The core of a quantum computer is a silicon chip no bigger than a coin, sealed inside a gold-plated copper disk. Nearly everything else in the machine exists to keep that chip cold. When Fortune toured a quantum computer built by the California startup Rigetti in 2019, it found layers of gold plates stacked like a wedding cake, a mixing chamber where two forms of liquid helium sloshed together, and five nested casings that served as thermal shields. At the bottom the temperature drops to one-hundredth of a Kelvin, hundreds of times colder than outer space.
Bob Sutor, who headed quantum strategy at IBM, pointed at the 20-qubit computer his company showed that January and said: “What you’re looking at is the world’s most expensive refrigerator.”
The idea behind it goes back to Richard Feynman. The American physicist, who shared the 1965 Nobel Prize in Physics with Julian Schwinger and Shin’ichirō Tomonaga for fundamental work on quantum electrodynamics, conceived of an ultrapowerful calculator in 1982 that would run on microscopic particles. He called it a “quantum” computer, after the laws that govern nature at the subatomic level. Other accounts place the idea in the early 1980s.
What makes it different
An ordinary computer stores information as bits, each a zero or a one. A quantum computer uses qubits, which can hold both values at once through superposition. When two qubits interact, they become entangled, and the pair can no longer be described separately. In theory that gives the machine access to every possible result of a calculation in one step, where a classical computer works through them one after another.
Algorithms arrived in the 1990s. In 1994 Peter Shor of MIT presented one that could factor any number into primes, enough to break most encryption in use. A problem that would take a classical computer a few billion years would take a quantum one minutes. Three years later Lov Grover of Bell Labs showed that a quantum machine could search a database of ten thousand items in about a hundred steps, not ten thousand.
The hard part
Qubits are fragile. Magnetic fields, light and heat can destroy their quantum state, a loss called decoherence, which French research agency CNRS called the main obstacle to building these machines. Sébastien Tanzilli of the Institut de physique de Nice, who leads quantum technologies for CNRS, said qubits must keep their properties for the whole calculation.
Researchers tried atoms, ions, molecules, electrons, photons and superconducting circuits. In 2003 Rainer Blatt of the University of Innsbruck built the first two-bit logic gate from calcium ions. In 2012 a team at the University of Bristol used a photonic device to factor the number 21 into 3 times 7. Modest, but a working proof of Shor’s principle.
Superconducting circuits became the most popular approach, and IBM and Google both chose them. Patrice Bertet of CEA Saclay said they are simple to produce, so many can be put on one chip. Trapped ions hold their state longer but are slow to manipulate. A third candidate, electron spins in silicon, was still fragile in 2019, with only two entangled so far.
Error correction multiplies the problem. Tanzilli said it could take 1,000 or even 10,000 physical qubits to make one reliable logical qubit. A useful machine, in that case, needs millions. “The manipulation and control of such a large number of qubits remains largely beyond reach,” Bertet warned.
The race
By 2019 IBM, Google, Intel, Microsoft and China’s Alibaba were racing to build one. Intel’s director of quantum hardware said it “feels like my generation’s space race.” JPMorgan Chase and Daimler had started testing the technology. China was building a $10 billion national lab for quantum computing, while U.S. lawmakers set aside more than $1 billion for a national quantum strategy.
In 2019 Google disclosed that it had run a calculation on a quantum computer in minutes that would take a classical machine 10,000 years. Chinese scientists claimed an even faster result the next year. The record for entangled qubits stood at 20 at the time of the CNRS report, though Google had announced a 72-qubit processor in 2018 without showing how many of them were entangled.
Scientists expected a modest demonstration of quantum advantage within a couple of years and predicted up to 10 years before the machines could handle meaningful tasks.
Sources
- Calculating Quantum Computing's Future | Fortune
- Dassault Systèmes blog
- Computers: Promises of the Quantum Dawn
- Richard Feynman
