The Quantum Race Is On: Microsoft, Amazon, and Google Compete for What Comes Next — Research

Quantum computing — machines that are supposed to finish problems classical computers will never finish — is now a three-house race. In a short stretch of months, Microsoft showed Majorana 1, Amazon showed Ocelot, and Google showed Willow. Whether that is the start of a useful quantum era or a tightly sequenced press calendar is the question. Here is what actually shipped, and what it does not yet do.

What quantum computing is, and why it matters

Ordinary computers store bits as 0 or 1. Quantum machines use qubits, which, through superposition, can be 0 and 1 at once. Entanglement lets that state space grow fast enough that, on paper, some calculations become exponentially shorter. A 2023 McKinsey report noted that a quantum computer could simulate complex molecules for new drugs in minutes — work that would take a classical supercomputer thousands of years. The catch is physical: qubits are easily knocked off by heat, vibration, or electromagnetic noise. Errors pile up. Scaling is still the hard part.

Microsoft and Majorana 1: a new state of matter

In February 2025, at CES, Microsoft presented Majorana 1 and claimed a “topoconductor”: a material meant to induce topological superconductivity, a new state of matter. The chip uses Majorana zero modes — quasiparticles Ettore Majorana sketched almost a century ago — to form topological qubits that should be more stable and need less error correction. The device has 8 qubits today. Microsoft says the architecture could reach a million on a single chip, enough for practical jobs. Sergey Frolov at the University of Pittsburgh has challenged the evidence, arguing that the Nature paper does not conclusively show those particles. The scientific community has not closed the file.

Amazon and Ocelot: cheaper error correction

Days later, Amazon answered with Ocelot, a 9-qubit chip from AWS, unveiled on 27 February 2025. Ocelot uses cat qubits, which suppress bit-flip errors by design. Amazon says the architecture makes error correction up to 90% more efficient than standard approaches, which would cut the number of qubits needed for useful work. Also published in Nature, it is still a prototype. Rajeev Puri at UC Davis calls scaling it a monumental problem, not a remaining engineering tweak.

Google and Willow: error rates that fall as the chip grows

Google joined in December 2024 with Willow, a superconducting-qubit chip with automatic error correction. In a blog post, Google said Willow halves error rates with each additional qubit — the trend you need if you want larger machines to get more reliable, not less. The latest test used 105 qubits. That is closer to a demonstration of practical quantum computing than the field had, and still far from commercial workloads at scale.

Watch a short on Willow

Real progress, or a well-timed press cycle?

The three bets differ: Microsoft is chasing inherent stability, Amazon cheaper correction, Google precision that improves with size. Optimism still runs into the same wall. Qubits remain fragile. None of these chips has reached the thousands or millions of qubits needed to beat classical machines on real problems — what the field calls useful quantum advantage. Criticism that Microsoft overclaimed, plus the absence of near-term applications, keeps the other reading alive: this may be as much about investors and headlines as about physics.

What is actually at stake

If any of the three scales, the effects are not subtle: current encryption becomes a liability, and some scientific search becomes much faster. Microsoft talks about useful quantum computers “in years, not decades.” Amazon says Ocelot could pull that timeline forward by five years. Google, more cautious, points to 2033. The race is open. The winner, if there is one, would set the computing stack for a long stretch of this century.