Quantum Keys Move Onto Production Routers

Quantum Keys Move Onto Production Routers

I co-led the telecommunications curriculum for Verizon's Next Step New England program and directed National Science Foundation, or NSF, funded Centers of Excellence at Springfield Technical Community College and the University of Central Florida through the transition from my Dad's Plain Old Telephone Service, or POTS, landlines to Internet Protocol, or IP, based voice, video, and data over copper, fiber and wireless. Our center worked closely with Cisco through that transition. The physics and the protocols changed rapidly snd it was a wild ride. What made the transition real was not the standards documents. It was carriers running the new transport on switches and routers in the central office, and technicians who did not need an advanced degree to keep it running.

Quantum key distribution, or QKD, is a way to generate an encryption key using the behavior of individual photons instead of math. Two machines exchange specially prepared light particles over fiber. If anyone taps the line and looks at those particles, the particles change in a way both ends can detect. That gives you a key exchange where eavesdropping does not stay hidden, which is a different from anything conventional encryption offers.

Aliro Technologies, the Vienna based quantum networking firm zerothird, and Cisco just ran a live version of this over Cisco's production routers at Cisco's Photonics Center in Vimercate, Italy. The hardware was Cisco's 8000 Series routers, the same platform Cisco sells into data centers today. That detail is the news. QKD has existed in labs for years. Running it on hardware a customer can already buy is the harder problem.

The system runs on the BBM92 protocol, which uses paired entangled photons rather than a transmitted key to establish a shared secret. Entangled photons are pairs of light particles created together so that measuring one instantly tells you something about the other, no matter the distance between them. A source creates these pairs and sends one photon from each pair to each end of the link. Both ends measure what arrives and use those measurements to build an identical key, without the key itself ever traveling across the fiber. zerothird supplies the hardware that does this: the photon source, the equipment that keeps the light polarized correctly, synchronizing clocks, and the software that cleans up errors and strengthens the final key. Aliro's Orchestrator software sits on top and manages the link, the way network management software already watches a conventional router. It tracks error rates and photon counts in real time and can reroute traffic or shut a link down safely if something looks wrong. The finished keys reach the routers through Cisco's Secure Key Integration Protocol, a standard interface, where they secure encrypted sessions between routers the same way a conventional key would, just generated a different way.

Diagram Gemini AI Generated

Chapter 1 of Quantum from the Ground Up covers the fiber problem in quantum networking through the University of Illinois work on ytterbium-171 emitters built for existing telecom infrastructure. That chapter is about getting a quantum signal onto fiber that already exists. This deployment answers the other half of the problem: getting the output of that signal into a router that already exists, with the monitoring and failover a network operations center can actually run day to day.

Chapter 14 frames quantum security as a race between two approaches. Post quantum cryptography, or PQC, keeps using math for encryption, just math that a quantum computer cannot easily break, and the National Institute of Standards and Technology, or NIST, has already published standards for it. QKD, the approach in this demonstration, does not rely on hard math at all. It relies on physics: any attempt to intercept the entangled photons changes them in a way both ends can detect. That is also its limit. A QKD key only protects the specific fiber link between two endpoints, while PQC can protect data anywhere the software runs. That is why Cisco is running both approaches rather than picking one. AT&T's coming quantum resilient Software-Defined Wide Area Network, or SD-WAN, service runs PQC on that same 8000 Series router line, which puts both approaches on the same hardware within the same product family.

The public announcement described the deployment as moving QKD out of isolated research setups and into standard enterprise infrastructure. Coverage of the announcement also framed the three way pairing as proof that quantum networking gear from separate vendors can interoperate in a live deployment, which matters more for enterprise adoption than any single performance number. A separate technical paper from the zerothird team tested the same entanglement based approach over a 22 kilometer fiber link between two data centers, which gives the enterprise demonstration a research paper trail worth reading alongside the press coverage.

What This Changes in the Book

Chapter 1 currently ends at the physics of getting quantum signals onto standard fiber. This deployment extends that story into the network operations layer: orchestration, telemetry, and automated remediation running on hardware already shipping. Chapter 14's framing of PQC and QKD as separate paths still holds, but the AT&T and Cisco pairing on the same 8000 Series router line is worth adding as a concrete case where one operator runs both approaches at once instead of choosing sides.

This post will fold into the next edition of the book, due September 1. The current edition is at gordostuff.com/p/quantum-from-ground-up-hardware.html.

POTS to IP took a decade of this kind of work: new transport riding on racked equipment. Quantum key distribution is passing the same tesst. The obstacle was never the physics. Here it's whether the keys can ride on a router Cisco already sells, watched by software a network operations center knows how to run.

Where the Jobs Actually Are

Some graduates spent the 2026 commencement season blaming AI for a job market that shut them out, loud enough that tech executives got booed at graduation ceremonies over it. Recruiters tell a different story.

Matt Walsh, CEO of the Phoenix search firm Blue Signal, works semiconductor hiring daily and says the problem isn't automation. "There aren't enough people," he says. The United States is heading toward what labor economists call the largest workforce shortage in its history, and it shows up hardest in the fields that build things.

The semiconductor industry expects to add close to 115,000 jobs by 2030. The Semiconductor Industry Association projects a shortfall of 67,000 technicians and engineers to fill them. That gap sits squarely in associate degree and bachelor's degree technical programs, not in the AI research labs getting most of the headlines.

Construction and the skilled trades show the same pattern. Branka Minic, CEO of the Building Talent Foundation, says fewer than half the workers needed in construction are entering the field, even with starting wages hitting $50 an hour in some markets. College graduates aren't matching that pay in comparable years of training.

Cybersecurity tells a similar story. CyberSeek, the workforce tracker built by CompTIA and NIST, counts hundreds of thousands of open cybersecurity positions in the U.S. against a supply of qualified workers that consistently falls short. The roles span network defense, security operations, and incident response, and they don't require a computer science PhD. A two-year degree with the right certifications gets a candidate into the field.

This is good news if you're building a technical career instead of chasing a headline. Employers in semiconductors, cybersecurity, advanced manufacturing, robotics, and skilled trades are competing for candidates, not filtering through thousands of applicants for one opening. Two-year technical programs, apprenticeships, internships and engineering degrees put graduates directly into that competition.

States have noticed too: several are merging workforce and higher education agencies or offering loan payoff incentives to pull people into these pipelines.

The AI panic makes for a cleaner headline than a demographic and skills pipeline problem. But the demand for people who can build, install, test, and maintain physical systems is not shrinking. It's the part of the labor market with the fewest applicants and the most openings.

ETH Zurich Builds Quantum RAM Out of Vibrations

Chapter 5 of Quantum From The Ground Up covers superconducting qubits by way of Josephson junction fabrication and IBM's 1,121-qubit Condor chip. It never had to answer a basic architecture question: where a superconducting qubit puts its data when it isn't actively working on it. Researchers at ETH Zurich just built an answer, and it doesn't look anything like a normal memory chip.

A team led by physicist Yiwen Chu, head of ETH Zurich's Hybrid Quantum Systems Group, split a quantum computer into the same two roles a laptop uses: a processor and a separate working memory. The design uses a superconducting transmon qubit as the processor and a mechanical resonator as memory, on a chip package 7.5 millimeters long. Instead of storing a qubit's state electromagnetically, the resonator holds it as a mechanical vibration, the way a guitar string holds a note, except this vibration follows quantum rules rather than classical ones.

Each resonator supports several distinct vibrational modes, and each mode works as its own memory slot. To run a computation, the qubit reaches into the resonator, pulls out a stored vibration, modifies it, and writes it back. Doctoral students Yu Yang and Igor Kladarić built the hybrid chip alongside Chu. The team validated the architecture by running a Quantum Fourier Transform and a period-finding algorithm on the hybrid chip, published in Science. That marks the first demonstration of mechanical resonators executing real quantum algorithms rather than just holding a state.

Superconducting qubits pack in tightly, but that density crowds out room for data. Electromagnetic memory schemes have historically traded a smaller footprint against coherence time. Mechanical resonators split that trade differently, offering higher storage density and longer coherence in less physical space. the approach still has to prove it scales beyond a single test chip, and Chu's group is continuing the work with that goal in mind.

What This Changes in the Book

Chapter 5's numbers don't move. IBM's Condor still holds at 1,121 qubits and 99.0 to 99.5 percent two-qubit fidelity, and nothing here challenges either figure. What changes is the chapter's scope. Chu's result adds a memory subsystem to the superconducting platform, a second engineering problem the chapter didn't previously address. It's a proof of principle, not a shipped component. A chip built for one qubit and one resonator still has to prove itself when both categories multiply.

This post will fold into the next edition of Quantum from the Ground Up, due September 1. The current edition is free to read at gordostuff.com/p/quantum-from-ground-up-hardware.html, and if it's useful to you, a coffee at ko-fi.com/gordostuff keeps it updated.

Dad and Doug

Lost my father and a close friend, Doug, within months of each other last year. Dad was fine right up until the last few weeks. Doug had been fighting for about five years. I wrote about each of them here already (for Dad, and for Doug), but never posted the obituaries themselves.

This blog has been mostly tech since 2005. But some things need to get written down while I still can, and this is one of them. Miss them both.

Write things down while people are still around to read them. The obituary is the last thing anyone writes about most of us. Everything before it is still yours to say.

The Free Path Into Quantum Work

The entire IBM Quantum Learning catalog is now open to the public, no partner account or IBM Quantum Network membership required. The catalog runs past ten courses, from the basics of qubits and circuits through algorithms for factoring and search, up to a course built around running experiments on processors with 100 or more qubits, all hosted free on the IBM Quantum Platform.

Path from free IBM Quantum Learning courses to certification to a quantum workforce role

The IBM Certified Quantum Computation using Qiskit v2.X Developer - Associate is a single exam built on the same Qiskit SDK and Qiskit Runtime the free courses teach. Since the program launched in 2021, more than 1,300 people across 71 countries have passed it. The exam runs up to $200 depending on location, with an optional $30 practice test through Pearson VUE. The preparation runs $0.

This matters most for programs without a quantum lab or a research grant behind them, which describes most community colleges and a fair number of universities. Qiskit needs Python and a working knowledge of linear algebra. Nothing else.

Every one of my students has a laptop and an internet connection. That's all you need to get started today.

This post rolls into the next edition of Quantum from the Ground Up, due September 1.

The Other Half of the Emerging Quantum Workforce

Chapter 2 of Quantum from the Ground Up covers the Massachusetts side of quantum
workforce building: over fifty million dollars in state investment, a quantum supply chain accelerator complex in Springfield, Massachusetts and a three tier pyramid running from technicians to PhDs. That chapter answers one question. How do you train people for jobs that barely exist yet? It does not answer a second question. Once someone is trained, where do they get time on a real quantum computer?

The federal government's answer is a program most people outside national labs have never heard of. The Quantum Computing User Program, or QCUP, runs out of the Oak Ridge Leadership Computing Facility. It owns none of the hardware it gives access to. Instead it brokers competitive, merit reviewed time on commercial quantum processors from vendors including IBM, Quantinuum, and Rigetti, for researchers doing open, fundamental science.

The mechanism is simple. A researcher submits a project proposal explaining what they want to run and why it needs quantum hardware. The Quantum Resource Utilization Council and independent referees review it for merit. Once approved, every user on that project applies for an account and gets assigned a Scientific Liaison, someone who understands both the science domain and the hardware, to help them get past the parts of quantum programming that have nothing to do with their actual research question.

The growth numbers are accelerating. A 2024 survey of the program found it had grown from 52 projects and 117 users in 2020 to 80 projects and 271 users by the end of 2023. Users range from national lab veterans to graduate students running their first circuit. Most projects are proof of principle work, not production science, which is exactly what you would expect from a field still figuring out what its hardware is good for.

Where the Access Turns Into Results

The hadronization work covered in my previous post, Anthony Ciavarella's simulation of quark binding on an IBM Heron processor, ran entirely through QCUP access. That is not a coincidence. It is the point of the program. Give enough researchers enough time on real hardware and some of them will publish in Physical Review D.

Demand is outrunning the obvious ways to measure it. The 2026 Quantum Computing User Forum at Oak Ridge, running July 20 through 24, closed registration before the event even started. A Fall 2026 hackathon open now for proposals gives existing users priority and asks new applicants to already hold an allocation before an August 7 deadline. This is not a program short on interested researchers.

What This Changes in the Book

Chapter 2 currently frames quantum workforce building as a state level, training focused story: Massachusetts building capacity so people are ready when jobs appear. QCUP is the federal, research access side of the same pipeline, and the book does not mention it yet. The two are not competing models. A pipeline that only trains people without giving anyone hands on time on real hardware produces graduates who have never run a circuit outside a simulator. QCUP is where some of that hands on time actually happens, for the research end of the workforce rather than the classroom end.

The next edition will add QCUP to Chapter 2 as the research access counterpart to the state workforce investment already covered there, with the growth numbers above and a note that at least one book worthy physics result, the Ciavarella hadronization paper, came directly out of that access.

This post will be folded into the next edition of Quantum from the Ground Up, targeted for September 1. The current edition is available at gordostuff.com/p/quantum-from-ground-up-hardware.html

A Quantum Computer Watches Matter Form From Scratch

Chapter 11 of Quantum from the Ground Up covers an IBM quantum chip that helped simulate a large protein molecule in May 2026. A new result uses a chip from the same family for something further from everyday life: watching the pieces of an atom's nucleus form.

Here’s some background - protons and neutrons, the particles that make up the nucleus of an atom, are not the smallest things out there. Each one is built from smaller particles called quarks, held together by a force called the strong force. Physicists call any particle built from quarks a hadron. The process of quarks locking together into a hadron is called hadronization. It happens constantly inside stars, inside particle colliders, and, billions of years ago, in the early universe. Nobody has ever watched it happen step by step, because it unfolds too fast and too small to see, even with instruments as powerful as CERN's Large Hadron Collider.

Anthony Ciavarella, a scientist at Lawrence Berkeley National Laboratory, used 104 qubits on an IBM quantum processor to simulate a piece of this process instead of observing it directly. He reached the machine remotely through a Department of Energy cloud access program called QCUP, rather than owning the hardware himself. We’ve described qubits in past posts - the basic unit of information in a quantum computer, similar to a bit in a regular computer, but able to hold more complex states. More qubits generally means a computer can model a bigger or more detailed problem.

The number 104 is an interesting choice and not a hardware limit. IBM's chip has 156 qubits, and Ciavarella used only some of them. In his setup, each qubit stands in for one point on a one-dimensional line of quark positions, so the qubit count is really a choice about how long a line to simulate. He picked a line of 104 points, long enough to fit a stretched gluon string spanning most of it, so something called the snap could happen out in the middle where it is easy to see clearly, rather than run into the edge of the simulation before it finished stretching.

The specific piece he simulated is called string breaking. Quarks are linked by something physicists describe as a string made of particles called gluons. Pull two quarks apart and that string stretches, the way a rubber band stretches, until it holds so much energy that it snaps. When it snaps, the energy does not just disappear. It turns into a brand new pair of quarks, which is how new hadrons get built. Ciavarella simplified the problem by using heavier quarks, which move around less and are easier to track, and he set up his simulation using a method he helped develop for preparing a quantum computer's starting state cleanly.

The simulation's answer matched what earlier work on ordinary supercomputers had already found. That match is the actual achievement here. Nobody is claiming the quantum computer beat a classical one. The result, published in Physical Review D, shows that a real quantum computer can reproduce a calculation even though the model was simplified down to one dimension. The data also hinted that the gluon string might briefly act like a heated gas right before it snaps, a detail worth checking again once the simulations get less simplified.

Why This Matters

Quantum computers are noisy. Every operation carries a small error, and those errors add up as a circuit grows. Nobody trusts a quantum result on a problem nobody can already solve until the machine first proves it gets the right answer on a problem people already know. Matching the classical calculation is that proof. It confirms the way Ciavarella mapped quark positions onto qubits, prepared the starting state, and corrected for hardware noise all worked correctly on real hardware, not only on paper.

Ciavarella's model used heavy quarks in one dimension because that version is still solvable classically. Light quarks, three dimensions, and watching a collision unfold in real time rather than checking a single snapshot are where classical computers run out of room, since those calculations carry exactly the kind of entanglement a regular computer cannot represent efficiently. That regime is also the one closest to what actually happens inside a real collision at a facility like the Large Hadron Collider. As Ciavarella put it, physicists already have the theory for hadronization; they have lacked a way to calculate predictions from it. This result is the first step toward a machine that can.

This was not the first attempt at something like this. In 2024, some of the same researchers, working with the University of Washington and Berkeley Lab, used 112 qubits to build and track a hadron directly, watching it move over time in an earlier, related model on an earlier version of this same IBM chip. The 2026 result trades that moving picture for a sharper look at the single moment when a hadron forms. 

What This Changes in the Book

Chapter 5, which covers this family of IBM chips, gets a new data point: the same chip line now has a published result in particle physics, using 104 qubits, on top of the protein simulation already covered in Chapter 11. Chapter 11's claim that this hardware has uses beyond biology gets a second example, and a more direct one, since this simulation ran entirely on the quantum processor rather than splitting the work between a quantum computer and a classical one.

This post will fold into the next edition of Quantum from the Ground Up, due September 1. The current edition is free to download at gordostuff.com/p/quantum-from-ground-up-hardware.html.