I wrote "no lasers" about Oxford Ionics. That was too clean, and their own paper says so.
I wrote "no lasers" about Oxford Ionics. That was too clean, and their own paper says so.
Electronic Qubit Control is real and it is not marketing. An antenna is built into the silicon chip. An oscillating current through it produces oscillating magnetic fields, and those fields drive the gates. No beam touches a qubit during computation. That is how they hold the fidelity records on single-qubit gates, two-qubit gates and SPAM at once.
But read their PRX Quantum paper on all-electronic control, and look at where the lasers still are.
Six wavelengths: 375, 397, 423, 729, 854 and 866 nm. Ion loading, cooling, initialization, measurement. The paper says it plainly: "in all other zones, all ion control is laser-free."
In all other zones. Not the central one.
That sentence is the whole story. The optical table did not disappear, it got concentrated into a single zone. Every shot, the ions are transported into that zone to be cooled, moved out to the target zone to compute, and brought back to be read.
Today the ions travel to the light.
Everything IonQ has bought since is about reversing that. It is not three bets, it is one ladder:
→ EQC takes electronics as far as it goes. Gates, done.
→ Integrated photonics takes what electronics cannot do and puts it on the chip. Waveguides and grating couplers under every site, light delivered locally, ions that stop commuting.
→ Photonic interconnects take what one chip cannot do and link chips together.
Nexus Photonics is rung two. Closed June 30 for about $76M. Founded at UCSB in 2018 for exactly one problem: photonic integration at short wavelengths, 400 to 1600 nm. Standard silicon photonics starts around 1100 nm, because below that silicon absorbs its own light. Barium runs at 493, 614 and 650 nm. Calcium at 397, 423, 729, 854 and 866. Almost the whole trapped-ion spectrum sits in the band nobody could integrate. That band is the company. Being precise: 375 and 397 nm still fall below their stated floor, so the hardest lines are not solved.
Their chairman is John Bowers. It was his group at UCSB that solved laser coupling on silicon in 2005, by bonding the laser material on top and bending the light down into the waveguide. He has already sold Terabit Technology to Ciena, Aurrion to Juniper and Aerius Photonics to FLIR. Nexus to IonQ is his fourth. h/t @netcreat, whose piece on what Nexus actually builds is the best thing written on it.
Lightsynq is rung three, and the memory part is the part people skip. A photonic link between two modules fails most of the time, because photons get lost in fiber. Without memory, every link in a network has to succeed in the same instant, which stops working the moment you have more than a few. With memory, a successful link is held while the others keep retrying. That is the difference between a two-node demo and a network.
Bhaskar, Machielse and Levonian built the first memory-enhanced quantum repeater in Lukin's lab at Harvard, Nature 580, 60, using silicon-vacancy centers in diamond. They founded Lightsynq in 2023, IonQ closed it in June 2025, and De Masi described the deal at the time as "the shift from experimental bulk optics to scalable optical chips."
He said the quiet part out loud fourteen months ago. Everything since has been execution on that sentence.
SkyWater, closed July 31, is where the three layers meet: the trap, the photonic circuit, and the control electronics.
It already lists trapped ions, photonics and high-performance waveguides among its quantum capabilities, plus advanced packaging: silicon fan-out and heterogeneous integration.
Being a DMEA Category 1A Trusted Foundry means the same line can serve both commercial and classified work.
Instead of ordering three different chips from three vendors on three schedules, IonQ can now co-optimize them in-house.
Then the people. I went through IonQ and its group companies role by role, before either July acquisition landed. More than 120 do photonics as their actual job, hired out of PsiQuantum, Quantinuum, Xanadu, Photonic Inc, ASML, MIT Lincoln Lab, JPL, Thorlabs, TOPTICA, Ayar Labs, Lightmatter. Bart Machielse now runs photonics technologies.
The newest one names the structure himself. Christian Dangel, Senior Staff Engineer and Lead Characterization since May, is hiring two senior photonic test roles for IonQ's Quantum Interconnect Division in Boston. He spent four years and one month at Photonic Inc in Vancouver: Quantum Device Engineer, Device Metrics Lead, Staff Quantum Device Engineer, Manager of Quantum Devices, then Director of Chip and Test Engineering. Before that, Cambridge, as a visitor in the Semiconductor Physics Group at the Cavendish. Physics at TUM and EPFL.
He left Photonic in May. Photonic closed over $200M at a $2B valuation that same month.
That 120 is a floor, not a total. It predates both closings. Nexus lands on top of it, and so does every process and integration engineer inside a semiconductor foundry, which is a different discipline I did not try to fold in.
Now put the published roadmap next to all of it. 10,000 physical qubits in CY27. 200,000 in CY29. Two million in CY30.
You do not deliver free-space beams to ten thousand ions from one central zone, and no single chamber holds two hundred thousand. Those are not targets, they are constraints, and each one names something that has to exist before the number does.
Call the roadmap aggressive. It is. Just notice that nothing on it has been left without a purchased answer.
Photonics is not what EQC replaced. It is what EQC left behind, and it is the next thing to go on the chip.