The machines that train today’s models are asking for more power, more water, more rooms. The curve does not look like something a city can keep feeding.
I am not presenting a quantum computer.
I am looking at approaches that treat quantum not as a distant promise, but as a different language for problems that classical machines either take too long to finish, or cannot speak at all — simulation, optimisation, and some forms of learning.
What holds my attention is the analog path: nature does not compute in discrete steps. A machine that evolves continuously, kept close to those who run it, sits closer to the philosophy I care about than another distant cloud.
Some discoveries
I have come across SME approaches in Barcelona that builds the stack end to end — theory, software, hardware — and puts analog and digital quantum processors next to classical machines. The other discovery is dedicated to quantum communications and their security, which no doubt will be a subject.
Qilimanjaro
A Barcelona team building full-stack quantum computers, with a distinctive analog chip based on fluxonium qubits, alongside a digital QPU and classical HPC.
They speak of a multimodal quantum data center: analog and digital processors sitting with classical machines, so a problem can meet the hardware that fits it. Systems have been placed at the Barcelona Supercomputing Center. A center has opened in the 22@ district. On-prem is part of the same idea — the machine can live where the work already lives, with European control of the data.
The link with AI is not a slogan. Training requirements keep doubling. Their argument is that analog quantum systems — quantum reservoirs, hybrid workflows — can take on parts of learning and inference that burn too much on GPU halls alone. SpeQtrum is the door they open to that stack, from a laptop to an analog QPU.
What holds my attention is not the race for more qubits. It is a European approach that tries to make quantum useful before error-corrected digital machines arrive at scale — and that can sit on-prem.
Dr. Marta P. Estarellas — CEO
Luxquanta
Also a Barcelona team, spun out of ICFO in 2021, that turns quantum physics into keys for the networks we already have.
Their path is Continuous Variable QKD. The keys are generated by the behaviour of light on ordinary optical fibre — the same fibre that already carries traffic. No dedicated dark fibre. The point is not a new network. It is a quantum-safe layer on the one that exists, for governments, hospitals, data centres, operators and anyone holding secrets that must still be secret in twenty years.
What holds my attention sits next to the computing story. If quantum machines become able to read what was encrypted the old way, the data that AI systems move and store is part of what breaks. LuxQuanta treats that as a present problem — harvest now, decrypt later — and answers it with physics rather than with another algorithm to be broken later. European control of the company is part of the same line: the keys, and who makes them, stay on this side of the water.
Vanesa Díaz — CEO