The Wrong Kind of Quantum

Let me say something that the quantum computing industry does not want to hear: the entire field, as it is currently being built, is heading in the wrong direction. Not wrong in the sense of scientifically dishonest — the physics is real, the mathematics is correct, the researchers are brilliant. Wrong in a deeper sense: the premise is wrong. The thing they are fighting is the very thing they need.

They are trying to harness quantum mechanics by hiding from the environment. And the environment always wins.

I want to make this argument carefully and at length, because it deserves more than a slogan. I want to walk through what quantum computing actually is (as I, personally, understand it to the best of my ability), what its current approaches actually do, why they are limited in a way that is not incremental but fundamental, and what a better path looks like. If this essay is longer than you expected, that is intentional. The world has been building the wrong thing for thirty years. The least I can do is explain why, completely.

The Promise — Stated Precisely

Richard Feynman proposed the idea in 1981. His observation was simple: simulating quantum systems on a classical computer is exponentially hard, because the state space of a quantum system grows as 2^N with the number of particles N, and classical computers must track all of it. A quantum computer, by contrast, could simulate a quantum system by being a quantum system — it could represent the exponentially large state space in the physical states of its own components. The simulation cost would be polynomial, not exponential.

David Deutsch formalized the quantum Turing machine in 1985. Peter Shor, in 1994, showed that a quantum computer could factor large integers in polynomial time — a result of immediate cryptographic significance, since virtually all public-key cryptography at the time rested on the assumption that factoring was hard. Lov Grover, in 1996, showed that a quantum computer could search an unsorted database of N items in roughly √N operations, a quadratic speedup over the classical O(N). These results are not conjectured. They are proven theorems. If you have a quantum computer, these algorithms run as advertised.

The operative word is “if.”

After Shor and Grover, the field pivoted toward implementation. The question became: how do you actually build a device that maintains quantum superposition and entanglement long enough to run a useful algorithm? And this is where the trouble begins — not because the implementation is difficult in a solvable engineering sense, but because the physics of the problem is fundamentally adversarial to the engineering goal.

What Decoherence Actually Is

A quantum bit — a qubit — is a physical system that can exist in a superposition of two states. The canonical example is the spin of an electron: up, down, or any quantum superposition of up and down. In a superposition, the qubit does not have a definite value. It is in both states simultaneously, with complex probability amplitudes whose magnitudes squared give the probabilities of finding the qubit in each state upon measurement.

The computational power of quantum computing comes from this superposition. A register of N qubits in superposition represents 2^N states simultaneously. A quantum algorithm operates on all 2^N states in parallel, and through carefully designed interference, amplifies the amplitude of the correct answer and suppresses the amplitudes of wrong answers. When you measure at the end, you get the right answer with high probability.

Decoherence is the process by which a quantum system loses its quantum character and becomes classical. It happens because every real quantum system is coupled to its environment — to surrounding atoms, phonons, photons, electromagnetic fields. This coupling causes the quantum state of the system to become entangled with the quantum state of the environment. Once that happens, looking at the system alone, you can no longer see the superposition — it has leaked out into the environment, spread over too many degrees of freedom to be recovered. The qubit has effectively become a classical bit.

The timescale on which this happens is called the coherence time, and it is brutally short in any physical system that operates at temperatures where humans live. At room temperature — 300 kelvin — the thermal energy kT is about 26 millielectronvolts. The energy gaps that define quantum states in most engineered systems are of comparable or smaller magnitude. Thermal fluctuations continuously pump energy into and out of the quantum system, scrambling the phase relationships that encode quantum information. The coherence time at 300 kelvin for a typical superconducting qubit configuration would be femtoseconds — 10⁻¹⁵ seconds. That is shorter than the time light takes to cross a hydrogen atom.

This is not a problem that better engineering will solve at room temperature with superconducting circuits. It is a consequence of the thermal energy scale at 300 kelvin being larger than the energy gaps of the system. The problem is dimensional.

The Industry Response: Make It Colder

The response of the superconducting qubit community is to make the problem go away by cooling. If you reduce the temperature to 15 millikelvin — 0.015 degrees above absolute zero — the thermal energy kT drops to about 1.3 microelectronvolts. The energy gaps of the superconducting qubit circuit, designed to be around 5–10 gigahertz (roughly 20–40 microelectronvolts), are now large compared to thermal energy. The thermal noise that would scramble the quantum state is suppressed.

This works, in a narrow sense. At 15 millikelvin, superconducting qubits can maintain coherence for microseconds. Google’s Sycamore processor, the one that achieved the 2019 “quantum supremacy” demonstration, ran at 20 millikelvin. IBM’s Heron processor, which represents the state of the art in 2024–2025, achieves coherence times on the order of hundreds of microseconds for its best qubits, with two-qubit gate fidelities around 99.5%.

Let me be specific about what 99.5% gate fidelity means in practice. A 0.5% error rate per gate sounds small. It is not small when you are running algorithms that require millions of gates. A circuit with 1,000 gates applied sequentially has, at 0.5% error per gate, a probability of completing without error of (0.995)¹⁰⁰⁰ ≈ 0.67% — fewer than one in a hundred computations complete clean. To do anything useful — to run Shor’s algorithm on a 2048-bit RSA key, which requires on the order of 10¹¹ to 10¹² operations — you need error correction.

Quantum error correction is one of the genuine intellectual achievements of the past thirty years. The surface code, the most practically promising approach, encodes one logical qubit in a two-dimensional array of physical qubits. By measuring stabilizer operators — without measuring the qubits themselves — the code can detect and correct errors without destroying the quantum information. The price is overhead: a fault-tolerant logical qubit with a surface code requires on the order of 1,000 to 10,000 physical qubits, depending on the error rate of the physical qubits and the depth of the computation.

To run Shor’s algorithm on a 2048-bit RSA key with a fault-tolerant quantum computer, current estimates require between one and four million physical qubits, depending on the physical error rate assumed. IBM’s roadmap targets 100,000 physical qubits by the late 2020s. The gap between 100,000 and 1,000,000 is a factor of ten, and the history of quantum computing roadmaps suggests that the gap in time will be considerably larger.

All of this at 15 millikelvin. Let me describe what that actually requires.

The Refrigerator Is the Computer

A dilution refrigerator — the device that cools a quantum processor to 15 millikelvin — is not a laboratory instrument in the usual sense. It is a major piece of infrastructure. A typical unit for a 50–100 qubit processor is roughly the size of a large wardrobe, weighs hundreds of kilograms, and requires several hours to cool down from room temperature. The cooling is achieved by circulating a mixture of helium-3 and helium-4 isotopes through a series of stages: a pulse tube cooler brings the temperature to around 4 kelvin; a Joule-Thomson stage brings it to around 1 kelvin; and the dilution unit brings it to the final 15 millikelvin.

The qubit chip itself sits in the mixing chamber, enclosed in multiple layers of electromagnetic shielding. Every wire that connects the chip to room-temperature control electronics is a potential channel for thermal noise, so the wiring must be attenuated and filtered at multiple temperature stages. A 1,000-qubit processor requires hundreds to thousands of individual coaxial cables threading through the cryostat, each one carefully designed not to carry heat into the cold stage. The tangled forest of wiring inside a large dilution refrigerator is one of the most striking images in modern technology — visually impressive and practically nightmarish to engineer.

This infrastructure is not incidental to the quantum computer. It is the quantum computer. The qubit chip is perhaps a centimetre square. The infrastructure that keeps it cold is the size of a room. To scale to a million qubits — the minimum for fault-tolerant cryptographic applications — you would need either a refrigerator of proportionally enormous size, or many refrigerators networked together (introducing new decoherence problems at the quantum communication links), or some fundamental architectural change that nobody has yet demonstrated.

The cooling is not a choice. It is a confession — a confession that the physics underlying the computation is inherently fragile.

The Other Approaches and Their Own Dead Ends

Superconducting qubits are not the only approach. The field is rich with alternatives, and each has its genuine advantages. The pattern across all of them is instructive.

Ion traps (IonQ, Quantinuum) use individual charged atoms suspended in electromagnetic traps. Coherence times of seconds to minutes are achievable, and gate fidelities of 99.9% or better. The physics is more favorable. The problem is scale: ions must be individually addressed by laser beams and interact via phonons of the trap, a channel that works beautifully for small numbers of ions but becomes progressively harder to control as the chain grows. Gate times are milliseconds, versus microseconds for superconducting qubits. The largest ion trap processors operate with dozens to low hundreds of qubits.

Photonic quantum computers (PsiQuantum, Xanadu) use photons as qubits. Photons are nearly immune to decoherence at room temperature, which is a genuine advantage. The problem is entanglement: photons do not naturally interact with each other. The most practical approach — linear optical quantum computing — achieves entanglement probabilistically, with any given entanglement operation succeeding well below 50% of the time. Scaling probabilistic entanglement to a large fault-tolerant computation requires enormous overhead and has not been demonstrated even at small scale.

Topological qubits (Microsoft) are conceptually the most elegant and practically the most elusive. Encoding quantum information in topological properties of non-Abelian anyons would protect it from local perturbations by construction. The difficulty is that non-Abelian anyons — specifically Majorana fermions at the ends of topological superconducting nanowires — are extraordinarily difficult to realize. Microsoft’s 2022 Nature paper claiming signatures of topological qubits was retracted in 2023; the data analysis was found to be flawed. As of 2025, a fault-tolerant topological qubit remains undemonstrated.

Neutral atoms (QuEra, Pasqal, Atom Computing) combine long coherence times with better scaling prospects than ion traps. QuEra’s 2023 Harvard collaboration demonstrated logical qubits with error rates below physical qubit error rates — a key proof that error correction can work in practice. But neutral atom processors still require ultrahigh vacuum and complex laser systems.

Silicon spin qubits (Intel, UNSW) encode qubits in electron spins in silicon quantum dots, promising compatibility with CMOS manufacturing. Coherence properties can be excellent, but addressing single electron spins requires tuning multiple electrostatic gates to nanometre precision, and scaling from a handful of qubits to thousands is a serious unsolved engineering challenge.

Every approach fights a different version of the same war: cold versus vacuum versus probabilistic shielding versus topological protection. None has won. None has demonstrated fault-tolerant computation on a useful problem.

What Nature Already Knows

Here is the fact the quantum computing industry prefers not to dwell on: nature has been doing quantum computing at room temperature for billions of years, and it is not fighting anything.

Photosynthesis achieves near-perfect energy transfer efficiency — approaching 100% — in the light-harvesting antenna complexes of plants, algae, and photosynthetic bacteria. The excitation energy must travel nanometres through a warm, wet, noisy biological environment at 300 kelvin. By any classical calculation, the energy should scatter and be lost. Fleming’s group at UC Berkeley showed in 2007, using two-dimensional electronic spectroscopy, that the energy transfer involves wavelike quantum coherence: the excitation exists in superposition of multiple pathways simultaneously, and quantum interference directs it preferentially to the reaction center. The efficiency is not despite the warm environment. The environmental fluctuations are themselves structured in a way that drives the coherence in the right direction — a phenomenon now called environment-assisted quantum transport.

This is quantum computation happening in spinach. At body temperature. Maintained not by isolation from the world but by integration with it.

The European robin navigates by quantum-entangled radical pairs in cryptochrome proteins in the retina, maintained at body temperature in a living eye. The spin entanglement exists for microseconds — long enough to produce a direction-sensitive signal, sustained not by any cooling or shielding but by the molecular design of the protein.

Enzyme catalysis involves quantum tunneling of protons through energy barriers at 37 degrees Celsius, demonstrated in aromatic amine dehydrogenase and alcohol dehydrogenase. The tunneling is not suppressed by protein thermal motion — it appears to be enhanced by it, with protein dynamics driving donor and acceptor into configurations that facilitate tunneling.

Quantum mechanics at room temperature is not exotic. It is everywhere. What is exotic is the premise that you can only access it by hiding from the world.

The Topological Argument

The fundamental error of current quantum computing is confusing quantum mechanics with quantum isolation. These are not the same thing.

A superconducting qubit is a fragile engineered state. It is defined entirely by its isolation: take away the cooling, let in electromagnetic noise, and the qubit is gone. The qubit is not a property of the matter. It is a property of the engineered isolation of the matter from everything else. And the engineering of isolation does not scale. Every new qubit adds new coupling paths, new control lines, new heat loads, new crosstalk. The isolation overhead tracks the qubit count. The architecture is structurally anti-scalable.

Contrast this with the quantum behavior of a metallic (8,8) armchair carbon nanotube. The electrical conductance of this tube is quantized — at room temperature, in contact with metal electrodes, surrounded by air. The conductance is 2G₀, where G₀ = 2e²/h ≈ 77.5 microsiemens is the conductance quantum, a fundamental constant of nature. This quantization is not approximate or statistical. It is exact, because it is a topological invariant of the electronic band structure.

The two conducting channels of the (8,8) tube are protected by the crystal symmetry of the armchair nanotube combined with time-reversal symmetry. You cannot scatter an electron out of a ballistic conductance channel by adding thermal noise, because such scattering would require a change in the topological quantum number that defines the channel — and that change is forbidden by the physics. The mean free path of an electron at room temperature is micrometres to tens of micrometres. In a short tube, transport is ballistic: the electron traverses without a single scattering event.

In my device — a (8,8) SWCNT inside a (13,13) MWCNT, photon-gated, AC terminal drive — the quantum transport is exactly this topologically protected ballistic conductance. When a 693 nm photon is absorbed by the MWCNT, it detunes the outer tube’s energy levels, weakening the inter-wall coupling and allowing the inner tube’s conductance to rise from 1.47 G₀ to 1.91 G₀. The AC terminal voltage drives ±74 microamps for the two polarities. SNR exceeds 2,000 — not 99% fidelity after error correction, but a deterministic output with signal-to-noise ratio of 54 decibels.

We have confirmed this transport not merely at 300 kelvin but across a wide range of temperatures. The device behavior is robust across conditions far removed from the single-temperature benchmarks that quantum computing demonstrations are held to. This temperature robustness is not incidental. It is a consequence of the topological protection: the physics does not know or care what the thermometer reads.

The quantum mechanics is real. The isolation is absent. The stability is intrinsic.

The Cold Machine and the Warm One

Consider what 15 millikelvin means on a cosmic scale. The temperature of the cosmic microwave background — the afterglow of the Big Bang, the coldest large-scale temperature in the universe — is 2.7 kelvin. The interior of a dilution refrigerator running a quantum processor is 180 times colder than the average temperature of the universe. This is the engineering answer to decoherence: make the smallest possible piece of matter the coldest large-scale object in the known universe, and hope that the quantum states last long enough to be useful.

The THATTE device is warm because it does not need to be cold. The quantum states it uses are stable at 300 kelvin because they are protected by topology, not by the absence of thermal noise. Room temperature is not a challenge to be overcome. It is the operating condition, confirmed by simulation across a temperature range far wider than 300 kelvin.

I am not claiming the THATTE device is a quantum computer in the IBM sense. It does not maintain qubit superpositions. It does not implement Shor’s algorithm. It is a ternary classical computer that uses quantum transport as its physical mechanism. But that is precisely the point: you can build a computing system whose physics is quantum mechanical without needing to maintain fragile superpositions — and the computing system you build that way will be deterministic, room-temperature, and stable in a way that no superconducting qubit ever will be.

If a quantum computer that works is ever built — one that scales, that operates reliably at ordinary temperatures, that does not require a building-sized refrigerator, that can be integrated into manufacturable electronics — it will be built on materials whose quantum behavior is intrinsic and topologically protected. Not on engineered fragile states that must be hidden from the world.

The right quantum computer will be warm. It will be made of carbon. It will use the photon as its clock and the AC transmission line as its signal bus. It will not fight decoherence because its quantum properties will not be coherences in the qubit sense. They will be something older and more robust: the quantized conductance of a ballistic channel, protected by symmetry, expressed at the speed of light.

The Quantum Advantage Question

Before the machines themselves, there is a prior question worth examining: what problem do quantum computers actually need to solve, and for whom?

The headline applications are well known. Cryptography: Shor’s algorithm breaks RSA and elliptic curve cryptography, the backbone of secure internet communication. Drug discovery: quantum simulation of molecular systems could model protein folding and drug-receptor interactions at a fidelity that classical force fields cannot achieve. Optimization and machine learning: proposed quantum speedups exist for certain combinatorial problems. Materials design: simulating correlated electron systems is intractable classically.

The cryptographic application is the most urgent and the most concrete. NIST finalized its first post-quantum cryptography standards in 2024; the migration of internet infrastructure to quantum-resistant algorithms is happening regardless of whether a cryptography-breaking quantum computer is ever built.

The non-cryptographic applications are more uncertain. The systems where quantum effects matter most — transition metal complexes, correlated electron materials, excited state chemistry — are also the hardest to simulate on near-term hardware. The active space required for a useful nitrogenase catalyst calculation exceeds what any near-term device can handle. Promised timelines have slipped consistently.

The 2019 Google “quantum supremacy” demonstration is instructive. Sycamore completed a random circuit sampling task in 200 seconds that Google estimated would take the world’s most powerful supercomputer 10,000 years. IBM immediately contested the estimate; subsequent classical algorithmic work brought the simulation time down further. More importantly: the problem was chosen because it was easy for a quantum computer and hard for a classical one. It has no practical application. It was a benchmark, designed to demonstrate quantum advantage on the benchmark while telling us nothing about quantum advantage on useful problems.

The honest assessment of quantum computing’s status in 2026: the theory is solid, the hardware is improving, and the gap between current hardware and hardware that could do something useful on an important problem is measured in orders of magnitude, not incremental progress.

The Scaling Wall

There is a number I find clarifying. It is the ratio between the physical qubits required for a fault-tolerant computation on a useful problem and the physical qubits that exist in the best processors today.

For Shor’s algorithm on RSA-2048, fault-tolerant quantum computing requires roughly 4 million physical superconducting qubits, based on optimistic error rate estimates and surface code overhead. IBM’s best processor in 2025 has around 1,000 usable qubits. The ratio is 4,000.

These ratios are not slowly declining. Physical qubit count grows, but the required count for useful applications is also being revised upward as resource estimation becomes more careful. The scaling gap has not narrowed consistently.

Compare this to the THATTE device. A single SWCNT@MWCNT device is approximately 2 nanometres in inner diameter and can be as short as tens of nanometres — the geometric scale of current CMOS manufacturing. The fabrication challenges are real but are challenges in materials engineering and process development, not challenges that require cooling to 15 millikelvin. The scaling destination — a room-temperature, manufacturable, dense device — is not separated from the present by a factor of 4,000 in any physical quantity.

The Question Nobody Is Asking

Here is the question the quantum computing field should be asking and is not: what would a computing system look like if, instead of forcing quantum mechanics to behave classically — by cooling qubits until they stop interacting with each other, by wrapping them in electromagnetic shielding, by maintaining them in hard vacuum — we designed the system around quantum properties that are already stable?

The answer to that question is not a superconducting qubit. It is a carbon nanotube.

The answer has been sitting in the materials science literature since the 1990s, quietly, while three generations of quantum computing researchers have built increasingly sophisticated refrigerators. Saito, Dresselhaus, and Dresselhaus showed the ballistic transport properties of armchair SWCNTs in 1992. Tans, Devoret, and Dekker demonstrated ballistic transport experimentally in 1997. The physics of the armchair nanotube as a topologically protected quantum conductor at room temperature has been known, in its essentials, for nearly thirty years.

I do not say any of this with contempt. The superconducting qubit route has produced extraordinary science. Quantum error correction is an intellectual achievement of the first order. The people doing this work are among the finest physicists and engineers alive. They are working on the wrong problem with great skill.

A path can be scientifically rich and simultaneously be the wrong path. Ptolemaic astronomy was mathematically sophisticated — epicycles upon epicycles, a machinery of extraordinary precision, calibrated to centuries of celestial observation. It was also wrong. Not wrong in its observations or mathematics, but wrong in its premise. The right premise changed everything.

Current quantum computing is Ptolemaic. The correct premise is not “use fragile quantum states and protect them from the environment.” It is “use quantum properties that are intrinsically stable because the environment cannot disturb them.”

The right path is materials. The right materials are those whose quantum properties are features, not bugs — properties that survive contact with a warm world because they are protected by the structure of the world itself.