Willow, Majorana, and Ocelot Google, Microsoft, and Amazon's Quantum Chips: Capabilities and Key Differences

Related Vendor

Wuxi InfiMotion Technology Co., Ltd.

Google, Microsoft, and Amazon have each unveiled major advancements in quantum computing chip technology within weeks of each other. But what sets these quantum chips apart, and how do their innovations compare? This article explores the key differences, breakthroughs, and future potential of quantum computing from the world’s leading cloud providers.

The development of quantum computers is advancing rapidly. While classical computers are limited to binary states, quantum processors open up entirely new computational possibilities through superposition and entanglement. However, one of the greatest challenges remains error tolerance—qubits are extremely sensitive and tend to lose information through decoherence.

However, if you believe the statements from Google, Microsoft, or Amazon, each of these companies has made new progress in overcoming these problems. All three providers rely on fundamentally different strategies to realize reliable and scalable quantum computers.

Google Willow: Error-corrected qubits with extremely long stability

In 2019, Google introduced its own quantum processor for the first time and immediately proclaimed "quantum supremacy" based on the state-of-the-art technology at that time. However, in the following five years, the company's efforts in the field had largely gone quiet.

With the quantum computer chip "Willow," the company returned in December 2024 and announced the achievement of a significant milestone: A logical qubit, consisting of 105 hardware qubits, was able to be kept stable for one hour. This remarkable stability is a crucial advancement for scaling quantum computers and for the ability to perform complex calculations error-free.

Google's technology relies on the so-called "surface code" for error correction. Numerous physical qubits are combined into a single logical qubit, allowing errors to not only be detected but also efficiently corrected. The critical factor here is that error correction becomes exponentially more effective the more qubits are integrated into the logical system.

By gradually increasing the "distance," or code distance, from 3 to 5 and then to 7, Google demonstrated that the error rate was halved each time. This scaling shows that the practical implementation of large, error-corrected quantum computers is becoming increasingly realistic.

Another key element is Google's new manufacturing facility, which allows for detailed control over the production process. Previously, quantum processors were manufactured in universal cleanroom labs. But now, according to the company, customized qubits with optimized structures can be produced.

In the process, the company has deliberately developed larger physical qubit components that are less susceptible to environmental disturbances. The combination of structural improvements and advanced error correction allows for an unprecedented period of stability.

This progress indicates that complex calculations, taking hours, may soon be realizable by quantum computers. The theoretical prediction that a significant increase in the number of qubits drastically improves error correction has now been experimentally confirmed. This takes Google a big step closer to realizing a usable quantum computer. How close they are to practical usability, however, remains a different matter. The quantum processor "Eagle," introduced by IBM in 2021, was already able to demonstrate 127 qubits.

Microsoft Majorana: Topological qubits with Majorana quasiparticles

Microsoft bases its quantum computer research on an entirely new physical foundation: Majorana quasiparticles. These exotic particles possess special topological properties that make them nearly immune to environmental disturbances. For a long time, the existence of these quasiparticles was considered purely theoretical, but Microsoft has now provided experimental confirmation of their existence.

Majorana quasiparticles arise in special nanostructures made of superconductors and semiconductors. They are localized at the ends of nanowires and behave like massless fermions. As a result, the resulting qubits are particularly resistant to the so-called decoherence and the storage of quantum information is especially stable.

Microsoft's first processor based on this, the "Majorana-1," introduced in mid-February 2025, features eight qubits. This is a low number compared to competing chips, but due to the inherent error-resistance, significantly fewer qubits may be needed to carry out error-corrected computations.

According to Microsoft, a central advantage of this architecture is scalability: theoretically, millions of these qubits could be accommodated on a single chip. This could allow Microsoft to develop high-performance quantum computers in the long term that require significantly fewer correction mechanisms.

Whether this is indeed the case remains an open question. "The paper is very good," said Klaus Ensslin, physics professor and semiconductor expert at ETH Zurich, in response to the Majorana announcement to Die Zeit. "But Majorana states were not demonstrated in it." He also believes that Microsoft has yet to prove that the chip is suitable as a qubit processor. Professor Ensslin is involved as a research group leader in the development of quantum processors at Zurich, which IBM is working on in collaboration with ETH Zurich.

Amazon Ocelot: Error resistance through cat qubits and transmon links

Just a few weeks after Microsoft's introduction of the Majorana-1, Amazon announced the release of its own chip named Ocelot, developed in collaboration with the California Institute of Technology (CalTech). In parallel, a paper on the topic of error correction in qubits was published in the journal Nature. Compared to the approaches of Google and Microsoft, the approach presented in the study "Hardware-efficient quantum error correction via concatenated bosonic qubits" pursues a hybrid strategy: instead of developing an entirely new qubit technology, the company relies on a clever combination of two different types of qubits: "Cat Qubits" and "Transmon Qubits."

The units Amazon refers to as "cat qubits" are a special form of superconducting qubits, where a single quantum state is distributed over multiple photons. This architecture makes them particularly resistant to bit-flip errors. However, as the number of photons increases, the risk of phase-flip errors rises.

To solve this problem, Amazon combines these Cat Qubits with classical Transmon Qubits. The Transmons act as error detectors, capturing phase-flip errors that can then be corrected. This mix allows for efficient error correction with a small number of qubits. The setup, developed together with CalTech, consists of two small silicon microchips stacked on top of each other. The company claims that the chip's design could reduce error correction costs by up to 90%.

The biggest disadvantage of this method is that the transmon qubits themselves are prone to errors. Additionally, the residual risk of rare bit-flip errors in the cat qubits remains, which could affect scalability in the long term.

Comparison of chip technologies

In summary, the three different approaches are characterized by the following features:

Google (Willow):

• Classic superconducting qubits with strong error correction

• Exponential error suppression through more hardware qubits

• Stability of a logical qubit up to one hour

Microsoft (Majorana-1):

• Use of topological qubits with high error resistance

• Potential scaling to millions of qubits per chip

• Simplified control through digital signals

Amazon (Ocelot):

• Hybrid approach for error correction

• High stability through the combination of robust qubits

• Challenges due to remaining error susceptibility

No question: Each of these presented technologies promises unique advantages for the field of quantum computing. However, which of them will become the leading technology in practical hardware has yet to be determined. According to experts, a quantum processor must have at least 1,000 qubits to be suitable for practical applications. Even Google's Willow, with 105 qubits, is still far from this mark. For comparison, IBM's quantum processor Osprey already reached 433 qubits at the end of 2022, leveraging quantum software for error correction and noise reduction.

These developments are exciting but also highlight another challenge: the architectural approaches among the companies are fundamentally different. The extent to which porting between architectures or scalability within the same is possible is currently unproven. Even if the magical threshold of 1,000 qubits may not be too far off, users are likely to face another issue when the first functional quantum computers arrive: a platform dispute among the various providers!

Quantum computing

First physical quantum computer in Switzerland inaugurated

Empa Young Scientist Fellowship

Young researcher makes quantum molecules glow