Scientists from more than two dozen universities, national laboratories, and quantum technology companies published a field-wide strategic roadmap last week that, for the first time, gives neutral-atom quantum computing a consensus answer to the question the industry has long dodged: exactly what would it take, in atoms and laser watts and error-correcting codes and years, to build a machine that can actually do something classically impossible and genuinely useful — and when? If the answer arrives on schedule, the world's most widely used encryption standard could be within reach of a quantum computer within a decade.

The paper, titled Strategic Plan for Neutral Atom Quantum Computation (arXiv:2607.21554), was submitted on July 23, 2026. Its authors span MIT, Harvard, Yale, Cornell, Stanford, the University of Chicago, the University of Wisconsin-Madison, ETH Zurich, UCLA, UCSB, Purdue, Northeastern, the University of Washington, the University of Waterloo's Institute for Quantum Computing, the Weizmann Institute of Science, Université Paris-Saclay, the Max Planck Institute of Quantum Optics, Ludwig Maximilian University of Munich, NIST, the Joint Center for Quantum Information and Computer Science, and commercial players including QuEra Computing, PASQAL, Infleqtion, planqc, and NanoQT. The paper grew out of a National Science Foundation town hall convened at MIT's Endicott House in January 2025.

The breadth of the coalition is itself a statement. This is not one company's product pitch. It is the field writing its own homework assignment — and attaching a realistic grade.

Why Neutral Atoms, and Why This Technology Competes at the Front of the Pack

Neutral-atom quantum computers work by trapping individual atoms — typically rubidium or cesium — in tightly focused laser beams called optical tweezers. The atoms are arranged into large, programmable arrays and can be physically moved during computation. When researchers need to perform logic operations, they excite atoms into high-energy "Rydberg" states, in which atoms interact strongly with one another over short distances, enabling the entanglement operations that give quantum hardware its computing power.

The platform has a structural advantage over fabricated solid-state devices like superconducting qubits: atoms of the same element are naturally identical. There is no manufacturing variation to engineer around. And unlike superconducting chips, neutral atoms can be physically repositioned mid-circuit — a capability that has proved vital for implementing the most efficient error-correcting codes, as IEEE Spectrum's 2026 neutral-atom overview documents.

Where neutral atoms currently lag is gate clock speed. A Rydberg two-qubit gate runs in hundreds of nanoseconds to a few microseconds; a superconducting CZ gate completes in 20 to 100 nanoseconds. That speed gap is real, but the movability advantage increasingly outweighs it as error-correction architectures demand long-range qubit connectivity that superconducting chips struggle to provide, as covered in TechTimes' report on the 2,000-tweezer platform.

Recent milestones have bolstered the technology's credibility. Leading experiments assembled arrays of thousands of atoms, while reported two-qubit gate fidelities surpassed 99.5%. In January 2026, QuEra, in collaboration with Harvard and MIT, published a landmark Nature paper demonstrating 96 error-corrected logical qubits using 448 physical atoms — a 4.7-to-1 physical-to-logical encoding ratio that represented the highest verified logical qubit count on any quantum hardware platform at the time of publication. The coalition's roadmap draws directly on this result as evidence that the path to scalable fault tolerance is credible.

A Technology Stack That Must Be Engineered as a System

The roadmap organizes its analysis into six interconnected domains: hardware scaling, integrated photonic control, quantum error correction, software and compilation, algorithms, and quantum networking. Its central argument is that progress in any one of these areas will deliver limited value without corresponding advances in the others.

Physical qubit scaling. The number of physical qubits in leading neutral-atom experiments grew at roughly 1.8 times per year over the past decade, while gate errors fell by a factor of approximately 0.6 annually — though the authors caution these are rough trends, not guarantees. Laser power is identified as the near-term constraint: current arrays of more than 3,000 rubidium atoms require approximately 15 watts of light near 850 nanometers. Commercial kilowatt-class laser systems, the paper argues, could support arrays of up to 100,000 atoms, though higher power introduces risks including heat, optical coating damage, and laser noise. Atom loss during computation — caused by imperfect gates, vacuum limits, or measurement errors — must also be managed through continuous reloading: a mechanism that replaces lost atoms without disrupting ongoing calculations.

Integrated photonic control. Perhaps the most significant engineering bottleneck at scale is the optical apparatus. Today's neutral-atom machines depend on spatial light modulators, acousto-optic deflectors, high-numerical-aperture lenses, and elaborate free-space optical setups. The paper says plainly that scaling this infrastructure to 100,000 qubits is impractical in its current form. Integrated photonics — chips that guide, switch, and modulate light through fabricated waveguides — could replace large free-space setups with compact devices containing thousands of individual control channels. PASQAL, one of the coalition's commercial contributors, is already working toward this: it targets gate operation rates of more than 100 cycles per second by 2028 through silicon nitride photonic integration — up from roughly one cycle per second today.

Quantum error correction: the code-choice that changes the encryption calculation. The roadmap highlights quantum low-density parity-check codes, known as qLDPC codes, as the most promising path to resource-efficient fault tolerance. Unlike surface codes — the dominant error-correction approach through most of the field's history — qLDPC codes can protect more logical information per physical qubit, because each error-check operation acts on only a small, fixed set of qubits, and each physical qubit participates in only a few checks. IBM's landmark 2024 Nature paper on qLDPC codes demonstrated this efficiency; the broader architecture of qLDPC codes versus surface codes is a major shift in how the field approaches fault tolerance.

That difference has a direct consequence for the cryptographic threat. Running Shor's algorithm to factor a 2,048-bit RSA number — the encryption standard protecting most internet banking, email, and digital certificates — could require as few as 10,000 to 100,000 physical neutral-atom qubits under optimistic qLDPC assumptions, according to estimates cited in the roadmap. Surface-code architectures previously required estimates in the millions of physical qubits. The movability of neutral atoms — their ability to be physically rearranged — is well-matched to qLDPC codes, which require long-range connectivity between qubits that superconducting chips cannot easily achieve.

Software and compilation. Compilers for neutral-atom systems face a problem with no good classical analogue: they must schedule not just logic gates, but the physical movement of atoms, accounting for each atom's location, its travel speed, which operations can proceed in parallel, and how to respond in real time to atom loss and error signals. The roadmap calls for closer co-design between hardware teams and software developers so that algorithms can be optimized to the specific strengths of neutral-atom processors, including their flexible connectivity, ability to perform parallel operations across the array, and qubit mobility.

Algorithms. The authors are notably candid here. Only a small number of known quantum algorithms offer a clear exponential advantage over classical methods, and most of those require enormous fault-tolerant systems that do not yet exist. Hardware progress may outpace the quantum software ecosystem's ability to identify useful problems. The roadmap calls for more algorithm development focused on neutral-atom-native operations and urges the community to adopt shared benchmark problems with long classical histories — molecular systems, reaction dynamics, material transport, and the two-dimensional Fermi-Hubbard model — so that advantage claims can be evaluated against state-of-the-art classical solvers rather than outdated baselines.

Quantum networking. Rather than placing every qubit in a single monolithic processor, the roadmap envisions modular architectures in which multiple neutral-atom processors are linked by quantum channels. Possible approaches include converting atomic quantum states into photons that travel between modules, or physically shuttling transportable atom arrays between processing zones within a shared vacuum system. Networking would ease constraints on the size and optical complexity of any individual machine, at the cost of introducing new sources of loss and error in the links.

How Do You Know When Quantum Advantage Is Real?

One of the roadmap's most consequential contributions is its proposed definition of what should actually count as meaningful quantum advantage — a term that has been applied with varying degrees of rigor across the field for nearly a decade.

The authors propose four criteria, all of which must be satisfied simultaneously: the computation must produce a correct result; it must perform a task beyond the reach of available classical hardware; it must have a scaling advantage over classical approaches; and it must address a problem that matters to people outside the group that built the machine. That last criterion is pointed. It disqualifies experiments designed primarily to be classically hard — a common strategy in published advantage demonstrations that has drawn sustained criticism for producing results with limited real-world value.

The current state of play, as of July 2026, is that the industry has clear evidence of quantum computational advantage on narrow benchmarks, contested evidence of quantum utility on small physics problems, and no confirmed demonstration of quantum advantage on a commercially relevant problem, as entangledfuture.com's quantum advantage analysis describes.

To enable more rigorous comparisons, the paper introduces a new unit called the "quop" — defined as an operation executable on one or two qubits within a single error-correction cycle. The measure is intended to allow fair comparisons between proposed applications while accounting for the hidden overhead of fault-tolerant operations. Under this framework, proving that a machine is genuinely quantum might require roughly 1,000 logical qubits and millions of quops. Simulating certain material systems could need hundreds of logical qubits and millions of quops. Factoring a standard RSA-2048 key could demand thousands of logical qubits and billions of quops.

The roadmap divides the path toward practical advantage into three operational stages. The first — weak, unverifiable quantum advantage — covers relatively small circuits and tasks such as random-circuit sampling that can strain classical machines but may also be difficult to verify. The second stage, early practical advantage, requires roughly one million to one billion quantum operations and encompasses certifiable random-number generation and selected quantum simulations. The third stage, broad practical advantage, demands one billion to one trillion quantum operations and includes potential applications in chemistry, materials science, nuclear physics, cryptography, and optimization.

The authors also propose organizing competitions modeled on the NIST post-quantum cryptography evaluation process, in which quantum and classical teams would attempt the same rigorously defined problems at specified accuracy levels. Such contests, they argue, would anchor advantage claims against strong contemporary classical methods rather than outdated baselines.

Verification, the authors acknowledge, remains difficult. A task that cannot be solved classically may also be hard to check classically. Factoring is an unusually clean case because multiplying the proposed factors confirms the answer directly. Results from quantum chemistry or materials simulations are harder to validate, and classical algorithms improve continuously — what appears beyond conventional hardware today may become routine after a better classical method is discovered, as the paper's verification section discusses.

What Is Your Encryption Waiting On?

The quop framework, when applied to RSA-2048, produces a number that matters to security architects far outside the quantum research community. Under the roadmap's own estimates, factoring a 2,048-bit RSA key requires thousands of logical qubits and billions of quops — hardware capabilities the coalition does not expect within a few years.

That timeline matters because the regulatory clock is already running in the other direction. The National Institute of Standards and Technology finalized its first post-quantum cryptography standards in August 2024 and has called for RSA-2048 to be deprecated from use after 2030 and formally disallowed after 2035, under NIST IR 8547. The National Security Agency's Commercial National Security Algorithm Suite 2.0 framework mandates that all new national security systems be quantum-safe by January 2027. The Quantum Computing Cybersecurity Preparedness Act requires federal agencies to inventory vulnerable systems and report migration progress annually.

If the roadmap's 10-year conditional projection to quantum utility holds, and if the quop thresholds for RSA-2048 are correct, the practical window for migrating cryptographic infrastructure is narrower than it might appear. The "harvest now, decrypt later" threat — in which adversaries capture encrypted traffic today with the intention of decrypting it once a sufficient quantum computer exists — means the relevant deadline is not when a cryptographically relevant quantum computer is built, but when the data being protected today needs to remain confidential, as The Quantum Insider's analysis of the current threat landscape explains.

Organizations with sensitive data that must remain confidential for a decade or more — governments, healthcare providers, financial institutions — may face exposure from a machine that this roadmap says could credibly exist within that window.

A Conditional Forecast

If progress continues at roughly the rates observed over the past decade, the paper projects that neutral-atom systems could reach quantum utility within 10 years. The projection is explicitly conditional: it depends on the field scaling processors into a range of roughly 100,000 to one million physical qubits while maintaining high-fidelity control over every one of them — a requirement that will test integrated photonics, laser engineering, error correction, and software in combination, as the roadmap's timeline section makes clear.

There are real constraints that could slow the timeline. Gate clock speed remains slower than superconducting alternatives, which matters for algorithms that must execute billions of operations within the coherence time of the machine. The photonic integration required to control 100,000 atoms in a single system has not been demonstrated; it is a goal for the next hardware generation. And the algorithm development ecosystem has not kept pace with hardware progress — the paper explicitly warns that hardware may outrun the field's ability to find useful problems.

The paper is also careful to note its own limitations. It is a strategic plan rather than a report of a single new experiment. Its timelines depend on continued improvement across areas that have not yet been demonstrated together in one machine. As a preprint, it had not undergone formal peer review as of its July 23 submission date. The quop thresholds for RSA-2048 specifically depend on optimistic assumptions about qLDPC code performance that have not yet been validated at scale.

What it represents, nonetheless, is something the quantum computing field has rarely attempted: a frank, technically detailed, multi-institution consensus on what the next decade of work actually requires. For a field accustomed to competitive roadmaps and carefully managed milestones, that kind of collective candor — 50 authors from academic and commercial institutions agreeing on a shared homework assignment — may prove to be its most valuable output.

Frequently Asked Questions

How many physical qubits does a neutral-atom quantum computer need before it can break RSA-2048 encryption?

Under the qLDPC error-correction architecture that the roadmap favors, breaking a 2,048-bit RSA key could require as few as 10,000 to 100,000 physical neutral-atom qubits — a dramatic reduction from the millions of physical qubits estimated under older surface-code designs, according to estimates cited in the roadmap. Reaching that qubit count with high enough fidelity to run the necessary billions of quops is the task the 10-year timeline is conditional on. Independent estimates from researchers outside the coalition put the likely timeline for a cryptographically relevant quantum computer at somewhere between 2030 and the early 2040s, with expert surveys assigning roughly 50% probability to the mid-2030s, as The Quantum Insider's post-quantum migration analysis documents. No such machine exists today; the largest demonstrated system with verified error correction used 96 logical qubits.

What is a "quop," and why does it matter more than a raw qubit count?

A quop — short for quantum operation — is a unit proposed in arXiv:2607.21554 for measuring the effective computational capacity of a fault-tolerant quantum computer. It counts one operation executable on one or two qubits within a single error-correction cycle. Qubit counts alone are misleading because they say nothing about how accurately the machine operates or how much overhead error correction consumes. A machine with 1,000 physical qubits and poor gate fidelity may be less capable than one with 100 highly accurate logical qubits. The quop framework lets researchers and evaluators compare proposed algorithms to hardware capacity on a common, meaningful scale — making it harder for vendors to claim quantum advantage based on qubit counts alone.

Should someone outside the quantum research community care about this roadmap right now?

Security architects and enterprise IT leaders should. The roadmap's 10-year timeline to practical advantage, combined with its explicit quop threshold for RSA-2048, means the window for migrating cryptographic infrastructure may be shorter than it appears. NIST's post-quantum cryptography standards were finalized in August 2024, and NIST is calling for RSA-2048 to be deprecated after 2030. The "harvest now, decrypt later" threat is active: adversaries are already archiving encrypted communications today with the intent of decrypting them once a sufficient quantum computer exists, as The Quantum Insider's migration timeline analysis documents. Organizations protecting data that must remain confidential for a decade or more — health records, financial transactions, classified communications — face meaningful exposure from a machine that this roadmap says could credibly exist within that window.

What distinguishes neutral-atom quantum computers from superconducting machines like those from IBM or Google?

The key architectural difference is qubit mobility. Superconducting qubits are fixed positions on a chip; they can only interact with their immediate physical neighbors. Neutral atoms sit in laser traps that can be repositioned mid-computation, enabling any qubit to interact with any other qubit in the array — a property called "any-to-any" connectivity, as IEEE Spectrum's 2026 neutral-atom overview describes. This makes neutral atoms uniquely well-suited to qLDPC error-correcting codes, which require long-range qubit interactions that superconducting chips cannot easily provide. The tradeoff is gate speed: superconducting gates complete in tens of nanoseconds; Rydberg gates take hundreds of nanoseconds to a few microseconds. Superconducting machines also benefit from a longer head start and deeper tooling ecosystem. The roadmap's central bet is that atom movability, combined with integrated photonics to replace today's bulky optical apparatus, will ultimately allow neutral-atom systems to scale more efficiently than any competing architecture.

The full preprint, "Strategic Plan for Neutral Atom Quantum Computation," is available at arXiv:2607.21554.