IBM has a clear plan for 100,000 qubits by 2033. What problems will a useful quantum computer solve and when will we get there?
❄️ A Refrigerator in New York
Somewhere at IBM's Thomas J. Watson Research Center, in upstate New York, a dilution refrigerator cools a metallic chip to 15 millikelvin — a temperature colder than outer space itself. Inside this chip lie superconducting transmon circuits, the qubits that IBM believes will change the world. The company isn't simply doing basic research. It has published a detailed roadmap stretching to 2033, targeting 100,000 qubits and the first truly useful quantum computer.
📈 From 5 Qubits to 1,121 — The Evolution in Numbers
In May 2016, IBM did something unprecedented: it opened the first quantum computer to the public via the cloud, the IBM Quantum Experience. It had just 5 qubits connected in a star-shaped pattern. Users could design quantum circuits through a graphical interface that IBM called a “quantum score”, like a musical sheet. It sounds primitive today, but at that moment it was the first time anyone could “touch” a quantum processor from their own computer.
Progress came at an impressive pace. IBM followed a series of processors named after birds:
- Eagle (2021): 127 qubits — the first to surpass 100
- Osprey (2022): 433 qubits — tripled in one year
- Condor (December 2023): 1,121 qubits — the world's largest quantum processor at that time
The number 1,121 was impressive. No competitor had that many qubits on a single chip. In practice however, qubit count doesn't automatically mean computational power. Qubits are extremely sensitive: noise from the environment, interactions between qubits, even cosmic rays can destroy calculations in milliseconds. IBM realized this quickly.
🔄 The Great Shift: Quality Over Quantity
Alongside the Condor, IBM unveiled Heron — a processor with just 133 qubits, but with dramatically improved gate fidelity. The move was strategic. Instead of simply chasing numbers, IBM bet on qubits that make fewer errors. Jay Gambetta, VP of IBM Quantum, explained that the future quantum computer won't be a massive monolithic chip, but rather a system of many smaller modules connected to each other.
This philosophy was realized in IBM Quantum System Two — a modular system that can house multiple processors within a unified computational framework. Think of it like a rack in a data center, but instead of classical servers, it contains dilution refrigerators with quantum chips that “talk” to each other.
🚀 The Road to 100,000 Qubits
IBM's detailed roadmap envisions the following until 2033:
- Flamingo: a processor designed for modular interconnection between chips
- Starling: a system with built-in quantum error correction (QEC)
- Blue Jay: the future processor that will reach 100,000+ qubits
Scaling to these levels requires solving quantum computing's greatest challenge: error correction. In a classical computer, if a bit changes value due to an error, you simply keep copies. In the quantum world, the no-cloning theorem forbids copying an unknown quantum state. The solution? Error-correcting codes like the surface code, which encode one logical qubit into dozens or hundreds of physical qubits.
In February 2023, Google showed that by increasing the number of qubits in a surface code, you can actually reduce errors — a critical milestone. In April 2024, Microsoft and Quantinuum achieved error rates in logical qubits 800 times better than the physical qubits. These results show that quantum error correction is no longer just theory — it's beginning to work in practice.
💻 Qiskit: The Language of the Quantum World
Hardware alone isn't enough without software. In March 2017, IBM released Qiskit — an open-source SDK (Software Development Kit) that allows researchers and programmers to write code for quantum computers. The code is compiled down to OpenQASM, an assembly language specifically designed for quantum circuits.
By 2025, over 400,000 users had used the IBM Quantum platform, publishing more than 2,800 research papers — from drug chemistry to financial portfolio optimization. As of June 2025, IBM offered 12 quantum devices for free on the cloud, with processors based on superconducting transmon qubits.
⚔️ The Competition: Google, Microsoft, IonQ
IBM isn't alone in the race. In September 2019, Google announced "quantum supremacy" when the Sycamore processor (53 qubits) executed a specific problem in 200 seconds — something a classical supercomputer would need (according to Google) 10,000 years. In 2025, Google's Willow processor achieved one of the first independently verifiable quantum advantages.
Microsoft follows a different path, betting on topological qubits based on Majorana particles. IonQ uses trapped ions instead of superconducting circuits, with the advantage of longer coherence times. Each technology has its strengths and weaknesses, and no one can guarantee which will ultimately prevail.
⏰ When Does It Become Reality?
Let's be clear: no quantum computer replaces a classical one today. Current quantum processors are noisy and can only execute “shallow” quantum circuits before errors accumulate irreversibly. This era is called NISQ (Noisy Intermediate-Scale Quantum).
IBM estimates that within the next decade, fault-tolerant quantum computers will be created. These will be able to simulate molecules for new drug discovery, break current cryptographic algorithms (via Shor's algorithm), and optimize supply chains that today require days of computation. IBM no longer speaks of “quantum supremacy” but of "quantum-centric supercomputing" — a hybrid model where classical and quantum computers work together.
The road is long, but the map exists. And the qubits, inside their refrigerators in New York, are slowly learning not to make mistakes.