Haiqu

Here is the standard business plan for a quantum computing startup. Step one: promise a machine that will crack encryption, design drugs, and price derivatives at speeds no classical computer can touch. Step two: explain that this machine requires millions of near-perfect qubits, which do not yet exist. Step three: raise money against the day they will. It is a good plan, in the sense that it has raised a great deal of money. It is a bad plan in the sense that, in the meantime, the actual quantum computers you can rent today are noisy, small, and mostly used to run demonstrations that a laptop could do faster.

Haiqu, a quantum software company founded in 2022 and now headquartered in New York, has decided to skip step three. Its founders - Richard Givhan, who studied engineering physics at Stanford, and Mykola Maksymenko, a physicist trained at the Max Planck Society and the Weizmann Institute - look at the noisy, small machines that exist right now and ask a more awkward question: what if the bottleneck isn't the hardware, but the software sitting on top of it? Their answer is a "hardware-aware operating system" for quantum, and in January 2026 investors handed them $11 million to go build it.

The pitch is refreshingly narrow. Haiqu is not making a quantum chip. It is making the layer between the chip and the person trying to use it - the middleware, the compiler tricks, the error mitigation, the orchestration - so that a given quantum processor can run bigger, more meaningful workloads before it drowns in noise. The company's tagline, "Full Stack Quantum Intelligence," is doing a lot of work, but the underlying claim is concrete: the same physical hardware, driven by smarter software, can do meaningfully more.

The three layers

Haiqu's product, which it launched into enterprise early access in 2026 as an "Agentic Quantum Operating System," comes in three layers, and the layers are the whole story. At the top is Agentic R&D: AI agents that carry Haiqu's quantum expertise and turn a rough research idea into a working prototype in days rather than quarters. In the middle is the Haiqu SDK, which does the unglamorous, essential work of loading data efficiently, compressing circuits, and mitigating errors - the company says it can pack more than a million data features onto 156 qubits and squeeze roughly 100 times more operations out of target hardware. At the bottom is the Haiqu Runtime, an orchestration layer that shuffles work across devices and, Haiqu claims, cuts the cost of running on a quantum processor by up to 1000 times.

Those are big multipliers, and the correct reaction to any quantum startup's big multipliers is a raised eyebrow. Haiqu's answer to the eyebrow is the boring, credible one: it points to work it did on anomaly detection - the kind of pattern-spotting a bank uses to catch fraud - that it says was validated in collaboration with IBM and the Bank of Montreal. In an industry where the currency is press releases, a co-signature from a chipmaker and a bank is worth more than an adjective.

Who is actually using it

The customer list reads like a survey of industries that would very much like quantum to work. Capgemini and Deloitte, the consultancies, are early-access users. HSBC and BMO, the banks, are circling the finance use cases - risk models and Monte Carlo simulations, the mathematics of not losing money. Airbus is on the aerospace side, where quantum simulation and computational fluid dynamics live. Each of them is running the same quiet experiment: can Haiqu's software make a near-term quantum computer earn its keep on a problem they actually have?

That is the interesting thing about Haiqu. It has positioned itself not as a bet on a distant quantum future, but as a toll booth on the road to it. If quantum computing turns out to be useful in five years, the teams doing the useful work will need a software layer between their ideas and the metal. Haiqu would like to be that layer. And if quantum takes fifteen years instead of five, the toll booth still collects, because every one of those years is full of researchers who need to run something, today, on the imperfect machines they have.

There is also the matter of where Haiqu is built, which is: everywhere. The company is Ukrainian-founded, with a core engineering team in Lviv, and it operates across New York, Waterloo in Ontario, and London, with people scattered further still. This is partly a talent story - the physics and quantum-engineering skills Haiqu needs do not cluster in one city - and partly a reminder that the hardest deep-tech problems tend to assemble their teams around the problem rather than around an office. Toyota Ventures liked the wager enough to invest in both the 2023 pre-seed and the 2026 seed, a rare double-down.

None of this guarantees that quantum computing pays off, or that Haiqu is the company that captures the value if it does. The field is littered with elegant software and unmet timelines, and "up to 1000x" is a phrase that will be tested by customers, not headlines. But Haiqu has done something unusual for the sector: it has picked a problem that exists right now, sold a product against it, and gotten serious companies to try it. In a market built largely on the future tense, that counts as news.