Key points

• Running the same circuit on two quantum vendors usually means two accounts, two SDKs, two billing systems and two ways of counting cost.
• EigenOrbit Router is a multi-vendor quantum API. It puts devices on Amazon Braket and IBM Quantum behind one API key and one prepaid balance, and shows the price of every job before it runs.
• With eigenkit, moving a circuit from a simulator to a real QPU means changing the device string.

Most quantum algorithms are developed on one vendor’s stack. The team learns its SDK, sets up its billing and gets used to its queue. When someone later asks how the same circuit would do on a different machine, the honest answer is often “we’d need a week to find out”.

That week has little to do with physics. It goes into a new account, a new SDK, another set of credentials, a different pricing unit and a fresh round of reading documentation. For a team that wants to compare hardware regularly, that overhead adds up fast.

Why is it hard to run the same quantum circuit on different hardware?

Every vendor packages access differently. The main differences are:

• Accounts and credentials. Each provider has its own sign-up, keys and permissions, and many companies need a procurement step for each one.
• SDKs and formats. Circuits written for one stack often need rewriting, and results can come back in different shapes or bit orders.
• Pricing. Some vendors charge per shot, some per task, some per second or minute of processor time. Comparing the cost of the same experiment across machines takes a spreadsheet.
• Spend control. A loop that submits more jobs than intended can run up a bill before anyone notices, and each vendor has its own way of setting limits.

None of these problems is hard on its own. Together, they make teams stay with the hardware they started on, even when another machine might suit the problem better.

What is a quantum router?

A quantum router sits between your code and the vendors. You send jobs to one endpoint with one key, and the router passes them on to the device you name.

EigenOrbit Router is a multi-vendor quantum API: a single access point to quantum computers and simulators from different providers. It sends jobs to devices on Amazon Braket and IBM Quantum, with one API key and one prepaid balance. The full list of devices is in the documentation.

How does pricing work?

Each job’s price is shown and reserved before it is sent. You can ask for an estimate at no cost, and when you submit, the estimated amount is held on your balance. After the job finishes, the actual cost is charged.

Every price is in one unit, US dollars, whichever vendor runs the job, so the cost of the same experiment on two machines can be compared directly. On a QPU, the final charge is never more than the amount held. The details are in the Router documentation.

If a job is rejected, fails, or is cancelled before it reaches the vendor, the hold is released in full. A job that fails inside the router after the vendor ran it is not charged.

Each key can also have its own spending limit. A student project, a CI pipeline and a research team can each get a key with a cap, and once a key reaches its cap, the router stops accepting new jobs from it.

How do you switch quantum hardware in one line?

Router is built into eigenkit, EigenOrbit’s open-source library for writing a quantum algorithm once and running it on many backends. Installation and setup take two lines:

pip install "eigenkit[router]"
export EIGENORBIT_ROUTER_API_KEY=eo_live_...

A circuit is built once and given a provider, which names the device:

provider = RouterConfig(device="braket/sv1", shots=1000)
circuit = Circuit(spec, all_z(2), provider=provider)
values = circuit(weights, inputs)  # one router job

To move from the Braket simulator to IQM’s 20-qubit Garnet processor, only the device changes:

provider = RouterConfig(device="braket/iqm_garnet", shots=1000)

The rest of the code stays the same. According to the documentation, eigenkit returns results in the same shape, data type and bit order as the same code run with your own vendor credentials. On real hardware the numbers will of course differ from run to run, as shot-based results always do.

Teams that do not use eigenkit can call the HTTP API directly: submit a job, check its status and fetch the result.

Where Router fits

Router is one part of a larger idea at EigenOrbit: comparing quantum hardware should be routine, not a project. Benchmarks helps choose which machine might suit a workload. Router then makes it practical to run the same code on that machine and on the one you used before, and to see what each run cost.

About EigenOrbit

EigenOrbit builds quantum labs for companies. Each lab is set up around the client’s own problems: it tests every quantum idea against the client’s classical solver and runs what wins on the right machine. Router, Benchmarks and the open-source eigenkit library are part of every lab, and some of them can also be used on their own.

Sources

• EigenOrbit Router
• Router documentation: quickstart
• Router documentation: billing
• eigenkit on GitHub

Key points

• A small in-house quantum team in Western Europe costs roughly €250K to €440K a year in salaries alone, before hiring time and employer costs.
• A one-off pilot is cheaper to start but answers one question, and the next question means a new contract.
• A licensed lab sits in between: an outside team works on your problems over a year, and the methods and code stay with you.
• The right choice depends mostly on how many quantum questions you expect to ask in the next two or three years.

The first quantum briefing usually goes well. A vendor or an adviser shows where the technology might matter for the business, someone senior asks to look into it, and a budget line appears.

A few weeks later, someone has to decide who will actually do the work. Most companies end up choosing between hiring their own people, buying a pilot from a vendor or consultancy, and licensing a lab that an outside team runs on their problems.

These options are rarely compared side by side, partly because each one leaves the company with something different at the end of the year. Below we put them next to each other, using public salary data, the estimates providers publish for pilots, and the published terms of one lab model.

How much does an in-house quantum team cost?

In Western Europe, a three-person quantum team costs roughly €250K to €440K a year in salaries, and hiring it usually takes months.

Three people is about the minimum for taking a business problem all the way to a quantum or classical verdict. One designs the algorithms. Another follows the research closely enough to tell a real result from a hopeful one. The third makes the code run on actual hardware.

In Germany, Switzerland, the Netherlands and France, recruiter DeepRec.ai puts mid-level salaries at €90K to €120K for a quantum algorithm engineer, €85K to €110K for a research scientist and €75K to €100K for a quantum software engineer. Senior people in the same roles earn €90K to €160K. A three-person team therefore costs about €250K to €330K a year at mid level and up to €440K at senior level. That is salary only. Employer contributions, equipment, cloud and hardware access come on top.

Finding these people takes time. In a 2026 survey by QuEra, reported by IntelligentCIO, 37% of organizations named the workforce shortage as a challenge to adoption. A company hiring here competes with quantum start-ups and hardware makers for the same small group of people.

In return, the company gets a team that learns its business and can switch between problems as priorities change. For a company planning to work with quantum for years, that continuity is hard to buy any other way.

The weak spot is that with two or three specialists, much of the know-how sits in a few heads. If one of them leaves, a large part of the capability can go too, unless methods and code are documented and shared from the first month.

How much does a quantum computing pilot cost?

Providers’ own estimates put a three- to four-month quantum proof of concept at around $100K, and a full pilot at up to $150K–400K, often with hardware time billed on top.

A pilot is the most common first step. A hardware vendor, a consultancy or a specialist firm takes one use case, for example a scheduling problem or a portfolio model, and works on it for a few months. At the end there is a report and, often, a presentation to management.

Pilot prices are rarely published, and the figures that exist come from providers themselves. Yuval Boger, CMO of quantum software company Classiq, has said a three- to four-month proof of concept can often be done for about $100,000 (Classiq). Development firm Gilzor gives broader planning ranges: $15K to $40K for a feasibility study, $40K to $150K for a hybrid proof of concept and $150K to $400K for a full pilot. Both are estimates rather than market data, but they give a sense of scale.

Hardware time is often billed separately. IBM lists pay-as-you-go access at $96 per minute of quantum processor time, and Quantinuum’s machines on Azure start at $125,000 a month, according to the price tracker quantumcomputingcost.com. A pilot that runs many experiments on real hardware can see this line grow quickly.

For a company that is still unsure, a pilot is the easiest way in. Nobody has to be hired, and the commitment ends with the project.

Its limits tend to appear afterwards. A pilot answers the question it was scoped for, and the next question is a new project with a new contract. When the pilot is run by a hardware vendor, the comparison is often between their machine and nothing, rather than between their machine and the best classical method the company already uses. And the code and tuned models frequently stay with the provider.

Pilots also seem to move budgets less than one might expect. In the same QuEra survey, only 9% of organizations said successful pilot results were the main reason their quantum spending went up.

What is a quantum lab, and how much does it cost?

A quantum lab is an outside team of scientists, with its own software, that works on one company’s problems for a longer period, usually a year. Prices vary by provider; ours is €150K a year.

Unlike a pilot, a lab is not tied to one question. It can take a routing problem in spring and a materials simulation in autumn, and it is expected to compare every quantum idea with a strong classical baseline. Whatever it builds along the way belongs to the client.

To give a concrete reference point, our own EigenOrbit Lab is priced at €150K a year, takes up to 12 weeks to set up and leaves the client with the methods, code, tuned classical baselines and a record of every run. Other providers will structure this differently, and the figure is here to make the comparison tangible, not as a market average.

The people in a lab are not your employees, so the company still needs someone inside who owns the topic and turns results into decisions. A lab also depends on the client bringing real problems and data. Without them, there is little for it to test.

How to choose

The most useful question is how many quantum questions the company expects to ask over the next two or three years.

If the answer is one, a well-scoped pilot may be enough. It is worth insisting that the pilot compares the quantum approach with the best classical method you already use, and that the code is handed over at the end.

If quantum is likely to become a permanent part of R&D, with several business units involved, an in-house team starts to make sense. It is the most expensive option and the slowest to start, but it builds capability that belongs to the company.

A lab fits the space in between: companies that expect a steady flow of questions but are not yet sure quantum deserves a permanent team. It can also serve as a bridge, building methods and baselines that a future in-house team inherits.

Whichever route a company picks, two things carry over from one option to the next. Every quantum result should be measured against a strong classical baseline, and whatever is learned should stay with the company.

About EigenOrbit

EigenOrbit builds quantum labs for companies. Each lab is set up around the client’s own problems: it tests every quantum idea against the client’s classical solver and runs what wins on the right machine. If classical methods win, the lab recommends classical.

Sources

• DeepRec.ai: Global Salary Guide, Quantum Computing
• IntelligentCIO: QuEra report shows quantum market shifting from hype to proof-driven investment
• Classiq: How to execute a quantum computing proof-of-concept project
• Gilzor: Quantum app development cost 2026
• quantumcomputingcost.com: Quantum computer price list 2026
• Quantum Computing Report: A four-step process for end users to leverage quantum technology
• EigenOrbit

Key points

• Junior research and analysis is the part of consulting AI absorbs first, and entry-level hiring at top firms has fallen since 2023.
• One in three large companies spent over $10 million on quantum in 2025 (McKinsey).
• Classical methods still win on many business problems today. Where they don’t should be settled by benchmarks.
• Advisory work is likely to become more automated and easier to check.

Ask a consulting team to cut a retailer’s delivery costs, and the proposal will usually name its tools: a routing heuristic, a commercial solver, perhaps a demand forecast on top. What it rarely says is whether anyone checked if part of the routing problem could run better on a quantum computer.

For most problems today, that check would come back in favour of classical methods. The result still has value. The client gets a measured baseline instead of an assumption, and a reason to look again in a year or two.

The missing line in the proposal is a small detail, but it points to a larger shift. Consulting is automating its own work faster than it is changing how that work is sold, and quantum has not yet made it onto the standard list of methods.

How AI is changing the consulting business model

Consulting firms have long been built like a pyramid. Partners win the work, managers run the projects, and a wide layer of analysts reads reports, builds models and writes first drafts. Most billable hours come from that bottom layer.

It is also the layer AI tools reach first. McKinsey says its internal assistant, Lilli, cuts search and synthesis time by up to 30 percent, according to The Register. The same article, citing Bloomberg, reported that the firm was discussing a few thousand job cuts over 18 to 24 months, mostly in support functions.

Hiring numbers point the same way. Revelio Labs found that demand for consultant roles at top firms is about 40% below its 2023 peak. Senior roles grew 55% since 2020 while entry-level hiring fell. By 2025 these firms had more people in AI roles than entry-level consultants, against almost four juniors per AI hire in 2015.

Pricing is where this gets awkward. Hourly billing made sense when analysis took people weeks. Once software does much of it, clients have more reason to ask for fixed outcomes, or for models they keep and rerun themselves after the project ends.

Why quantum computing belongs in strategic advice

Large companies are already spending on quantum, whether or not their advisers bring it up.

McKinsey’s Quantum Technology Monitor 2026 counts more than 300 organizations working with quantum companies, Airbus, E.ON and JPMorgan Chase among them. A third of large companies put more than $10 million into quantum in 2025. Start-ups in the field raised $12.6 billion that year, over six times the 2024 figure.

Without an independent view, most of what a company learns about quantum comes from vendors. Each vendor is building a different machine and has good reason to show it at its best. Comparing their claims across a dozen platforms takes specialist time that few companies have in-house.

On many business problems, well-tuned classical methods still come out ahead. Hardware is improving fast, though, and research on where quantum can help keeps producing new results, especially in optimization and in simulating molecules and materials. Companies already hire consultants for exactly these kinds of problems.

So the useful question is less whether quantum works and more where it beats what the client already runs, and from when. Answering it takes a measurement. The same problem runs on the client’s classical solver and on quantum hardware available today, and the results are compared on time, cost and quality. Often classical wins for now, and the client keeps a benchmark to rerun next year. Sometimes one part of the problem does better on quantum, which is worth knowing early.

What the future of consulting could look like

Advisory work later this decade may look less like a team on site for three months and more like a set of automated tools run by a few senior people.

The tools would take over the repeatable part. A business problem gets turned into a mathematical model, run on suitable classical and quantum hardware, and reported in the same format every time, so a result from 2026 can be compared with one from 2028.

In quantum, this only works if the code is not tied to one vendor, because nobody yet knows which architectures will suit which problems. The tests also have to be easy to rerun. A benchmark ages as fast as the hardware it was measured on.

People would spend their time on what tools handle badly. Picking the problem worth solving is one example, and it is often not the problem in the original brief. Deciding what a better result means for a particular business is another. Then there is the slow work of getting an operations team to trust a plan it did not build.

Questions worth asking

AI and quantum have one thing in common here. Both make advice easier to check. If the analysis is automated and the benchmark can be rerun, a client can test a recommendation instead of taking it on trust. A decision to stay classical can be revisited when new data arrives, and so can a forecast that quantum will pay off by 2028.

For consulting firms and in-house strategy teams, a few questions follow. Does quantum appear in the standard evaluation of optimization and simulation problems, even when the likely answer is to wait? Can the measurements behind a technology recommendation be repeated by someone else? Does the client keep anything reusable once the project ends?

Each company will answer differently. The questions themselves are likely to come up more often over the next few years.

About EigenOrbit

EigenOrbit builds quantum labs for companies. Each lab is set up around the client’s own problems: it tests every quantum idea against the client’s classical solver and runs what wins on the right machine. If classical methods win, the lab recommends classical. Methods, code and tuned classical baselines stay with the client.

The lab combines EigenOrbit’s scientists with its own tools, including QUASAR, an AI research agent that tests quantum ideas against the strongest classical methods, and the open-source library eigenkit, which runs the same code on 12 simulators and quantum computers. The team of nine is headquartered in Luxembourg.

Sources

• Revelio Labs: The industry that advised disruption is being disrupted
• The Register: Hot for its bot, McKinsey may cut thousands of jobs
• McKinsey Quantum Technology Monitor 2026