# Y2K: what it costs to be right about a disaster that then does not happen > Somewhere between two hundred and six hundred billion dollars was spent, midnight passed, almost nothing broke, and the people who did the work have been called frauds ever since. The honest account is more interesting than either side's version: some of the fear was well founded, some of it was demonstrably not, and one comparison has never been satisfactorily answered. Source: https://ronutz.com/en/learn/y2k-and-the-prevention-paradox Updated: 2026-09-08 --- ## The bug It began as a sensible economy. Storage and memory were expensive, and a year could be written in two digits instead of four. A programmer in 1970 who wrote `74` rather than `1974` saved bytes on every record, on machines where bytes were the constraint that mattered, and had every reason to assume the program would be replaced long before it mattered. The programs were not replaced. Date arithmetic is everywhere in business software - interest accrual, contract expiry, pension eligibility, inventory shelf life, billing cycles - and all of it is subtraction. Subtract `99` from `00` and you get minus ninety-nine years. A loan matures in the past. A ninety-year-old is minus ten. A perishable item expired a century ago or never will. That much is not in dispute by anyone, and it is worth stating plainly before the argument starts, because the argument is not about whether the bug existed. ## What was spent Estimates vary widely and it is more honest to give the range than to pick a number. Peter de Jager, who did more than anyone to raise the alarm, estimated in 1993 that fixing it worldwide would cost fifty to seventy-five billion dollars. That proved far too low. The United States alone is put at around a hundred billion, including eight and a half billion by the federal government, on the Commerce Department's accounting. Worldwide totals given in the literature run from three hundred billion to six hundred billion dollars, depending on who counted what. Then midnight came, and very little happened. A man in New York received a video rental late fee of ninety-one thousand dollars. Minor glitches were corrected. The infrastructure everyone had been warned about kept running. ## The sceptical case, stated at its strongest It is easy to caricature the sceptics as people who do not understand what averted disasters look like. It is more useful, and more honest, to state their best arguments properly, because two of them are strong. **The embedded systems fear was substantially wrong.** With something between seven and forty billion microcontrollers in the world - the estimates themselves varied enormously - early projections suggested two to three per cent might fail on date rollover. By late 1999 that risk had been revised down to about one thousandth of one per cent, and even that was probably too high. A review of commercial aircraft found no essential systems that were date-sensitive at all. This was the part of the panic that drove people to buy generators and stockpile food, and it did not survive contact with actual testing. **And then there is Italy.** Italy had plenty of computers and a Y2K remediation effort that visibly lagged the American one, and its problems on 1 January 2000 were no worse than anyone else's. The usual rebuttal - that less-prepared countries simply had fewer computers - does not apply to Italy, and that is the single hardest fact for the case that the spending was necessary at the scale it happened. About a third of small American businesses made no preparations whatsoever, and they were mostly fine too. ## The case for the work, also at its strongest **The bug was real and demonstrable.** This was not a theoretical risk being modelled; it was code that could be run in a test environment against a system clock set forward, producing wrong answers on demand. Organisations found the failures before the date arrived because they went and looked. That is a materially different epistemic position from most disaster predictions. **The failures that did occur clustered in systems nobody had fixed.** That is the pattern one would expect if remediation worked, and it is the reason the practitioners insisted afterwards that the quiet was the result rather than the refutation. **And the largest institutions were also the most exposed.** Banks, insurers, utilities and governments ran the oldest date-arithmetic code on the largest scale, and they were the ones that spent most. Whether the marginal hundred billion was necessary is arguable; whether a global banking system running unexamined 1970s COBOL through the rollover would have been fine is a bet nobody was willing to take. ## What cannot be resolved, and why that is the point There is no experiment. The world in which the money was not spent does not exist to be compared against, and Italy is one country rather than a control group. Anyone who tells you the answer is settled - in either direction - is telling you about their priors. That irreducible uncertainty is the actual subject of this article, because it is the daily condition of security and operations work. **A successful prevention produces no evidence of itself.** The reward for the patch that stopped the outbreak, the segmentation that contained the intrusion, the backup that was never needed, is a budget line that looks unjustified next year. The [worm era](https://ronutz.com/en/learn/the-worm-era) shows the other half of the same coin: nine years of outbreaks whose fixes existed and were not applied, because applying them cost something and not applying them appeared, right up until the morning it did not, to cost nothing. Y2K is the largest instance ever recorded of an industry being asked to spend on the basis of a prediction, doing so, and then being asked why it bothered. ## What a practitioner should take from it **Make the failure demonstrable, not theoretical.** The one genuine strength of the Y2K case was that you could show it. A test system with the clock advanced produced the wrong number in front of the finance director. Anyone arguing for a security budget should look for the equivalent: not a threat report, but this system, in this lab, doing this wrong thing. **Scope the fear to the evidence.** The embedded-systems panic did enormous damage to the credibility of the people who were right about the mainframes, and it did so because the claims outran what testing supported. Being right about the important part is not protection against being loudly wrong about an adjacent part, and the audience remembers the second one. **Expect no credit, and say so in advance.** The honest way to ask for prevention money is to state at the outset that success will be indistinguishable from the problem never having existed, and to agree in advance what evidence would count. That conversation is uncomfortable and it is far better held before the spending than after. **And notice what the money actually bought.** A great deal of Y2K expenditure went on systems that were overdue for replacement anyway, and organisations came out of it with a documented inventory of their own software - in many cases the first they had ever had. The [2038 problem](https://ronutz.com/en/learn/the-year-2038-problem) is the same class of defect arriving with decades of notice, and it is being handled quietly, in advance, by people who will also get no credit. ## Sources - [Britannica: an estimated 300 billion dollars was spent, almost half in the United States, to upgrade computers and applications to be Y2K compliant; as it became apparent that systems were intact, relief was followed by accusations that the likely incidence of failure had been greatly exaggerated, while those who had done the compliance work maintained the threat had been real and the continued viability of the systems was proof the effort had succeeded](https://www.britannica.com/technology/Y2K-bug) - [EBSCO research starter: de Jager estimated in 1993 that addressing the problem would cost between 50 and 75 billion dollars worldwide and was far too conservative; the United States alone spent 100 billion including 8.5 billion by the federal government according to the Commerce Department, with the worldwide bill estimated between 500 and 600 billion; with between 32 and 40 billion chips in use their potential for trouble was enormous; about a third of small American businesses made no preparations whatsoever; critics accused the experts of exaggerating the danger to scare businesses into unnecessary spending](https://www.ebsco.com/research-starters/computer-science/y2k-crisis) - [The Straight Dope, on the sceptical case and its limits: remediation is estimated at between 100 and 600 billion dollars; the argument that less developed countries used fewer computers is not credible for Italy, which had plenty of computers and a lagging Y2K effort yet fared no worse; of the 7 to 25 billion embedded chips worldwide initial estimates suggested 2 to 3 per cent might fail, but by late 1999 the risk had been downgraded to 0.001 per cent and even that was likely high; a review of commercial aircraft found no essential systems were date-sensitive; much expenditure may have been litigation defence, and much of the software was due for revamp anyway](https://www.straightdope.com/21343978/was-all-that-money-spent-on-y2k-wasted) - [Slate, January 2000: governments and companies spent between 200 and 500 billion dollars over five years; sceptics point to minor glitches such as a 91,000 dollar video rental late fee, and to countries such as Italy and Paraguay which spent little and appear to have fared as well as the United States, which spent 100 billion; the emerging consensus among computer experts was that the expenditure was worthwhile, beginning with the argument that a lack of preparation would have cost more](https://slate.com/news-and-politics/2000/01/were-the-y2k-preparations-in-vain.html)