# Will Correlated Noise Prevent Quantum Error Correction From Ever Working?

Mathematician Gil Kalai has spent two decades building a rigorous case that quantum computing cannot scale — not because of engineering limitations that better hardware will eventually overcome, but because of structural properties of noise that may be fundamentally incompatible with [fault-tolerant quantum computing](https://quantumintel.tech/glossary/fault-tolerant-quantum-computing). In an August 1, 2026 interview with Yuval Boger on The Quantum Insider, Kalai — a retired professor from the Hebrew University of Jerusalem now at Reichman University in Herzliya, and a former adjunct professor at Yale University — laid out two independent theoretical lines of attack against quantum scalability: correlated noise that defeats quantum error correction, and a complexity-theoretic argument that [NISQ](https://quantumintel.tech/glossary/nisq) devices cannot achieve [quantum advantage](https://quantumintel.tech/glossary/quantum-advantage) even under standard noise models. Both arguments, if experimentally validated, would close off the two main paths the industry is currently pursuing. Kalai entered the field in 2005, initially motivated by earlier work on noise sensitivity and noise stability conducted with collaborators Nati Linial, Jeff Kahn, Oded Schramm, and Itai Benjamini. His position has remained consistent and empirically falsifiable: current-generation hardware should already be sufficient to test his core conjectures.

---

## The Two-Pronged Argument Against Scalable Quantum Computing

Kalai's framework rests on two structurally distinct arguments developed over roughly two decades of work, and it is worth separating them clearly because they attack the problem from different angles and carry different implications for experimental falsification.

### Argument One: Correlated Noise Defeats QEC

The first line of argument, which Kalai describes as occupying roughly his first seven years of work starting in 2005, targets quantum error correction directly. The core conjecture is that realistic noise in quantum systems is not independent across qubits — it is correlated. Specifically, Kalai posits that **entangled pairs of qubits must have correlated errors**. This is not a claim about engineering imperfection; it is a claim about the physics of noise in quantum systems that, if true, would mean the foundational assumptions underlying surface codes and other QEC architectures are violated in practice.

Standard [error threshold](https://quantumintel.tech/glossary/error-threshold) theorems — the mathematical bedrock justifying the fault-tolerance roadmaps of [Google Quantum AI](https://quantumintel.tech/companies/google-quantum-ai), IBM Quantum, and others — assume that errors on individual qubits are sufficiently uncorrelated that redundancy can suppress them. Kalai's conjecture, by contrast, asserts that the act of creating entanglement necessarily introduces correlated error channels that cannot be suppressed by adding more physical qubits. The more entangled the system, the more systematically the errors co-vary — exactly the regime where QEC is supposed to kick in and help.

Kalai is careful to describe these conjectures as speculative rather than proven, but argues they are now experimentally testable. With current NISQ hardware generating meaningful entangled states across tens to hundreds of qubits, the correlation structure of errors in entangled pairs is a measurable quantity. He explicitly calls for this experimental work.

### Argument Two: Complexity Theory Bars NISQ Supremacy

The second argument is structurally independent and does not require any new or exotic noise model. Developed partly in collaboration with Guy Kindler — initially in the context of boson sampling — this line of reasoning uses standard computational complexity arguments applied to the standard noise model that the field itself uses. The conclusion, as Kalai states it, is that NISQ computers cannot achieve quantum supremacy under these standard assumptions.

The argument then extends: because quantum error correction is a harder task than quantum supremacy (it demands lower noise rates and more precise operations), if NISQ devices cannot demonstrate supremacy, they also cannot perform error correction of the quality required to bootstrap toward fault-tolerant computation. The implication is a kind of double lock: the noise rates that NISQ devices currently operate at are, by complexity-theoretic reasoning, insufficient for either the benchmark tasks or the QEC tasks that would allow the field to progress.

This second argument is particularly sharp because it sidesteps the debate about whether correlated noise is physically realistic — it accepts the field's own noise model and still arrives at a negative conclusion.

---

## What This Means for Google's 2019 Supremacy Claim

Kalai and Boger discuss [Google Quantum AI](https://quantumintel.tech/companies/google-quantum-ai)'s 2019 quantum supremacy result — in which Google reported that its Sycamore processor completed a specific sampling task in a time that classical supercomputers would require far longer to replicate — as a direct test case. Kalai's complexity-theoretic argument predicts that claimed demonstrations of quantum supremacy should be classically simulable or otherwise explicable without invoking genuine quantum computational advantage. The source transcript does not detail the specific technical objections Kalai raises to the 2019 result, but his framework would predict that the sampling distributions produced by noisy intermediate-scale devices are not, in fact, beyond classical reach.

This is not a fringe position in the abstract: subsequent classical simulation work by researchers at various institutions did partially undercut the original classical runtime estimates Google cited in 2019. Whether those classical improvements fully close the gap remains contested terrain — but Kalai's framework provides a principled reason to expect they should.

---

## The Scientific Value of Falsifiability

One of the more intellectually honest aspects of Kalai's position, as presented in the interview, is his explicit acknowledgment that his conjectures are falsifiable and should be tested. He does not argue that the industry should halt experimentation — he argues the opposite: that current hardware is now sophisticated enough to directly probe whether correlated errors in entangled qubit pairs exist at the levels his theory predicts.

This is important context for the broader industry. Skeptical theories that make specific, testable predictions are scientifically valuable regardless of whether they are ultimately correct. If experimentalists design clean tests of error correlation in entangled pairs and find the correlations Kalai predicts are absent or below the threshold his theory requires, that is a meaningful result that strengthens the case for fault-tolerant scalability. If the correlations are present, the field needs to know.

Kalai also raises a question that the field rarely asks publicly: what scientific insights would emerge if quantum computing ultimately fails? Understanding why a physical system cannot be controlled at scale is itself a scientifically rich outcome — one that would teach physicists something fundamental about [decoherence](https://quantumintel.tech/glossary/decoherence), entanglement, and the structure of noise in complex quantum systems.

---

## Industry Trajectory: Why Skeptical Voices Matter Now

The timing of this interview is not incidental. The quantum computing industry is at an inflection point where fault-tolerant roadmaps are being treated as engineering problems rather than open scientific questions. Major hardware players are committing to [logical qubit](https://quantumintel.tech/glossary/logical-qubit) milestones on multi-year timescales, and venture capital has followed accordingly.

Kalai's framework, if even partially correct, would have severe implications for those roadmaps. A noise correlation structure that scales adversely with entanglement depth would mean that the hardware improvements being pursued — higher physical qubit counts, lower individual gate error rates, longer [coherence time](https://quantumintel.tech/glossary/coherence-time)s — are necessary but not sufficient. The error budget would grow faster than the error correction capacity.

Scott Aaronson, one of the field's most rigorous theoretical voices, has noted that each experimental advance makes Kalai's position less plausible in his view. That is a legitimate scientific assessment. But "less plausible" is not "falsified," and the specific experimental tests Kalai calls for — direct measurement of error correlations in entangled pairs — have not been designed and executed with his conjectures as the explicit null hypothesis.

For enterprise buyers and investors evaluating quantum platforms: the honest answer is that Kalai's conjectures remain live scientific questions. The [error threshold](https://quantumintel.tech/glossary/error-threshold) theorems that underpin fault-tolerant roadmaps assume independent noise. Whether that assumption holds in practice at the scale required is an open empirical question — one that Kalai has been asking, with mathematical rigor, for over twenty years.

---

## Key Takeaways

- **Gil Kalai** (Hebrew University / Reichman University / former Yale adjunct) has maintained since 2005 that quantum computers cannot scale, citing two independent theoretical arguments.
- **Argument one**: Entangled qubit pairs must have correlated errors, which would defeat quantum error correction regardless of how many physical qubits are added.
- **Argument two**: A complexity-theoretic analysis using the field's own standard noise models predicts NISQ devices cannot achieve quantum supremacy — and therefore also cannot perform the QEC needed for fault-tolerant computation.
- **Google's 2019 supremacy result** is a direct test case for Kalai's framework; he predicts such results should be classically reproducible.
- **Kalai explicitly calls for experimental tests** of his core conjecture about correlated errors in entangled pairs, arguing current hardware is capable of running them.
- **The scientific community** — including Aaronson — acknowledges Kalai's arguments while generally viewing them as increasingly constrained by experimental progress, but the specific tests Kalai proposes have not been run against his conjectures as an explicit null hypothesis.
- **Investors and enterprise buyers** should note that fault-tolerant roadmaps assume independent noise; whether that assumption holds at scale remains an open empirical question.

---

## Frequently Asked Questions

**What is Gil Kalai's main argument against quantum computing?**
Kalai argues on two fronts: first, that realistic noise in quantum systems is correlated across entangled qubits in a way that defeats quantum error correction; second, that a complexity-theoretic argument using standard noise models shows NISQ devices cannot achieve quantum supremacy. Both arguments, if correct, would close off the two main paths toward useful quantum computation.

**Does Kalai's theory claim quantum mechanics itself is wrong?**
No. Kalai's theory does not dispute the validity of quantum mechanics. It disputes whether engineered quantum systems can be controlled at scale with sufficient noise suppression. The claim is about the behavior of noise in physical implementations, not about the underlying physics.

**How does correlated noise defeat quantum error correction?**
Standard fault-tolerance theorems — the mathematical basis for surface codes and similar QEC architectures — assume that errors on individual qubits are sufficiently uncorrelated. If entangled qubits experience systematically correlated errors, redundancy cannot suppress them in the way the theory predicts, and the error threshold cannot be reached.

**What experiment would falsify Kalai's conjecture?**
Kalai explicitly states his conjectures are experimentally testable with current hardware. A clean measurement of error correlations in entangled qubit pairs that found correlations absent or below the levels his theory requires would constrain or falsify his noise-correlation conjecture. This specific experiment, designed with his conjectures as the null hypothesis, does not appear to have been formally executed.

**How does Kalai's view relate to Google's 2019 quantum supremacy claim?**
Kalai's complexity-theoretic argument predicts that sampling tasks run on noisy intermediate-scale devices should be classically simulable. The 2019 Sycamore result is therefore a direct challenge to his framework — though subsequent classical simulation work has complicated the original runtime comparisons. The debate remains scientifically unresolved.