Disclosed Defence Money for Quantum Computing Roughly Doubled to Tripled in 2026 on Two Announced DARPA Deals

Defence money in quantum computing is easy to overstate. Headlines quote caps of $300 million and deals worth billions, but the money actually promised to named firms is far smaller than either. Counting priced awards on the register or announced by firms, defence buyers put $80 million into the field in 2023, $38 million in 2024 and $67 million in 2025.

Then 2026 changed the scale. By 11 October the disclosed total stood at $157 million or $188 million, depending on whether PsiQuantum’s new agreement includes its earlier one. All of the rise is announced money from two DARPA agreements, with PsiQuantum and Diraq, while obligated money on the register fell from $35 million in 2025 to $12 million so far. Two more DARPA agreements that year carry no stated value, IBM’s place in the final stage of its test and a programme on mixing qubit types. IonQ’s and Atom Computing’s final-stage agreements state only the $300 million cap, which is counted as a ceiling.

Disclosed defence money for quantum computing ran at tens of millions of dollars a year until 2026

The count below includes every defence award for quantum computing that names a price in US dollars. Most come from USAspending, the federal register of contracts and grants, which records the money set against each award, and the rest are sums that firms gave in news releases and filings. The table keeps the two kinds apart. Ceilings, unpriced deals and other currencies are left out.

The first three years did not move together. The total fell from $80 million in 2023 to $38 million in 2024, then rose to $67 million in 2025 when PsiQuantum announced its first $32 million DARPA agreement. Each year brought 4 to 5 itemised awards plus between 13 and 24 smaller grants and study contracts. Obligated money on the register fell each year, from $55 million in 2023 to $35 million in 2025.

The totals are also uneven. An award’s whole lifetime value sits in the year it started, even if the work runs for three years. One large grant can therefore move a whole year, as Maryland’s $27 million did in 2024.

The year before is itemised only. There was no keyword sweep, so 2022 is not directly comparable. In 2022 three priced awards came to $67 million. The largest was $32 million for the Johns Hopkins University Applied Physics Laboratory to help run DARPA’s earlier test of quantum computer designs. That is work done for the government, not a payment to a quantum firm. The Air Force also paid PsiQuantum $22 million for photonic chip work and IonQ $13 million for a trapped-ion computer, both through its research laboratory.

Two DARPA agreements, with PsiQuantum and Diraq, account for almost all of 2026

In July 2026 PsiQuantum said it had signed “a new $125 million expanded agreement” with DARPA. It falls under the Quantum Benchmarking Initiative, the agency’s test of whether any company can build a quantum computer worth more than it costs by 2033. On 7 October (8 October in Sydney) Diraq, an Australian firm building silicon qubits, said an initial US$51 million funding tranche would pay for its next year in the same programme. Two small Air Force laser contracts and about $8.4 million of smaller grants, all to universities, make up the rest.

The PsiQuantum figure has two possible readings. In September 2025 the firm announced a $32 million agreement to continue testing in the same final stage, and its July release does not say whether the $125 million includes it. If it does, the 2026 total is $157 million. If it does not, the total is $188 million. Our report on the PsiQuantum agreement gives the company’s account.

The multiple also depends on two counting choices, so the table below shows each. IonQ’s 2024 contract and Rigetti’s 2025 contract are for networking quantum systems, and we leave them out of the main count, though a case can be made for either. Maryland’s second award started on 2024-12-26, six days before 2025. Moving it changes the comparison with 2025 but not the comparison with the 2023 to 2025 average, which is why the average is the better yardstick.

Under the main count, 2026 is 2.5 to 3.1 times the 2023 to 2025 average and 2.3 to 2.9 times 2025. The average is steadier. Counting all three networking deals, it is 1.9 to 2.4 times the average and 2.1 to 2.7 times 2025. With Maryland in 2025 instead, the 2025 comparison falls to 1.6 to 2.1 times. So “roughly doubled to tripled” is the claim every version supports.

There is a fair objection. DARPA’s earlier agreements with Microsoft and PsiQuantum in its precursor test carried no published price. PsiQuantum’s chief scientist dates the relationship to “our first engagement with DARPA several years ago”, long before any price was published. So part of the 2026 rise may be a rise in disclosure rather than in money. The series cannot separate the two until DARPA’s R-2 budget books, the line-by-line justifications it sends to Congress each year, can be read, and we could not reach them.

Diraq is the only firm in the final stage to announce a first tranche. Its release also says funding of up to US$300 million would pay for independent checks, known as verification and validation, of its route to a large, error-corrected machine. That second number is a ceiling. Our report on Diraq’s move covers the announcement.

DARPA’s final stage holds six firms, and its $300 million figures are caps, not payments

DARPA’s programme page sets three stages with different limits. Firms can be awarded up to $1 million in the first, up to $15 million in the second and up to $300 million in the third. In April 2025 DARPA chose nearly 20 companies for the first stage. Rigetti’s filing described its own place there as worth up to $1 million on completion of milestones.

Six firms now sit in the final stage. Microsoft and PsiQuantum moved in from an earlier DARPA test, and Atom Computing, Diraq, IBM and IonQ joined them on 7 October 2026. IonQ calls its deal one with a potential value of up to $300 million, and says funding beyond the first payment depends on future appropriations and DARPA’s own funding decisions. Atom Computing speaks of potential funding of up to $300 million. IBM and Microsoft have given no amount.

Six times $300 million is not a budget. DARPA publishes no programme total, so a sum of the six would describe only the most that could ever be paid, and only if every firm reached its full cap. Of the six, only PsiQuantum and Diraq have named money. Our reports on the IonQ and IBM moves give each company’s account.

The middle stage is larger and cheaper. By our count of DARPA’s announcements, eleven firms reached it, each able to win up to $15 million, and by the same count seven remain there, among them Quantinuum, QuEra and Xanadu. Congress wrote the programme into law in the FY2025 defence act without a dollar figure. DARPA also runs smaller computing work, such as a 2026 programme on combining different kinds of qubit, which named 19 teams from 15 groups and no amounts.

The Air Force signs the most awards, while DARPA is larger by dollars

Four paymasters cover every priced award since 2023 once the smaller grants are counted. The Air Force, mostly through its research laboratory and science office, signed 55 of the 99, worth $135 million in all. It is the most frequent buyer. DARPA signed only 4. It is larger by dollars, at $176 million or $208 million, because its agreements with PsiQuantum and Diraq are each far bigger than any grant. The Army signed 25, worth $21 million, and the Navy 15, worth $8.7 million, most of them research grants.

Two Army cooperative agreements are left out as intelligence money. The University of Sydney’s was made for IARPA, the spy agencies’ research arm, and its title says so. Innsbruck’s probably was too. IARPA lists both as lead teams in the same programme, and both awards were signed the same day with the same dates of work.

IonQ’s deals sit on both sides of the line. Its 2023 filing says the $26 million deal with the Air Force Research Laboratory is to “deploy two barium-based trapped ion quantum computing systems”, so it is counted. Its $55 million contract of 2024 is for “the scaling, networking, and deployability of quantum systems”, and like Rigetti’s $5.8 million networking contract of 2025 it is left out of the main count. Neither IonQ deal appears under its name in the federal register. IonQ says that following its first laboratory contract in 2022 it won “three additional contracts … with a total value of over $94.4 million”. That figure overlaps these deals and is not added to them.

Some security money comes as shares. In-Q-Tel, the venture fund of the US spy agencies, is named among the backers of D-Wave and Rigetti, but neither sum has been made public. It is intelligence money, so it is left out of every total.

The biggest company recipient is PsiQuantum. Its defence awards with a price since 2023 come to $136 million or $168 million, including a $11 million Air Force contract in 2025 for optical parts for its light-based chips. Rigetti won a run of smaller Air Force orders and grants.

Universities took most of the money we could count in 2023 and 2024, then companies took over

Universities did well at first. Of the money we could count, they received $53 million of $80 million in 2023, about 66 per cent, most of it in smaller research grants. The largest single award went to the University of Maryland for an Air Force project on running test quantum computers safely for research. The Army also paid New South Wales in Australia $4.6 million for qubit research.

In 2024 Maryland won a second phase worth $27 million, and universities took about 90 per cent of the year’s total. In 2025 and 2026 every itemised award went to a company, as DARPA’s test of company designs and Air Force orders for lasers and other parts took over the list. Universities still won about $13 million and $8.4 million in smaller grants. That was about 20 per cent of the 2025 total and 4 to 6 per cent of 2026’s.

Outside America the computing money is in pounds and euros, and mostly undisclosed

Britain’s defence laboratory, Dstl, paid three firms for quantum computing studies in 2022, worth about £530,000 together. PA Consulting looked at how quantum computing might speed up military operations, while Frazer-Nash studied how to run sensors and QinetiQ looked at uses in surveillance and scouting. In 2023 a further £85,000 went on two small awards, one of them for light-based work that may not be computing. These are pounds, kept out of the dollar totals.

Britain also owns a machine. In 2022 the Ministry of Defence bought a small ORCA Computing system at an undisclosed price, which the company and the BBC called the government’s first quantum computer. Elsewhere the numbers thin out. France’s PROQCIMA defence programme has a ceiling, not a budget. It allows a maximum of €500 million over at least 10 years, with €60 million for phase 1 and up to €150 million for phase 2. Pasqal left the programme after its first phase, as it said on 28 September 2026.

India’s defence research agency signed a deal with Zero mK, a start-up in India that builds very cold fridges for quantum machines, but neither side gave a value. Germany’s Cyberagentur runs a programme to build a mobile quantum computer for defence and security, with no budget or firm named. Neither has a price.

Industry deals add more gaps. Lockheed Martin has research links with PsiQuantum, IBM and Xanadu, and Anduril and Davidson Technologies are working with D-Wave on missile defence. None of these deals has a stated price. Davidson, a defence firm in Alabama, has also hosted a D-Wave machine since November 2025, D-Wave’s annual report says, again at a price nobody has disclosed.

The money buys tests, parts and research hardware, and only obligations on the register are firm

DARPA’s money pays for an outside team to check whether a firm’s design could work at scale, and PsiQuantum’s 2025 agreement was to continue on-site testing. The Air Force buys parts. In 2026 it paid N.P. Photonics and Cybel for lasers tuned to the barium ions used in trapped-ion quantum machines.

Other awards pay for research. The Air Force bought two IonQ systems in 2023. Rigetti’s 2023 award was a subscription ordered under its Air Force contract for custom quantum chip fabrication, in effect a foundry subscription. The Army paid SEEQC for chip design work, and DARPA paid QuEra for a small testbed study. Eaton, the power equipment maker, won a $7.0 million Air Force contract that the register describes as “quantum computing for resiliency and security”.

Rigetti’s smaller awards show the same mix. The Air Force paid it for custom chip fabrication and, through its basic research office, for a $2.0 million grant on flaws in the Josephson junctions at the heart of its qubits. Of that grant, $1.5 million passes to 3 subawardees.

No armed force has publicly ordered quantum computers for use in operations. IonQ’s two systems went to a laboratory, and Britain’s ORCA machine was a single small system bought to learn from.

For an investor, firmness matters more than size. The firmest figure is an obligation on the federal register, which anyone can check. Next comes an announced tranche such as Diraq’s $51 million, which the company says will fund its next year of work but which no public register shows yet. Weakest are caps and “expanded” totals, which may repeat money already counted.

A $300 million cap is not money in hand. IonQ says funding beyond its first payment depends on future appropriations and DARPA’s own decisions, so a firm that plans its hiring against the cap is counting on stages it has not yet passed. Government work can still be most of a small firm’s revenue. Rigetti’s annual report says governments of every kind supplied 90 per cent of its 2025 revenue and US government bodies 48 per cent, though it does not say how much of that is defence.