AWS unveils Ocelot to slash quantum error correction costs

The dirty little secret of quantum computing is that cramming more physical qubits onto a chip is the easy part. The real nightmare is keeping them from losing their minds the second a Wi-Fi signal or stray heat wave passes by.

To keep calculations from crashing, researchers have historically surrounded each fragile qubit with a massive entourage of extra qubits just to monitor and correct errors. It is like hiring nine supervisors to watch a single factory worker do their job. It works on paper, but the sheer cost makes scaling to a practical machine almost impossible.

AWS just unveiled a new prototype chip called Ocelot that takes an axe to that overhead. Developed at the AWS Center for Quantum Computing at Caltech and published in Nature, Ocelot slashes the resources needed for quantum error correction by up to 90%.

Putting Schrödinger’s cat on a silicon stack

Instead of taking an off-the-shelf qubit design and stitching error correction on top later, the AWS hardware team rebuilt the foundation around what are known as cat qubits.

Named after the famous Schrödinger's cat thought experiment, cat qubits have a clever physical property: they intrinsically suppress certain common quantum errors right out of the box. By tackling error suppression at the hardware level, the system needs far fewer backup parts to clean up mistakes.

Physically, Ocelot is built as a 3D sandwich of two integrated silicon chips, each measuring roughly 1 square centimeter, bonded together and electrically linked.

It features 14 core components: five cat qubits for computation, five buffer circuits to keep them stable, and four auxiliary qubits dedicated to spotting remaining errors. The cat qubits rely on ultra-stable oscillators crafted from thin films of tantalum, a superconducting material AWS engineers learned to pattern using conventional microelectronics manufacturing processes.

Cutting the error budget by 90%

According to Oskar Painter, AWS director of Quantum Hardware, scaling Ocelot to a full-fledged, fault-tolerant system could require as little as one-tenth of the resources demanded by standard approaches.

Painter estimates that future production chips built on this architecture could cost roughly one-fifth of current setups. More importantly, the team believes this design shortcut could accelerate the timeline to a truly practical quantum computer by up to five years.

Make no mistake: Ocelot is still an early prototype. AWS is treating this like its Graviton CPU journey, planning a multi-stage engineering flywheel that will take years of testing and refinement before it powers commercial workloads.

Why it matters

We think AWS is making the right bet by shifting the quantum race away from raw qubit vanity metrics and toward usable, fault-tolerant efficiency.

For years, the industry celebrated whenever a lab announced a hundred or a thousand physical qubits. But without robust error correction, those qubits produce mostly noise. By choosing cat qubits and designing error prevention into the physical components from day one, AWS is attacking the single biggest cost bottleneck holding back quantum supremacy.

For enterprise developers, this is not something you will deploy tomorrow morning. However, you can already build and test quantum algorithms on AWS via Amazon Braket using existing third-party systems and simulators. The real takeaway is that practical quantum applications, like drug discovery and materials science, might arrive before the decade is out rather than in some distant science-fiction future.

This move puts serious pressure on rivals like IBM and Google. If cat qubits prove that hardware-level error suppression can deliver reliable logical qubits at a fraction of the silicon real estate, brute-force error correction roadmaps could suddenly look hopelessly expensive.