the Numbers Actually Say

The reduction from up to 42,000 GKP qubits (or up to 16,400 single photons in alternative prior designs) to a few thousand GKP qubits per station is not marginal — it is roughly one order of magnitude. For a 1,000km link with 9km spacing, that implies roughly 111 repeater stations. At a few thousand qubits each, the total photonic qubit count for the repeater chain sits in the low hundreds of thousands. Prior designs at 42,000 qubits per node would have pushed that past four million.

That distinction matters enormously for the commercial trajectory of quantum networking. Photonic integrated circuit manufacturers and quantum photonics startups currently working on scalable [photonic qubit](https://quantumintel.tech/glossary/photonic-qubit) platforms face a very different engineering target at a few thousand qubits per node versus tens of thousands. The former is within the roadmap horizon of several active hardware programs; the latter is not.

---

## Industry Trajectory

The broader quantum networking field has been searching for a third-generation repeater design — one that requires no quantum memories, operates purely on photons, and can be deployed in existing fibre infrastructure — since the theoretical framework was established. This UMass Amherst architecture is an academic result, not a deployed system, and Photon Queue Inc is a small collaborator rather than an established photonic hardware manufacturer. The paper does not report experimental hardware demonstrations; it presents a modeled architecture with calculated performance parameters.

What it contributes is a credible resource-reduction target. The field now has a design that fits a few thousand GKP qubits per node at 9km spacing, achieves 15.4 dB squeezing as a performance anchor, and uses a well-understood combination of GKP and Steane codes. That gives hardware teams a concrete specification to engineer toward rather than an open-ended requirement.

For [fault-tolerant quantum computing](https://quantumintel.tech/glossary/fault-tolerant-quantum-computing) architectures that envision distributed quantum processors connected over metropolitan and intercity distances, this is the kind of networking result that belongs in roadmap planning — with the caveat that the path from modeled architecture to fielded infrastructure typically takes years and encounters loss budgets that simulations underestimate.

---

## Key Takeaways

- **1,000km quantum communication** is theoretically achievable with all-photonic repeaters spaced 9km apart, per the UMass Amherst architecture
- **Resource reduction is the core result**: a few thousand GKP qubits per station versus up to 42,000 GKP qubits or 16,400 single photons in prior third-generation designs
- **15.4 dB squeezing** is the achieved noise-suppression level underpinning the fidelity performance
- **GKP + Steane code combination** provides synergistic error correction, addressing continuous-variable photon loss and discrete-variable bit/phase-flip errors simultaneously
- **Discard windows** filter corrupted GKP states below a defined threshold, trading throughput for fidelity
- The architecture is **modeled, not yet experimentally demonstrated** in hardware; realistic fibre conditions remain to be benchmarked
- Collaborators include Manning College of Information and Computer Sciences and **Photon Queue Inc**

---

## Frequently Asked Questions

**What makes the UMass Amherst quantum repeater architecture different from previous designs?**
Previous third-generation all-photonic repeater proposals required up to 42,000 GKP qubits or up to 16,400 single photons per station for equivalent performance. The UMass Amherst design, combining GKP and Steane error correction codes, reduces that to a few thousand GKP qubits per node while maintaining 1,000km range with 9km repeater spacing — roughly an order-of-magnitude improvement in resource efficiency.

**What is a GKP qubit and why does it matter for quantum networking?**
GKP (Gottesman-Kitaev-Preskill) qubits encode quantum information in the quantum state of a harmonic oscillator — a continuous-variable approach. They are inherently robust against small displacement errors caused by photon loss, making them well-suited to optical fibre transmission where photon loss is the dominant error mechanism.

**Has this 1,000km quantum repeater architecture been demonstrated in hardware?**
No. The UMass Amherst result is a modeled architecture with calculated performance parameters, not an experimental hardware demonstration. The team acknowledges that future work must address realistic conditions including fibre imperfections and temperature fluctuations.

**What is the significance of the 15.4 dB squeezing level?**
Squeezing reduces quantum noise in one quadrature of the electromagnetic field, directly improving the signal-to-noise ratio for GKP qubit transmission. The 15.4 dB figure achieved in this architecture represents the noise-suppression performance required to maintain qubit fidelity across 9km repeater segments — a level that is challenging but not unreachable for current laboratory squeezers.

**Why does repeater spacing of 9km matter practically?**
Shorter repeater spacing means more stations but easier per-segment fidelity requirements. At 9km, a 1,000km link requires roughly 111 repeater nodes. The key question for deployment is whether the per-node resource cost — a few thousand GKP qubits — can be manufactured and integrated at that scale. This architecture's resource reduction makes that question answerable with near-term photonic hardware roadmaps in a way that prior designs did not.