S-81 Isaac Peral: The Decimal Point That Weighed 75 Tonnes

In May 2013 the Spanish state shipbuilder Navantia confirmed the worst kind of news for any submarine programme: the four boats of the S-80 class, already years into construction and among the most expensive defence projects in the country's history, were between 75 and 100 tonnes heavier than their design allowed. A submarine is unlike almost any other vehicle in one respect that made this fatal rather than inconvenient. Its buoyancy is fixed by the volume of water its hull pushes aside, and that volume was welded shut years earlier. If the weight of the boat exceeds the weight of the water it can displace, the boat does not float lower, the way an overloaded ship does. It does not float at all. Engineers described a vessel that could dive but might never come back up — a submarine that would behave, in the words of one report, like a stone.

~17 min read  ·  Published October 4, 2026

By EnginStack Editorial Team | Checked against Navantia and Spanish Ministry of Defence statements, the U.S. Naval Institute's programme record and contemporaneous Associated Press reporting About →

This is the seventh accident in this series, and the first in which nobody misread a unit. At Gimli the number 22,300 was correct and the unit behind it was wrong. On the Mars Climate Orbiter the impulse values were arithmetically perfect and silently denominated in pound-force-seconds. In the Dhahran Patriot case the units were right on both sides of the calculation and the failure lived in a register that could not hold a tenth of a second. The Spanish submarine programme that produced the S-80 class is a separate family again. Nobody wrote metres where feet were required. Nobody read pounds as kilograms. The tonnes, the metres and the cubic metres are all exactly what they claim to be. What failed was the size of a number.

That distinction is what makes this the most transferable case in the series, and the most uncomfortable. Every other accident here can be reduced to a labelling problem: the quantity was right, the label was missing or wrong, and a machine that could not read the label acted on the wrong meaning. A decimal point in the wrong place is not a labelling problem. It is the same arithmetic operation that sits inside every unit conversion on this site, performed with the wrong power of ten. Converting kilograms to pounds is multiplying by 2.2046. Converting metres to feet is multiplying by 3.2808. A misplaced decimal point is a multiplication by ten, or by a hundred, or by a tenth, where none was intended and nobody noticed. It is not a mistake in the unit at all. It is a mistake in the coefficient, and the coefficient is where all the danger of a conversion actually lives.

Quick answer

Spain ordered four S-80 class submarines in 2003 for about €1.76 billion. In May 2013, with the first boat five years into construction, Navantia disclosed that the design was between 75 and 100 tonnes too heavy, and that as built the boats might not be able to surface reliably. The most widely reported cause came from a former Spanish defence official, who said someone had put a decimal point in the wrong place in a weight calculation and that nobody had reviewed the arithmetic — a claim Navantia never confirmed; the company formally acknowledged only “deviations related to the weight balance”. Whatever the precise origin, the fix was a ten-metre hull extension at €7.5 million per metre, a larger pressure hull, and €16 million to enlarge the Cartagena dock that the longer boat no longer fitted. The programme cost rose from €1.76 billion to €3.94 billion and the first boat, S-81 Isaac Peral, entered service on 30 November 2023 — twelve years after the date originally planned.

Failure A weight value in the design calculation carrying the wrong power of ten — reported as a misplaced decimal point — against a buoyancy allowance fixed by the hull
Quantity Light-ship weight or a component within the weight budget
Overweight 75 to 100 tonnes, on a submerged displacement of about 2,430 tonnes — roughly 3 to 4% by mass
Consequence Insufficient buoyancy to surface reliably — the boats could dive but might not rise
Fix Hull lengthened 10 m (71 m to 81.05 m), displacement raised from about 2,200 to 2,695 tonnes surfaced
Fix cost €7.5 million per metre of hull; €14 million to the U.S. consultancy Electric Boat; €16 million to enlarge the dock
Programme total €1,756 million (2004) → €2,212 million (2010) → €3,935 million (2018, four boats)
Delivery Planned 2011; S-81 Isaac Peral commissioned 30 November 2023

The Number That Was Ten Times Too Big

Start with what a misplaced decimal point actually does, because the phrase is so familiar that its arithmetic gets lost. Moving a decimal point one place to the right multiplies a value by ten; one place to the left divides it by ten. There is no other effect. The digits do not change, the unit does not change and the sign does not change. Eight point three becomes eighty-three. Eighty-three becomes eight hundred and thirty. The number still looks like an ordinary engineering number. It still has plausible magnitude. It still passes the eye of anyone reading the spreadsheet row on its own.

This is why the error is so much harder to catch than a unit confusion, and the reason belongs to the structure of the two mistakes rather than to carelessness. A unit mistake leaves a visible absence: a field that should have read kilograms and reads nothing, or a column header that says pound-force where the specification said newton. You can hunt for unit errors by hunting for missing labels, and the whole discipline of unit checking exists because that hunt is tractable. A power-of-ten error leaves nothing to hunt for. The label is present. The value is present. The only thing that is wrong is how large the value is, and size is a property that a number does not carry on its face.

There is a constraint that follows, and it is the most useful single sentence in this article. If the total overweight of the S-80 design was between 75 and 100 tonnes, and if that discrepancy came from a single item multiplied by ten, then the item in question was about 8.3 to 11.1 tonnes — because ten times a weight minus the original weight is nine times the original, and 75 ÷ 9 ≈ 8.3 while 100 ÷ 9 ≈ 11.1. A generator set, a length of main ballast piping, a battery section, a block of structural steel: any of these sits comfortably in that range on a submarine of this size. That is the kind of number that disappears into a weight budget of two thousand tonnes and reappears four years later as an emergency.

A misplaced decimal point is a multiplication by ten the digits never change, and neither does the unit — only the size does 8.3 t the intended value a plausible component ×10 83 t what the design carried +74.7 t of error ×10 830 t implausible and obvious Why the middle box survives a design review 8.3 t and 83 t are both believable weights for submarine equipment. Only one is right, and nothing on the line itself tells you which. The value that betrays itself — 830 t is the third shift, the one nobody made. A 75 to 100 t total discrepancy is arithmetically consistent with one ×10 slip on an 8–11 t item.

It is worth being explicit about the difference between this and the mistake everyone assumes it is. When people say a project failed on a decimal point, they usually picture a single keystroke — a slip of the thumb, detectable by anyone who looks. That is not the failure mode here, and treating it as one leads to the wrong defence. The value 83 in a weight table is not detectable by looking. It is detectable only by knowing, independently, that the item weighs about 8.3. A review that reads each line and asks “is this plausible?” answers yes, because 83 tonnes of machinery is entirely plausible on a submarine. The review has to ask a question the table does not prompt: not is this number possible, but is this number the right one — and that requires a second, independent estimate of every weight on the list.

Why a Submarine Is the Wrong Place to Lose 75 Tonnes

The reason a 3 to 4% weight error is an emergency rather than a footnote is a piece of physics that submarines share with almost nothing else. Buoyancy is not a number that can be adjusted. It is a volume. A hull displaces a quantity of seawater determined by its own shape and size, and the upward force that keeps the boat afloat is that volume multiplied by the density of the water. When the pressure hull was welded shut, that volume was fixed. It does not care how much the contents weigh, and it cannot grow to accommodate them.

Weight is the opposite kind of quantity. It accumulates. Every drawing adds something — a pipe run, a console, a battery, a pump, a late-requested modification — and every addition is individually small and individually justifiable. There is no point in the process at which the sum announces itself as dangerous, because the sum is not compared, continuously and in one place, against the ceiling. A surface ship tolerates this by sitting lower in the water: load it further and it simply takes on more draught. A submarine has no such slack, because it must be able to hold depth without touching the bottom or the surface, and it must be able to come up. Its margin is the reserve of buoyancy it can command by blowing water out of its ballast tanks, and on a conventional submarine that reserve is on the order of ten to fifteen per cent of the displacement.

Set 75 to 100 tonnes against that margin and the numbers finally produce the right reaction. Against the full submerged displacement of about 2,430 tonnes, the overweight is 3 to 4% — small enough to sound harmless in a meeting. Against the reserve buoyancy that actually determines whether the boat can surface, it is worse than the ratio suggests, because the missing tonnes come out of a reserve that was small to begin with. That is the whole accident in one sentence: the error was a small fraction of a large number, subtracted from a much smaller one.

Weight accumulates; buoyancy does not the ceiling is fixed by the welded hull — every new tonne has to fit underneath it BUOYANCY CEILING — set by hull volume, cannot move design 2,200 t + equipment + AIP & growth + the error as built A ship sits lower when it is heavy. A submarine whose weight crosses the ceiling does not sit lower — it stops being able to rise.

Now the design decision that made the error irreversible becomes visible. To add buoyancy to a submarine you must add volume, and the only practical way to add volume to a boat that is already building is to make it longer. Lengthening the hull adds a cylinder of displaced water for every metre of new length, and that cylinder is the only currency in which the error could be paid. The Spanish programme's answer, developed with the U.S. submarine builder Electric Boat, was a ten-metre extension — the single most expensive metres of hull in the programme, at €7.5 million each. It worked, and it was applied to all four boats. It also closed the loop on the accident's cost: the error was cheap to make and enormously expensive to undo, which is the signature of a mistake that survived review.

What the Record Actually Says

Here the article has to slow down, because the phrase at the heart of this case is better documented than the thing it describes. The decimal-point explanation is the one that travels. It comes from Rafael Bardají, a former director of the Strategic Assessment Office at the Spanish Ministry of Defence, who told the Associated Press in 2013 that someone had put a decimal point in the wrong place in the calculations and that “nobody paid attention to review the calculations”, calling it “a fatal error”. It is a clean story: one keystroke, one missing check, one catastrophe.

What Navantia actually said is narrower and, for anyone trying to learn from the case, more interesting. The company acknowledged “deviations related to the weight balance” — a statement about the effect, not the cause. It never confirmed a decimal point, and no public investigation report of the kind that produces a numbered finding on a numbered page has established one. Contemporaneous technical coverage offered an alternative explanation that has nothing to do with a slip of the thumb: that the boats had grown heavier through the accumulation of new technology, in particular the air-independent propulsion system and other equipment added after the original weight calculation was closed. On that reading the programme did not suffer a typo. It suffered the same slow, unmanaged growth of weight against a fixed ceiling that sinks ordinary projects, at a sensitivity — buoyancy as a difference — that turns a normal margin erosion into a life-safety event.

The arithmetic then does something useful: it lets you test the simple story against the size of the fix. One hundred tonnes of extra weight requires about 2.3 metres of additional ten-metre-diameter-level hull to offset, by a first-order estimate — 100 tonnes of seawater occupies about 97.6 cubic metres, and the S-80's pressure hull, roughly 7.3 metres in diameter, gains about 41.9 square metres of cross-section per metre of length, so 97.6 ÷ 41.9 ≈ 2.3. The boats were lengthened by ten metres. That is more than four times what the overweight alone demanded, and the extra volume is not necessarily waste: the extension also had to carry its own structure and ballast, restore a positive reserve, and accommodate the programme's other weight growth. But the size of the gap between a 2.3-metre requirement and a 10-metre fix is a piece of evidence in its own right. It says that a single misplaced decimal point, on its own, is unlikely to be the whole story, and that the honest description of the S-80 overweight is closer to a weight budget that lost control against a ceiling that could not move than to one keystroke.

That conclusion is less quotable than the decimal point, and more useful. The decimal point is a story about a person. The weight budget is a story about a process, and processes are the part you can actually change.

The Fix That Cost More Than the Error

The repair was neither quick nor clean, and its own timeline is part of the lesson. Navantia brought in General Dynamics Electric Boat in May 2013 on a technical-assistance contract reported at around €14 million over three years. The consultancy's recommendation was to add length, and by late 2014 the redesigned boats had been stretched by ten metres: from an original length of 71 metres to 81.05 metres, about 233 to 266 feet. Displacement rose with them — from roughly 2,200 tonnes surfaced and 2,430 tonnes submerged in the original design to 2,695 tonnes surfaced and 2,965 tonnes submerged as built, so that the class now sits in a different weight bracket than the one its original weight budget was written for.

Then came the discovery that turned an engineering problem into a running joke about Spanish defence procurement. The lengthened boats no longer fitted the dock at Cartagena, the naval base where they had been built. The basin had been dimensioned around submarines of up to about 78 metres; the S-80 Plus was now over 81. Modifying the dock added a further €16 million to a bill that had not been caused by a single welding defect but by a number nobody re-derived. The class's timelines drifted the same way. The first boat had been expected in 2011. Isaac Peral was named in April 2021 and launched on 7 May 2021, began sea trials in 2022, completed a first static dive in March 2023 and a full test-depth dive in October 2023, and was commissioned into the Spanish Navy on 30 November 2023 — twelve years late.

The money is the least surprising part of the story and the clearest measure of it. The four-boat contract was signed in 2004 for €1,756 million. By 2010 that had risen to €2,212 million, before the overweight was even public. By 2018, for four redesigned boats, the programme figure stood at €3,935 million — close to €984 million per submarine, a unit cost in the neighbourhood of a nuclear boat rather than the diesel-electric one that was ordered. None of that money was spent to correct the arithmetic. It was spent because arithmetic corrected too late has to be paid in steel, metres and years.

The Check That Would Have Caught It

Because a power-of-ten error cannot be seen by reading the number that contains it, the defence against it cannot be reading. It has to be an independent estimate of magnitude, which is what engineers call an order-of-magnitude check. The check does not ask whether the value 83 is correct; it asks whether the item should weigh about eight tonnes or about eighty, using a source other than the number under review. On a submarine weight budget that is a tractable question, because most components belong to families with known size ranges: a diesel generator, a battery, a length of pipe, a console. Line something up against the family it comes from and a tenfold error stands out immediately, because a component that is ten times heavier than everything in its class is not a subtle correction — it is a different object.

Two more habits fall directly out of this case, and neither requires new software. The first is a sum invariant: a weight budget is by construction a total made of parts, so the total must equal the sum of the parts, and the total must sit inside the platform's ceiling. A single item inflated tenfold breaks both relations, and a one-line reconciliation of total against ceiling — run at every design review, not once at the start — would have rung like a bell. The second is a difference-sensitivity rule. When the quantity that matters is a difference between two large numbers, as buoyancy is, or the interval between two timestamps as in the Dhahran case, an absolute error is the wrong thing to track. What matters is the error measured as a fraction of the margin, not of the total. Three per cent of a 2,400-tonne submarine sounds like a rounding note. Three per cent of a reserve that decides whether the boat can surface is a different conversation, and it is the only one that should have been happening.

There is a final, structural point that generalises past submarines. The S-80 programme is an example of a system whose capacity is fixed and whose demand is not: the ceiling was set by geometry, and the weight grew by addition, review after review, with no mechanism forcing the two to meet until construction had made the meeting inevitable. The same shape appears wherever an unbounded accumulation runs against a bounded allowance — a memory limit filled by features, a weight-and-balance envelope filled by options, a schedule slack consumed by small delays, a 32-bit second counter filled by the passage of time. The unit is different in each case. The mechanism is not, and neither is the reason it goes unnoticed: nothing in the day-to-day decisions announces the sum.

What This Case Teaches That the Other Six Cannot

The six earlier accidents in this series all teach the same family of lessons: label every number, check every interface, convert in both directions, treat a metrication transition as a safety event. Those lessons are about units, and units are the part of a conversion you can see. The S-80 teaches the part you cannot.

A wrong unit leaves a hole; a wrong power of ten leaves nothing. This is the deepest point in the case and the reason it deserves to sit alongside a missing label and a truncated register. A unit error can be hunted by looking for what is absent — the field with no unit, the column with the wrong header, the interface with no unit field at all. A magnitude error cannot be hunted that way, because everything is present and everything is plausible. The only way to find it is to have an independent notion of how big the number should be, and to compare. If your verification process never forms that independent estimate, it is not checking magnitude, no matter how carefully it reads.

Know which of your quantities is a difference. Buoyancy, tolerance, margin, headroom, budget remaining, error interval, time to deadline — these are all subtractions, and in a subtraction a small absolute error becomes a large relative one. Identify them explicitly, and review them with a tighter standard than the quantities around them. The S-80 programme reviewed the boat's weight, which was a very large number that tolerated a three per cent error. It needed to review the boat's reserve, which was a much smaller number that did not.

A fix that costs more than the error is the normal case, not the exception. The overweight was created by arithmetic that would have taken minutes to re-derive. It was removed by ten metres of pressure hull at €7.5 million a metre, a €14 million consultancy contract and a €16 million dock. That asymmetry — cheap to make, expensive to undo — is precisely why order-of-magnitude checks pay for themselves even though they almost always find nothing. A check that fails once in a hundred reviews and saves a submarine programme is not overhead. It is the cheapest thing on the drawing board.

And a last note in the spirit of this series, which is to prefer the honest version of a story over the satisfying one. The decimal point makes a better headline than a weight budget that drifted, and it may well be true. But the record does not establish it, and the arithmetic of the fix suggests the truth is broader. If you take only one thing from the S-80, let it be the version that survives scrutiny: a number can be entirely correct in its digits, its unit and its sign, and still be wrong in the only way that matters — its size. For the units themselves, and the exact constants that keep a conversion honest, the tools below are the companion to this article.

Related Converters

Every number in this accident is a conversion between tonnes, metres and cubic metres — the three units that set a submarine's fate. These are the tools for working them out precisely:

Related Guides

Frequently Asked Questions

What went wrong with the Spanish S-80 submarines?

The four boats of the S-80 class were designed and built about 75 to 100 tonnes heavier than their weight budget allowed. Navantia disclosed the problem in May 2013, when the first boat was already five years into construction. Because a submarine's buoyancy is fixed by the volume of water its hull displaces, the extra weight could not simply be carried: the boats risked being unable to surface reliably. The fix was to lengthen the hull by ten metres to create additional buoyant volume, at a reported cost of 7.5 million euros per metre, and to enlarge the dock at Cartagena that the longer boats no longer fitted.

Was the S-80 problem really caused by a misplaced decimal point?

That is the most widely reported explanation, but it is not one Navantia confirmed. It came from Rafael Bardaji, a former director of the Strategic Assessment Office at the Spanish Ministry of Defence, who told the Associated Press in 2013 that someone had put a decimal point in the wrong place and that nobody had reviewed the calculations. Navantia's own statement acknowledged only deviations related to the weight balance, without naming a cause, and other contemporaneous coverage attributed the growth to new equipment and the air-independent propulsion system rather than to a single keystroke. The size of the repair, which lengthened the boats four times more than the overweight alone would require, suggests the truth is broader than one decimal point.

Why does 75 tonnes matter on a submarine that displaces 2,400 tonnes?

Because the quantity that decides whether a submarine can surface is not its total displacement but its reserve of buoyancy, which on a conventional submarine is only about 10 to 15 per cent of that displacement. An overweight of 75 to 100 tonnes is only 3 to 4 per cent of the boat, but it is taken out of a much smaller margin, so it consumes a large share of the reserve. A surface ship can absorb extra weight by sitting lower in the water; a submarine cannot, because the volume of water it displaces is fixed by its hull and the boat must be able to hold depth and rise again.

How much did the overweight problem cost?

The direct repair costs included 7.5 million euros per metre for the ten-metre hull extension, roughly 14 million euros for technical assistance from the U.S. shipbuilder General Dynamics Electric Boat, and around 16 million euros to enlarge the Cartagena dock to take the longer hull. The programme total rose from 1,756 million euros for four boats at contract signature in 2004, to 2,212 million by 2010, to about 3,935 million by 2018, a unit cost near 984 million euros, and the first boat entered service on 30 November 2023, about twelve years later than originally planned.

How did Navantia fix the S-80 submarines?

By adding buoyant volume, which for a submarine means adding length. Working with General Dynamics Electric Boat, Navantia lengthened each hull by ten metres, from about 71 metres to 81.05 metres, and added a pressure hull ring. Surfaced displacement rose from roughly 2,200 tonnes to 2,695 tonnes and submerged displacement from about 2,430 tonnes to 2,965 tonnes. The redesign solved the buoyancy problem but increased the cost and the delay, and created a further problem: the finished boats exceeded the 78-metre limit of the dock at Cartagena, which then had to be modified.

What is an order-of-magnitude check?

An order-of-magnitude check estimates roughly how large a value should be, using a source independent of the value itself, and then asks whether the number in front of you is consistent with that estimate. It does not verify the exact figure, only its size, which is exactly what a misplaced decimal point corrupts. On a submarine weight budget the check asks whether a component should weigh about eight tonnes or about eighty, using the known size range of comparable components. Because a tenfold error still produces a plausible number, only an independent estimate of magnitude can expose it, which is why reading each line of a weight table for plausibility can never substitute for re-deriving it.

Sources

Primary sources

Secondary references