Google’s new quantum chip can do 10 septillion years of computing in 5 mins

Google’s Quantum AI team has announced an advanced quantum computing chip, named Willow, that can run random circuit sampling (RCS) benchmark — the most challenging benchmark available for quantum computers — in less than five minutes. Google believes that the same computation, if executed by one of the fastest supercomputers, will take more than 10 septillion years, which is a million times longer than the age of the Universe. 

Willow marks a major milestone in quantum computing for Google, which achieved quantum supremacy for the first time five years ago. In 2019, Google Sycamore chip-based quantum computer performed a mathematical calculation that would have taken classical computers 10,000 years to perform. 

With Willow, Google has also used a new method to reduce errors in quantum computers. 

“This cracks a key challenge in quantum error correction that the field has pursued for almost 30 years,” said Hartmut Neven, founder and lead at Google Quantum AI. 

Quantum computers work by managing quantum mechanical systems called qubits, which are highly sensitive and susceptible to calculation errors by decoherence or other forms of noise including a ray of light. As quantum computers grow, so does the error. According to Google, even the most advanced quantum computers will typically fail at least once in every thousand operations.

This issue is addressed with a method called quantum error correction, which protects quantum information by encoding and spreading it across multiple physical qubits to form logical qubits. 

In a paper published in the Nature journal, Google researchers wrote that Willow is the first chip where error-corrected qubits improve exponentially with size. Doubling the encoded qubits, from a 3x3 to a 5x5 or 7x7 lattice of physical qubits, halves the encoded error rate. Willow has 105 qubits. 

That said, larger lattices can also increase the error rate and “overwhelm” the error correction. However, with sufficiently low physical qubit error rates, error correction improves significantly as more qubits are added. The threshold error rate marks the point where error correction transitions from harmful to beneficial.

“Willow represents a significant leap forward in quantum hardware. It maintains the tunability of our previous architecture, Sycamore, while improving the average qubit lifetimes,” said Google researchers in a blog post. 

“With error correction, we can now in principle scale up our system to realize near-perfect quantum computing,” they added. 

The researchers said that they might need more than a thousand physical qubits per surface code grid to realize relatively modest encoded error rates of 10-6. 

A regular computer processes information with bits (0 and 1), which means that two bits can be in four possible states (00, 01, 10, or 11) but they can represent only one of these states at any given time. Whereas, a quantum computer allows two qubits to represent all four states at the same time. This is called superposition and is similar to four computers running simultaneously, which is what makes quantum computers much faster than regular computers.

With their superior computing capabilities, quantum computers have the potential to accelerate research in many fields, including chemistry, drug discovery, climate modeling, machine learning (ML) and cryptography. 

Addition of more qubits can lead to exponential increase in the computing capabilities of a quantum computer. IBM, which also owns a quantum computer, is planning to launch a modular quantum computer with a 4,000 qubit processor by 2025. It released the 433-qubit Osprey chip in 2022, followed by the more powerful 1,121-qubit Condor chip a year later.

Image credit: Google Quantum AI Lab