IBM Quantum

The roadmap everyone measures against

Ask anyone in the field to sketch where quantum computing is heading, and they will most likely reach for a version of IBM’s published schedule. The IBM quantum roadmap has become the reference document against which rivals, investors and government programmes set their own expectations, which is why the company’s tagline of being the roadmap everyone is measured against reads as plain description rather than salesmanship. IBM publishes dated hardware targets years in advance and then reports, in public, how far it has got.

The current hardware

The current generation is Nighthawk, a 120-qubit processor built around 218 tunable couplers. A qubit is the quantum equivalent of a bit, able to hold a blend of nought and one until it is measured; a coupler is the switchable link that lets two qubits interact so that a two-qubit gate, the basic operation of entanglement, can be applied. The more reliable those links, the longer the circuit a machine can run before noise swamps the result. Nighthawk is specified to support circuits of up to 5,000 two-qubit gates, a figure the IBM quantum roadmap expects to raise to 7,500 by the end of 2026 and 10,000 in 2027. Depth of that kind is the practical measure of what a processor can actually compute.

Alongside the workhorse chips sits Loon, an experimental processor built to demonstrate the components of fault tolerance rather than to add raw scale. Every physical qubit is noisy, so a fault-tolerant machine spreads one protected unit of information, a logical qubit, across many physical ones and uses error correction to catch mistakes as they happen. Loon is where those building blocks are being tried in silicon before they are committed to a production design.

The 2029 target

The design they lead to is Starling, the IBM quantum roadmap’s marquee target for 2029: 200 logical qubits able to run more than 100 million operations in a single job. That is the threshold at which a quantum computer stops being a laboratory instrument and starts being useful for problems classical machines cannot reach. Getting there depends on the error-correction work happening on Loon paying off at scale.

Nearer term, IBM has committed publicly to demonstrating verified quantum advantage by the end of 2026, meaning a computation that a quantum machine performs and that a classical computer can confirm was genuinely hard to reproduce. The word verified carries weight, because earlier advantage claims across the industry have often been walked back once classical methods caught up. To its credit, IBM runs an open community tracker so that any such claim can be contested in public rather than taken on trust, a stance that fits a company whose whole strategy rests on the credibility of its published dates. Readers new to the underlying ideas may find our quantum explainers a useful companion here.

Beyond the processors

The ambition is not confined to the processors themselves. IBM is spinning off what it describes as America’s first dedicated quantum chip foundry, backed by $2bn in combined federal and private funding, an attempt to secure domestic manufacturing for a technology that has so far depended on a handful of specialist fabrication lines. The move reads as a bet that whoever controls the supply chain will shape the next decade of the field.

The company is also exporting the hardware. It is commissioning one of India’s first physical quantum computers in Amaravati, due by September 2026, a Quantum System Two housing a 156-qubit Heron processor. Placing full systems in new national programmes turns the schedule from a corporate plan into shared infrastructure that others build upon.

What to watch

What to watch over the next eighteen months is simple enough to state and hard to achieve: whether the verified-advantage claim lands on time, whether Nighthawk’s gate depth climbs as promised, and whether Loon’s fault-tolerance components mature fast enough to keep Starling on track for 2029. Each is a checkable milestone on a public IBM quantum roadmap, and each will tell the rest of the sector whether the schedule everyone measures against is holding.