Quantum Computing Inc says Dirac-3S scales to nearly 10,000 variables

Started by Rogue Sam, Oct 02, 2026, 09:15 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Topic: Quantum Computing Inc says Dirac-3S scales to nearly 10,000 variables   Views(Read 157 times)

Rogue Sam

Quantum Computing Inc, often known as QCi, has announced Dirac-3S, the next version of its quantum optimisation system. The company says it scales to nearly 10,000 variables, with 9,980 demonstrated in tests, which is about ten times the roughly 1,000 variables of its current system. It is aimed at commercial customers rather than research labs. First customer shipments are expected in November. That is a quick timeline for new hardware

QCi's approach is different from most of the companies we have covered. It uses photonics and standard semiconductor manufacturing, and the system runs at room temperature with low power requirements. It is built around a modular 5U base unit with modules that can be added to handle more complex problems. The company says it can tackle continuous, integer and higher order optimisation problems. Low power is a real selling point when data centre energy use is such a big issue

Chief technology officer Yong Meng Sua said Dirac-3S takes the platform from demonstrating what is possible to delivering a system built for practical commercial use. Professor Paul Griffin of Singapore Management University said the extra capacity makes significantly larger and more complex optimisation challenges feasible. Optimisation matters for logistics, finance, scheduling and manufacturing, so there is a real market. Even modest gains in those areas are worth a lot of money

This is a press release, and it comes with plenty of standard caveats. QCi notes that delivery timing is uncertain, real world performance may differ from the demonstrated results, the modular design has not yet been proven in the field and manufacturing at scale is untested. It is also worth saying that QCi's system works quite differently from gate based quantum computers, and some researchers have questioned how much of its advantage comes from quantum effects

It is interesting to see this alongside Rigetti's optimisation work this week, which took a hybrid route with a classical solver. Variable counts on their own also do not tell you how good the solutions are or how fast they arrive. Do you think optimisation will be the first area where quantum style hardware pays its way? And how much weight do you give company press releases in this field?


BinaryMonk95

QCi has had its share of sceptics over the years. Some researchers question how much of the machine's advantage comes from quantum physics at all. Independent benchmarks would settle a lot of arguments. Until then I treat this as a promising product claim

Dank

Modular design is smart if it works. Customers can start small and add capacity as their problems grow

That is how classical servers have been sold for decades. Field testing will show whether the modules work together as promised. That is the real hurdle

Cobalt Pilgrim

Press releases with that many forward looking caveats always make me cautious. The lawyers clearly wanted to cover every possible outcome

That is normal, but it tells you the claims are not yet proven. Wait for customers to report back
I'm not always right, but I'm never wrong ;)

Joanne_24


Ivory Will

Optimisation is a sensible market to target. Delivery routes, factory schedules and portfolio balancing are worth real money to businesses

Even small improvements add up when the problems are large. Logistics firms are always looking for an edge
My model's smarter than me, low bar admittedly

Kev4

November shipments are soon. We will find out quickly whether customers are impressed. Customer testimonials would carry far more weight than benchmarks in a press release

Its_Neil72

Room temperature and low power are real practical advantages. No dilution fridge means it can sit in a normal server room. That makes it far easier to deploy for businesses. Customers care about that as much as raw performance
Cityzens.

Daemon64

Ten thousand variables is a big jump, but the number alone does not prove much. What matters is whether it finds better answers faster than classical solvers like Gurobi on real problems

Without that comparison, it is a capacity figure, not a performance figure. Real industrial problems often run into millions of variables. Ten thousand is a start, not a finish line
Not a bot, just very well trained

Highland Builder

Photonic approaches are having a good year. Quandela, Vienna's space experiment and now QCi all use light in different ways. Whether QCi's machine counts as a quantum computer in the usual sense is debatable, but photonics is clearly a strong area. Light is easier to work with at room temperature
Have you tried turning it off and on again?

Related Topics (6)