Almost every serious quantum computing company now publishes a roadmap. These are glossy diagrams with named chips, dated milestones, and confident arrows pointing toward a future of fault-tolerant machines. They are part engineering plan, part marketing, and part recruiting pitch. Read carefully, they reveal how each company thinks the hard problems will be solved, and they expose where the optimism outruns the physics.
Why roadmaps matter in this field
Quantum computers are not improving on a smooth curve the way classical chips did under Moore's Law. Progress comes in jumps tied to specific breakthroughs: a better qubit, a working error-correction code, a way to wire thousands of control lines without melting the fridge. A roadmap is a company's bet on the order in which those jumps will happen. It tells customers when to expect useful machines, tells investors that there is a plan, and tells engineers what to build next. It also locks the company into public commitments that the rest of the industry will measure it against.
IBM: a parade of named processors
IBM popularized the modern quantum roadmap. For years it shipped a sequence of superconducting chips with bird names and rising qubit counts, moving from the 127-qubit Eagle to the 433-qubit Osprey to the 1,121-qubit Condor. Then the strategy shifted. Rather than chase ever-larger single chips, IBM pivoted toward modular designs with its Heron processors and a plan to link multiple chips together with classical and quantum interconnects. The headline target is a large-scale, error-corrected machine built around a code that is friendlier to fixed wiring than the standard surface code. The roadmap frames fault tolerance not as a distant dream but as a concrete engineering destination at the end of the decade.
Google: six steps to a million qubits
Google's published roadmap is structured as a series of milestones rather than a chip catalog. The first milestone, demonstrating a quantum processor that could outpace classical simulation on a contrived task, is behind it. The defining milestone since has been showing that adding more physical qubits to a logical qubit lowers the error rate instead of raising it, the crossover point that makes error correction worthwhile. The destination is explicit and enormous: a fault-tolerant computer with roughly a million physical qubits supporting around a thousand reliable logical ones. Google has been candid that the gap between today's hardware and that goal is years of work, and that each milestone unlocks the next.
Quantinuum: fidelity first
Quantinuum, built from Honeywell's quantum unit and Cambridge Quantum, takes a different posture. Its trapped-ion machines carry fewer qubits than the superconducting giants, but they emphasize quality. The company's roadmap leans on the high gate fidelities and all-to-all connectivity that ion traps allow, plus a clever architecture that physically shuttles ions around the chip. Its public timeline points toward a fully fault-tolerant machine around the end of the decade, reached by steadily raising qubit counts while keeping error rates low enough that fewer physical qubits are needed per logical one.
How to read them without getting fooled
The roadmaps share a common shape: today's noisy, intermediate-scale devices on the left, a valley of error-correction engineering in the middle, and useful fault-tolerant computers on the right. The differences are in the bets.
- IBM bets that modular superconducting chips plus a hardware-efficient code will scale fastest.
- Google bets that brute-force physical qubit counts, paired with a proven surface code, will get there.
- Quantinuum bets that starting from very clean qubits reduces the overall mountain to climb.
A few cautions apply to all of them. Dates near the present tend to be reliable because the chips already exist in a lab. Dates four or five years out depend on breakthroughs that have not happened yet, so treat them as aspirations rather than schedules. Qubit-count targets mean little without matching error rates, and almost no roadmap promises a specific commercially valuable application by a specific date, because nobody can honestly guarantee one.
The real value of the timelines
For all their marketing gloss, these documents serve a genuine purpose. They force companies to state, in public, what problem they are solving next and roughly when. That accountability is healthy in a field prone to hype. When a roadmap slips, the slip is visible. When a milestone lands, it is checkable. The smartest way to follow quantum computing is not to memorize the promised dates but to watch which arrows on each diagram actually get drawn solid.
