Will Quantum Be the Manhattan Project of Modern Science?

The quantum Manhattan Project is a phrase people reach for when they want quantum computing to sound urgent. In March 2026 the former British security minister Tom Tugendhat wrote that quantum “is a modern-day Manhattan Project”. On the evidence the comparison fails. The wartime project spent $2.2 billion between 1942 and 1945, about $10 billion a year in 2025 prices, and as a share of the American economy it equals about $83 billion a year today. America, the biggest spender on quantum that publishes its budget, reaches about a third of the first rate and under a twentieth of the second.

The deeper gap is in what the money buys. A Manhattan Project is an order, in which a government specifies a machine, pays for it to be built and takes delivery. No government has placed one for a quantum computer. The Department of Energy wants a fault-tolerant machine for science by 2028, but it offers prizes for milestones rather than buying the machine. Only $2.5 million of its $215 million came from the 2026 budget year. The rest depends on Congress.

The Manhattan Project cost about $10 billion a year in today’s prices and far more as a share of the economy

The Manhattan Project began in earnest in 1942. That year President Roosevelt approved an initial allotment of $500 million. By the end of the war it had spent $2.2 billion and, at its peak, employed about 130,000 people, according to the Department of Energy. The US consumer price index stood at 18.0 in 1945 and 321.9 in 2025 on the Minneapolis Federal Reserve’s table. On that basis the project cost about $39 billion. That is roughly $10 billion a year. Veterans of Los Alamos went on to devise the Monte Carlo method, published in 1949 and covered in our history. Richard Feynman, who in 1982 made the case for a quantum computer, was one of its youngest group leaders.

That yardstick flatters the present. The American economy produced about $821 billion in total over 1942 to 1945, on the Commerce Department’s figures, so the project took about 0.27 per cent of national output in those years. The same share of the $30.9 trillion economy of 2025 would be about $83 billion a year. The choice of yardstick decides the comparison. By prices it was a mid-sized programme, and by weight in the economy it was one of the largest the country has run.

Apollo, the other programme people reach for, was larger again. The Planetary Society puts Project Apollo at $25.8 billion between 1960 and 1973, about $309 billion in 2025 dollars by NASA’s own inflation index. What the two shared was not size. Each had one goal anyone could state in a sentence, one government customer that would take the result, and a deadline set by events outside the lab.

The world’s public quantum money adds up to one Manhattan Project only on the weakest yardstick

The strongest objection to this guide is that public quantum money, taken together, now rivals the wartime project. It does, on one measure. McKinsey’s Quantum Technology Monitor, the most widely used count, found about $42 billion in public quantum funding announced worldwide by April 2024. That is close to the Manhattan Project’s $39 billion in today’s prices. It is also the sum of more than a decade of announcements by dozens of governments, much of it not yet spent, and its largest single part is China’s disputed $15.3 billion.

Year by year the gap is wide. The table turns each programme into a yearly rate at 2025 average exchange rates from the Federal Reserve. The largest is America’s federal quantum research budget, which the National Quantum Initiative report puts at $1,006 million for 2024, about a tenth of the Manhattan Project’s yearly spend in today’s prices. Germany’s plan comes next at about $848 million a year and Britain’s at about $330 million.

Add everything Washington signed or announced in 2026 to its research budget, at face value, and the federal total comes to about $3.4 billion. That counts the Commerce stakes, DARPA’s $125 million and the Energy Department’s $215 million prize. It is a third of the wartime rate. Counting commitments as if paid in full overstates what America spends, so the gap shown here is the smallest the evidence allows. As a share of the economy it is under a twentieth, because the economy the money comes from is now so much larger.

So the scale objection holds only on the weakest yardstick. Priced by consumer prices, America in 2026 reaches a third of the wartime rate, and only when every commitment is counted as if paid in full. Priced by weight in the economy it is nowhere near, and in either case the money buys research, prizes and shares rather than one ordered machine.

China is the strongest counter-case, and its real spending cannot be seen

If anyone is running a quantum Manhattan Project it would be China, where the state directs research and industry together. The figure usually quoted is $15.3 billion in announced national investment, from McKinsey’s 2022 count. It is disputed. In 2023 Chao-Yang Lu, a leading quantum physicist at the University of Science and Technology of China, was quoted by Foreign Policy as saying the real money was maybe a quarter of that. That would put it near $4 billion.

Neither figure settles it. China does not publish a line-by-line quantum budget, and some of its work is military. What can be seen points away from a single crash programme. Chinese teams publish machines of different designs, such as the Zuchongzhi 3.0 superconducting chip and the Jiuzhang light-based machine. A single wartime programme would pick one. Our guide to China’s quantum companies is further reading. The honest conclusion is that China may be spending more than anyone, but no evidence shows it ordering one machine on a wartime footing.

Quantum has no single goal, so it has no single test of success

Ask what quantum computing is for and the answers split. Some backers want to break today’s encryption, the scenario our guide to Q-Day covers. Others want to simulate molecules for drug makers or materials for batteries, which may need a differently built machine from code-breaking. Still others want the neighbouring technology of quantum sensing, clocks and motion sensors that work without satellites, which our guide to quantum sensing explains. Each points to a different machine.

The Manhattan Project could chase two bomb designs at once because both served one end. Quantum firms chase several ways of building a qubit, the basic unit of a quantum computer, and those designs suit different jobs. With no agreed goal there is no agreed test. So every company can claim progress on its own terms, and no outside judge can say who is ahead on the one measure that would matter to a buyer.

America’s federal push is the nearest thing to a Manhattan Project, and it buys tests and factories, not machines

Since 2025 Washington has acted more like a single buyer than any other government, and it has done so mainly through tests. Its sharpest tool is DARPA’s Quantum Benchmarking Initiative, which asks whether any company can build a quantum computer worth more than it costs by 2033. In April 2025 it chose nearly 20 companies for a six-month first stage. Six firms now sit in the final stage, where an independent team tests the hardware. PsiQuantum and Microsoft moved across from an earlier DARPA programme in February 2025, and Atom Computing, Diraq, IBM and IonQ joined them on 7 October 2026. In July 2026 PsiQuantum signed a $125 million agreement with DARPA. That is the Manhattan method of setting the government’s own bar, without the Manhattan step of buying the winner, and so far the test has kept six different kinds of qubit in the race.

Around it sit the money and the dates. In November 2025 an executive order launched the Genesis Mission, a plan to use AI and big computers for science. By July 2026 the White House had announced more than $5 billion in federal commitments to it from over 15 agencies. The Department of Energy set a goal of a fault-tolerant quantum computer for science by 2028, one that fixes its own errors well enough to run long calculations. Its contest offers up to $215 million for a machine with at least 100 such logical qubits, but only $2.5 million came from the 2026 budget year.

The bigger cheques are for factories and shares. In May 2026 the Commerce Department signed letters of intent for about $2 billion to nine quantum firms in return for minority stakes, as D-Wave’s filing for its own $100 million share confirms. In September it finalised a $1 billion award to Anderon, a new IBM foundry for quantum chips, backed by a further $1 billion from IBM. That is the most coordinated quantum effort Washington has run. It is still not an order for a machine.

Public sources supplied 3 per cent of start-up money in 2025, and Washington answered with equity stakes

Investors now pay most of the bills. McKinsey’s 2026 monitor found that quantum start-ups raised $12.6 billion in 2025, more than six times the year before. Public sources supplied about a third of start-up investment in 2024 and only 3 per cent in 2025. Part of that fall reflects a few very large private rounds, since the total grew more than sixfold while public money stood still. Equity and grants also differ. Investors expect a return and governments usually do not.

Washington’s answer was to become an investor too. The Commerce letters of intent take minority stakes rather than giving grants, so the state is buying a share of the upside instead of buying a product. That inverts the Manhattan model. In 1942 the state paid for everything because no market for an atomic bomb existed. In 2026 the state is one shareholder among many.

Most quantum science is published, though export controls and classified work are growing

The Manhattan Project’s sites did not appear on public maps. Quantum results appear on preprint servers within days, company roadmaps are posted online, and the field’s best-known algorithm, Shor’s method for factoring, has been public since 1994. The race between nations is real, but it is run mostly in the open.

Mostly is the right word. Spy agencies and armed forces do quantum work that is kept secret, and in September 2024 the United States added quantum computers and parts to its export control list. Neither changes the picture of the civil field. The companies that would build a machine publish their designs to win investors and government tests. Governments are moving to post-quantum cryptography because they expect a code-breaking machine one day. They do not act as if one has been built in secret.

One veteran builder has already walked away, which a wartime project would not allow

In a wartime project no contractor could simply decide to stop. In quantum they can. In September 2026 Nikkei Asia reported that the Japanese technology group NEC had stopped developing gate-based quantum computer hardware, after research going back to the 1990s, because a return would take too long. It is keeping its work on quantum annealing, a narrower kind of machine.

That freedom is not a weakness in itself, because it lets money move away from designs that are not working, something a single project often finds hard to do once it has committed. But nobody can promise the outcome. A Manhattan Project was a commitment to finish, and quantum is a set of bets that any backer can call off.

It would take a named order, tens of billions a year and a chosen design

Three things would have to change. A government would have to name one goal, such as a machine that can break a specific code or simulate a specific molecule, and commit to buy it. Germany came closest on paper. Its 2023 plan set a target of a home-built machine with at least 100 qubits by 2026, but it funded research groups towards it rather than ordering one. It would have to fund that goal at tens of billions of dollars a year rather than hundreds of millions. And it would have to pick a design and a team rather than run a contest among many.

None of that has happened, and there are good reasons it may not. No one yet knows which qubit will win, and picking one now could mean backing the wrong horse at a cost no finance ministry would want to explain. If the American mix of DARPA’s test, the 2028 target and Commerce’s stakes turns into a funded order for a machine by a date, the comparison will start to fit.

Policymakers and investors should stop planning around a quantum Manhattan Project

For a policymaker the lesson is about words and money. Calling for a quantum Manhattan Project commits a government to a named order on the scale of tens of billions a year, and no government has shown it will pay that. A call that is not matched by an order and a budget only raises expectations the programme cannot meet.

For an investor or a company buying quantum services, the lesson is about timing. No government has promised to buy a finished machine at scale, so revenue for builders will come from research contracts, cloud access and private customers for some years yet. Plan on the money that exists, such as test programmes, factory awards and equity, rather than on a wartime order that has not been placed.