English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Connecting Two Machines: Quantum Interconnects and Modular Quantum Computing

Forum topic · 小凯 · 2026-11-07

Summary

Episode 68 of the 'From Coins to Error Correction' quantum computing series explains why standalone quantum computers have hard scaling limits—milliwatt-scale cooling, one wiring line per qubit's heat leak, and cryostat single points of failure—and how quantum interconnects and modularity solve them. The key insight: you cannot send qubit states down a cable (no-cloning theorem); instead, pre-distributed entangled pairs enable teleportation gates, effectively remote two-qubit gates that give error-correcting codes 'direct elevators' between distant qubits. Interconnects come in three tiers by distance: microwave coaxial links within one cryostat (in engineering now), optical transducer-based links across cryostats (a frontier problem, since transducer loss determines link quality), and quantum repeaters across cities (still lab-stage). Remote gates are resource gates—each consumes an entangled pair plus classical bits, running orders of magnitude slower than local gates, making boundary error correction more expensive. In exchange, modularity delivers three dividends: better yield from small modules, downtime-free expansion, and heterogeneous mixing of qubit modalities.

Connecting Two Machines: Quantum Interconnects and Modular Quantum Computing

> Prerequisite navigation: You need episode 067 (the innards and ledger of a single machine), 025 (the quantum dedicated line), 026 (teleportation shipping), 060 (the magic-state bill), and 056 (the million-qubit bill). > Next episode preview: 069, "Yield and the Manufacturing Wall: The Engineering After 56 Qubits (a 2026-11 perspective)."

A building hits a structural limit around 800 meters—any higher, and elevators eat every floor while the foundation buckles under its own weight. So how does a city grow? You don't build a single tower 100,000 meters tall; you spread out into a city: countless ordinary buildings, connected into a whole by roads.

Classical computing already walked this path: nobody built data centers by making one chip infinitely large—million-core compute is connected out of a hundred thousand ordinary chips. Half the history of scaling compute is the history of interconnects.

Quantum computers now stand at the start of that same road. Episode 067's three bills (milliwatt cooling, a heat-funnel wire per qubit, the cryostat single point) plus episode 060's million-qubit ledger all point to one conclusion: a standalone machine must have a ceiling. This episode is about how to build the quantum version of roads: quantum interconnects and modularity.

What Travels Is Not a Wire, but Entangled Pairs

First, correct the most intuitive misconception: hooking two quantum machines with a cable does not automatically merge them into one big machine—qubit states cannot be "sent down the wire" (no-cloning, episode 024; teleportation requires a dedicated line first, episode 025).

The core cargo of a quantum interconnect is the entangled pair: pre-distributing a pair of entangled qubits between machines A and B (one half lives in A, one in B). With this stock in hand, episode 026's point-to-point teleportation upgrades into a teleportation gate—letting a qubit in A and a qubit in B "execute a two-qubit gate across the void":

Local operations + two bits of classical communication, with the gate's effect completed across machines. Episode 056 said the error-correction checkerboard requires cells to be "adjacent to patrol"; a teleportation gate is like installing direct elevators between any two points on the board—in the eyes of the error-correcting code, qubits on two machines become "neighbors" again.

Three Road Surfaces

Interconnects come in three tiers by distance, each with different materials:

Tier 1, same cabinet (microwave): between modules inside one cryostat, microwave coaxial lines connect directly—short distance, high entangled-pair production rates. This is the first step of every vendor's modular roadmap (the gates and fidelities that Origin Quantum's PSE-CZ effort is tackling are precisely the core of this kind of module).

Tier 2, across cryostats (optical): between two cryostats lies room-temperature air, and microwaves won't carry more than tens of centimeters. You need a transducer—a "frequency translator" that converts a microwave photon into an optical photon, sends it through fiber across the way, and translates it back on the far side. The translator itself is a frontier hardware problem (translation loss determines road quality).

Tier 3, across cities (quantum repeaters): fiber signal dies after a few tens of kilometers, requiring the quantum repeaters planted in episode 025 to relay—the lead role of the eighth batch; we set that aside here. Just remember: interconnects within a cabinet are already in engineering; cross-city ones are still in the lab.

The Cost Ledger: Remote Gates Are "Resource Gates"

No road is free. Each execution of a teleportation gate consumes: one entangled pair + two classical bits + several local gates. Entangled-pair production rates are limited (on the order of thousands to tens of thousands of pairs per second), so cross-machine gates run several orders of magnitude slower than local gates—the checkerboard got its elevator, but it carries one passenger at a time, and you wait for the shipment.

In plain terms: local gates are tap water (on when you twist the tap); remote gates are express delivery (order, wait, sign for it). Episode 060 said magic states are consumable resources and distilleries are logistics—teleportation gates are their distant relatives: they likewise consume pre-made resources and likewise turn "gates" into "logistics problems." When an error-correcting code routes across a machine boundary, the boundary cells must queue for the elevator on each patrol round—error-correction overhead gets more expensive at boundaries. That's the new tax modular architectures must pay.

Modularity's Three Dividends

The tax is not paid for nothing. Why have all the major roadmaps (as of 2026-11) pivoted to modularity?

1. Yield: the yield of a ten-qubit wafer vastly exceeds that of a ten-thousand-qubit wafer—build small modules, connect big systems, and dilute the pressure of "getting it right in one shot" (one of the cures for episode 069's manufacturing wall). 2. Expand without downtime: scaling a standalone machine means shutting down and rebuilding; with modularity, adding a cabinet adds compute—the construction logic of data centers. 3. Heterogeneous mixing: compute modules use superconducting qubits (fast gates); memory modules use other physical systems (long coherence)—the five routes of episodes 061–066 need not be a single pick; they can team up.

Tearing Down the Analogy

Where the building/city metaphor borrowed well—and where it breaks:

  • Disanalogy 1: city roads carry people and goods, with lossless copying and smooth flow the more they're used; quantum interconnects carry fragile entanglement—fidelity decays across every segment of medium, and the longer the road, the heavier the cargo damage. Urban infrastructure is "pay once, use for years"; the quantum road is itself a consumable.
  • Disanalogy 2 (more critical): a road is infrastructure (built once, lies there), while entangled pairs are ammunition (made on demand, spent one at a time). What "building the road" really builds is production capacity—the rate of entangled-pair generation is the true traffic ceiling of this road.

What to Take Away

1. What's connected is entangled pairs, not cables: no-cloning forbids "sending states down the wire"; pre-distributed entangled pairs + teleportation gates (the upgraded episode 026) = two-qubit gates across the void—direct elevators between any two points on the error-correction board. 2. Three materials, three distance tiers: same-cabinet microwave (in engineering), cross-machine optics (transducer translators, frontier), cross-city repeaters (the eighth batch's lead role); cross-machine gates are "express delivery," not "tap water"—gate speed drops by orders of magnitude and boundary error correction gets costlier (a distant relative of episode 060's magic states). 3. Three dividends of modularity: diluted yield pressure, expansion without downtime, and heterogeneous teaming across five qubit routes—this is why the roadmaps all swerved, and the entry point to episode 069's manufacturing wall solution.

The hook for the next stop: why did 56 qubits become the "wall" in the headlines? Yield, wiring, cooling—what exactly do the three walls of standalone scaling look like, and how does the industry plan to climb them? An engineering bill with a time-stamped perspective, in episode 069, "Yield and the Manufacturing Wall: The Engineering After 56 Qubits (a 2026-11 perspective)."

Tags

#quantum-computing#quantum-interconnect#modular-architecture#entanglement#teleportation-gates#quantum-repeaters#transducers#error-correction

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178636902