IBM Corp. said today it has taken a massive step forward toward its goal of delivering the world’s first fault-tolerant quantum computer by 2029, after successfully linking and cooling down the first pair of modules based on a new, highly scalable cryogenic architecture.

The new system makes it possible for hundreds of quantum processors to be linked together into a giant cluster, paving the way for large-scale quantum computers that can actually make a difference.

IBM said the achievement is a key milestone on its roadmap toward the delivery of IBM Quantum Starling, which is a new quantum computing system that’s currently under development and slated to launch in three years. It’s envisaged that Starling will be able to perform about 20,000 times more calculations than today’s existing quantum computers, making it powerful enough to deliver the Holy Grail of “quantum advantage.” That’s the threshold where these machines will be able to solve complex, real-world problems that existing classical supercomputers cannot do.

Quantum computers derive their incredible computing power from so-called “qubits,” which are the quantum version of traditional bits found in classical computers. With classical computers, the bits can represent either a 1 or a 0, whereas qubits can be a 1 or a 0, or both at the same time. They can do this because they take advantage of a principle in quantum mechanics known as “superposition,” and that enables them to process vast amounts of multidimensional data at such ungodly speeds.

The problem is that no one has been able to build a functional quantum computer so far, or at least, not at a scale that’s big enough to make a difference. That’s because these qubits are so incredibly fragile. They’re susceptible to basically any amount of “noise,” including things as subtle as a slight drop in temperature, electromagnetic interference, vibrations caused by someone walking across a room, or even the fluctuations of the Earth’s magnetic field. When they’re disturbed by this noise, qubits suddenly lose their quantum state – a process known as decoherence – and much of the data they store is lost, resulting in errors in quantum calculations.

IBM is trying to solve this challenge with “logical qubits,” so rather than use a single, error-prone qubit, engineers bunch together multiple physical qubits together with advanced error correction codes, which allows them to act as a single, reliable qubit. It’s a form of redundancy that ensures quantum computers can still process calculations correctly, even if some of the individual qubits fail.

But this solution creates new problems – to build a quantum computer that can achieve quantum advantage, IBM needs to have thousands of qubits all working in tandem. To protect these qubits, they must be housed in shared spaces and cooled to extreme temperatures close to absolute zero.