Hi there,
For most of the past years, IonQ felt like a quantum ponzi scheme to me.
Big roadmap numbers, slick marketing decks, acquisitions announced faster than physics gets done. Every quarter another headline; every quarter the underlying device looked roughly the same. The marketing was outpacing the physics, and the gap was widening.
But now in 2026, I have to walk some of that back. The engineering decisions behind the scenes have changed quite dramatically, and we are pivoting back to seeing great physics alongside the press releases, as well as an updated roadmap reflecting it.
The first thing worth noticing about IonQ’s updated roadmap is what it stopped saying. The old IonQ targets were denominated in #AQ, so called algorithmic qubits, a metric they invented and that quite frankly, nobody outside IonQ used.
The new roadmap drops it. Targets now run in physical and logical qubits: 64 in 2025, 256 in 2026, ten thousand in 2027, twenty thousand in 2028, climbing to two million physical and eighty thousand logical by 2030.
Credits: IonQ
#AQ was a way to compete on a custom benchmark while the rest of the field competed on qubit count and fidelity. Switching is IonQ playing on the same scoreboard as everyone else. You don't make that switch unless you think you can win on it. And you don't think you can win on it unless something in the engineering stack has materially changed.
What changed is Oxford Ionics.
The acquisition delivers three pieces of technology that, together, make the new roadmap technically coherent:
Electronic Qubit Control replaces free-space lasers with integrated antennas on the chip, aiming to sidestep photon scattering, optical addressing complexity, and the crosstalk that punishes any attempt to densify a laser-controlled trap.
The trap-on-a-chip 2D fabric breaks ions out of the 1D linear chain and into a dense planar layout, with up to 50x more qubits per trap; without this, the 2027 target of ten thousand physical qubits is not a roadmap, it's a wish.
The repeatable eight-qubit cell is supposed to deliver scale-invariance the way classical fabs do: copy-paste a working unit, don't redesign every generation.
Take any of these three away and the new roadmap stops working. Together, they are the technical reason the roadmap is worth reading at all.
Where I'm still skeptical: the cell tiling. The unit cell is eight qubits. The 2026 milestone is 256. The 2027 milestone is ten thousand. That's a roughly forty-fold jump in twelve months. Tiling looks easy on a slide, but how easy is it really on silicon ? Yield, variability across cells, defect handling, calibration of an array. Every fab generation in classical computing has hit walls between the prototype cell and the volume product, and quantum has tighter tolerances than any classical process.
I am not saying it cannot be done. I am saying that "we have an eight-qubit cell that works and we'll have a ten-thousand-qubit chip in twenty-four months" is the part of the IonQ story I still don't believe on faith.
Show me a 256-qubit chip running at the fidelities the eight-qubit cell hits and I'll update again.
Until next time,
