← Writing

Quantum Is a Bill Before It's a Business Case

McKinsey sizes quantum computing at up to $2.7 trillion by 2035. For most companies the first five years are a security bill with federal dates on it. Here's what's known about the size of that bill, and where it goes in a 2027 budget that's growing 3.7 percent.

Is quantum computing coming fast, or is it something most companies don’t need to worry about yet?

Almost every executive I talk to is heads-down on AI right now, and I get it. Quantum has only just started coming up in my conversations, and that question is how it comes up.

My answer is both. As a growth story, quantum can wait for most companies. As a line in the budget, it can’t, and the first version of that line belongs in the 2027 plan you’re closing this quarter.

The payoff from quantum depends on a machine nobody has built yet. The bill depends on deadlines that are already in writing. So the bill gets here first.

New to quantum? Click here for Quantum 101 in plain English. Two minutes, no physics.

Quantum 101

If a sentence in this box needs a physics degree, I failed.

What is a quantum computer? It isn’t a faster version of the computer you own. It’s a different kind of machine that is spectacularly good at a short list of problems and useless for almost everything else. It won’t speed up your email or fix your spreadsheets.

Why does anyone outside a lab care? One problem on that short list is cracking the math behind most of today’s encryption. That’s the math that lets your browser talk privately to your bank and lets your phone check that a software update is genuine. A big enough quantum computer could work out the secret keys.

Does that machine exist? No. Nobody has built one big enough, and serious people disagree about when anyone will.

Then why is this a problem now? Two reasons. Encrypted data can be copied today and unlocked later, so anything that has to stay secret for years is exposed to a machine that doesn’t exist yet. And swapping out encryption across a whole company takes years.

Is there a fix? Yes, and it’s less exotic than it sounds. The fix is new math that runs on the computers you already own. NIST, the U.S. standards agency, published the first versions in 2024. People call it “post-quantum” or “quantum-safe” encryption, and those mean the same thing. The hard part is finding every place your company uses the old math.

Two terms you’ll see below. Encryption does two jobs. Key establishment is how two computers agree on a secret so nobody else can read what passes between them. A digital signature is how one computer proves to another that a message or an update is the real thing. The federal deadlines treat the two separately.

The forecast and the deadline

On April 28, McKinsey published its Quantum Technology Monitor for 2026. It says quantum computing “has reached a commercial tipping point” and that “innovation-driven companies can no longer afford to wait and see.” It sizes the prize at up to $2.7 trillion of economic value worldwide by 2035.

Eight weeks later, on June 22, the White House signed Executive Order 14412. It gives federal agencies until December 31, 2030 to switch their most important systems to quantum-safe encryption for key establishment, and until December 31, 2031 to do the same for digital signatures.

$2.7Tof economic value quantum computing could create worldwide by 2035McKinsey, 2026
2030first federal deadline for quantum-safe encryption on the government’s most important systemsExecutive Order 14412, 2026

I worked in technology strategy at McKinsey, and I have no quarrel with the forecast. My argument is about order. One of those documents describes money a few industries might make in the 2030s. The other is homework with a due date, and homework with a due date gets done first.

The payoff belongs to a few industries, and it comes later

Read the forecasts closely and they say this themselves. McKinsey’s sentence is about “innovation-driven companies.” BCG told chief executives in June that quantum is getting real, and in the same piece wrote that the initial value pools “will be resource- and IP-intensive and therefore concentrated among a few, large players.” BCG’s long-term estimate is $450 billion to $850 billion in value creation. Its estimate for the quantum market itself is approximately $2.5 billion by 2030, with $5 billion as the optimistic case. McKinsey counts more than $1 billion in revenue for quantum computing companies in 2025.

If you run a drugmaker or a chemicals company, the advice that follows from those numbers may be right for you: pick the problems quantum could solve, find a hardware partner, build a team. If you run a regional bank or a manufacturer, those early winnings aren’t yours, and the first thing quantum will do to your company is cost it money.

I watched the AI forecasts play out the same way. In June 2023 McKinsey estimated that generative AI could add the equivalent of $2.6 trillion to $4.4 trillion annually. Two years later MIT researchers found that roughly 95 percent of generative AI pilots were failing to deliver measurable P&L impact. I wrote about what that number means in the spring. Both findings can be true at once. A forecast tells you what a technology could be worth to the whole economy. It doesn’t pay for anything at your company, and it doesn’t tell you what to do first.

The takeaway: the quantum value forecasts are about a handful of industries and a horizon past 2030. For everyone else, treat them as background and budget for what has a date.

The deadlines are already written down

Nobody can tell you when a quantum computer will be able to break today’s encryption. In its latest survey, the Global Risk Institute asked 26 experts and reported that they consider such a machine “quite possible (28-49%) within the next 10 years” and “likely (51-70%) in the next 15.” The report says those experts believe the timeline has sped up. Read the same numbers the other way and the experts think it’s more likely than not that the machine doesn’t show up this decade.

The people who set deadlines read the same odds and picked dates anyway.

Nobody on that list is predicting the machine arrives in 2029. These are dates for finishing the switch. The people who set them decided the switch takes so long that they can’t afford to wait and find out.

The dates in writing
Nobody knows when the machine arrives. The calendar is already full.
Oct 22, 2026Federal agencies owe OMB a migration planand an estimate of money and peopleDec 2026Proposed rule for federal contractors due(180 days from the executive order)2027New commercial national security systemsmust support quantum-resistant algorithms2029Google’s and Cloudflare’s targetsfor their own migrationsDec 31, 2030Key establishment on high-value federal systems.Proposed compliance date for covered contractors.Dec 31, 2031Digital signatures on high-value federal systems2035Final phase of the federal migration
Sources: Executive Order 14412, June 22, 2026. OMB Memorandum M-26-15, June 24, 2026. NSA, October 1, 2026. Google, March 25, 2026. Cloudflare, April 7, 2026. The October 22 and December dates are my count of 120 and 180 days.

Why the hurry, if the machine might be 15 years away? Because of a trick with an ugly name: harvest now, decrypt later. Somebody copies your encrypted data today and sits on it until a machine exists that can open it. The executive order names the risk in its first section, describing adversaries “collecting United States information now” and decrypting it later. Google’s security engineers wrote that the threat to encryption is “relevant today” for the same reason. I haven’t read a source that documents a specific theft of this kind, so treat it as an assumption a careful board should make. Quantum may not feel like it’s on our doorstep. Our data is already on theirs.

The cryptographer Michele Mosca gave boards the cleanest way to think about this back in 2015, and it fits on a napkin. Take the number of years your data has to stay secret. Add the number of years it will take you to switch to quantum-safe encryption. If that sum is larger than the number of years until the machine exists, then in his words, “we have a serious problem today.” You can’t know the third number. You can know the first two this quarter.

The third number has also been moving in one direction. In 2019 a Google researcher co-published an estimate that breaking 2,048-bit RSA, a widely used kind of encryption, would take a quantum computer with 20 million qubits. Qubits are the quantum version of a computer’s bits, and more of them means a bigger machine. In May 2025 the same researcher cut his estimate to less than a million.

Quantum may not feel like it’s on our doorstep. Our data is already on theirs.

The takeaway: the dates that will reach your contracts, your audits and your regulators are 2027, 2030, 2031 and 2035. Plan to those dates and let the physicists argue about the machine.

What’s known about the size of the bill

Less is known than I’d like. I looked for a sourced benchmark for what this switch costs a single company and didn’t find one I’d put in front of a board.

Two public numbers give a sense of scale. In July 2024 the White House projected about $7.1 billion, in 2024 dollars, to migrate the federal government’s prioritized information systems between 2025 and 2035. That figure leaves out national security systems, and the report calls it “a rough order of magnitude rather than precise calculations.” Then last October, Forrester predicted that quantum security spending would exceed 5 percent of the overall IT security budget in 2026, on things like consulting services and tools that find and list where a company uses encryption. That one is a prediction about this year, and I haven’t seen it measured.

$7.1Bprojected to migrate prioritized federal systems, 2025 to 2035, national security systems excludedWhite House, 2024
5%+of the IT security budget predicted to go to quantum security in 2026Forrester, 2025

What I can describe is the shape of the bill, because it’s the shape of every large technology migration. The first part is cheap: finding where your encryption is and deciding which data matters most. The second part is expensive: fixing the old systems nobody wanted to touch. AI spending followed the same curve. The costs that hurt arrived after the introductory pricing ended and the pilots had to run for real.

That shape tells a board how to get a number it can trust. Ask for the estimate for the whole job, as a range, before approving the cheap first step. A board that approves the first step on its own will be surprised twice.

Some of the news is good. The replacement encryption exists. NIST finalized the first three standards in August 2024 and encouraged administrators “to start integrating them into their systems immediately.” Cloudflare says more than 65 percent of the human traffic on its network is already protected by post-quantum encryption, which means your browser did part of this for you while you weren’t looking. In January, CISA, the federal cybersecurity agency, published the product categories where quantum-safe options are widely available, including cloud services, browsers and web servers, messaging tools and full-disk encryption. Much of the remaining work will arrive as vendor updates, as long as your contracts ask for them.

The takeaway: nobody can price this for your company yet, and that includes me. What a board can do today is ask for the whole estimate before funding the first step.

Where it goes in a budget growing 3.7 percent

Last week I wrote about the 3.7 percent problem. Gartner projects that IT budgets will grow an average of 3.7 percent in 2027 while funding for agentic AI grows 31.8 percent, so every AI dollar in next year’s plan comes out of another line. I suggested writing down what each AI line replaces.

The quantum line can’t pass that test, and it’s better to say so up front. It doesn’t replace anything. It prevents a loss nobody can put a date on, which makes it the hardest budget conversation in technology and one that AI never had to have. Run it like Y2K with a moving date.

In a flat budget, a line with no return and no fixed date gets funded last. That’s how AI overruns get paid for now: late in the year, by whoever is short. A job with a 2030 date on it shouldn’t be funded that way, so the line needs to be in the plan before the AI lines eat the whole raise.

The 2027 version of the line is small. It pays for a list and a contract clause.

The Quantum Line in a 2027 Plan

  1. An owner. One named executive, inside the cybersecurity oversight you already have. Don’t form a quantum committee. Nobody needs another committee.
  2. An inventory. A list of where encryption is used in the systems that matter most and in every connection to the outside world, including an honest note on what the list can’t see yet.
  3. A shelf-life map. How many years each kind of data has to stay secret. This is the number a board should argue about.
  4. A procurement clause. Anything you buy that will still be running in five years supports the new NIST standards, or comes with a documented upgrade path. Your security team knows the standards as FIPS 203 and 204.
  5. The whole estimate. A range for the full switch through 2031, shown before the first step is approved.

The growth side gets a watching brief. One person spends a quarter-day a month on it and writes the committee a short memo twice a year. Give that brief one trigger for funding a pilot: a vendor shows, with proof, that a quantum machine beat the best conventional method on a problem your company has today.

That’s less than the big firms recommend. BCG’s June piece advises chief executives to identify three to five high-value use cases, invest early in a team or center of excellence, and form strategic partnerships now. For the few, large players, that may be the right first move. For most companies it’s the same sequence that produced the AI pilots. The right advice here is boring, and boring is hard to invoice.

The best argument against

The strongest objection is that the machine may be a long way off. The experts in the Global Risk Institute survey put the ten-year odds below 50 percent. Spending real money against a threat with no date is how security budgets fill up with software nobody uses.

I’d grant most of that, and the 2027 line still stands, for two reasons. First, the deadlines don’t depend on the machine. If you sell to the federal government, a proposed rule with a 2030 date is on its way to your contracts, and I’d expect large customers to borrow the language. Second, these changes take years. NIST notes that an older security algorithm called SHA-1 has been in use since 1995, and its final phase-out date is the end of 2030.

Being early is also cheap. A list of where your encryption is and a map of how long your data has to stay secret are things a well-run company should have anyway. Not knowing where the data was is what stalled a lot of AI pilots.

Before your 2027 budget closes, ask management two questions. Where is our encryption? How long does each kind of data we hold have to stay secret? If nobody can answer, you’ve found the first line item.

A board doesn’t need to understand qubits. It needs to know where its encryption is, how long its secrets have to last, who owns the plan and what the plan costs.


How RLK Can Help

I brief boards and leadership teams on quantum in plain English, and I help CIOs and CFOs put a defensible first number on the migration. Start a conversation.


Sources

Ryan King

About the author

Ryan King

Fifteen years in technology strategy at McKinsey and Deloitte. Now running RLK Consulting: enterprise-caliber tech strategy, one strategist doing every hour of the work. Over $10B in documented value capture across 50+ engagements and 12 industries.