The other half of the record
Expensive mistakes.
A list of winners quietly implies their judgment was sound throughout. It was not. These are documented, costly errors by the same people — kept to the same standard as everything else here: what happened is a fact with a citation, what it teaches is our interpretation and is labelled as one.
The record
Ordered as recorded. Each entry separates the documented event from the lesson we draw from it.
The Fire Phone
2014Amazon launched a smartphone at $199 on contract, built around 3D 'Dynamic Perspective' features few customers wanted. It was discontinued within roughly a year and cut to 99 cents.
Cost
$170M write-down, plus $83M of unsold handsets
In their own words
Bezos told the project's lead not to lose a minute of sleep over it, and later wrote that as a company grows, the size of its failed experiments has to grow too — 'if the size of your failures isn't growing, you're not going to be inventing at a size that can actually move the needle.'
The failure was building from Amazon's own strategic need — a direct channel to mobile customers — rather than from a customer need, which inverted the customer-obsession the company otherwise runs on. The salvage is the more interesting part: the hardware team and its learnings were redirected into Echo and Alexa. Treating a dead project as a trained team rather than a sunk cost is the transferable move.
Buying Berkshire Hathaway itself
1964Buffett bought into a failing textile company out of irritation. Its CEO had verbally agreed to tender his shares at $11.50, then sent an offer at $11.375. Buffett responded by buying control and firing him — then spent two decades tied to a dying textile business.
Cost
Buffett's own estimate: roughly $200bn in forgone value
In their own words
He has called it the dumbest stock he ever bought, and said the holding company would be worth about twice as much today had he put the same money into a good insurer instead.
The most expensive decision of a famously rational career was made in a moment of spite over an eighth of a point. Worth sitting with before assuming that the people in this dataset optimise coldly: the documented failure mode is emotional, and it cost more than any analytical error he has admitted to.
Over-automating the Model 3 line
2018Tesla built a heavily robotic assembly line for the Model 3, including a complex conveyor network. Robots handled flexible parts — trim, hoses — badly, and the line became the bottleneck. Tesla tore it out and put people back on those stations.
Cost
Months of 'production hell'; a near-existential cash crunch
In their own words
'Yes, excessive automation at Tesla was a mistake. To be precise, my mistake. Humans are underrated.' (April 2018)
First-principles reasoning has a failure mode, and this is it: the argument that a task *should* be automatable is not evidence that it can be, at cost, now. The reversal is the notable behaviour — he named the error publicly, attributed it to himself, and ripped out the capital already spent rather than defending it.
Missing mobile, then buying Nokia
2007–2015Ballmer publicly dismissed the iPhone's prospects on launch, citing its price and lack of a keyboard. Microsoft's mobile share collapsed over the following years. In 2013 it bought Nokia's handset business for roughly $7.2bn; in 2015, under Nadella, Microsoft wrote the unit off almost entirely and cut 7,800 jobs.
Cost
$7.6bn impairment — more than the purchase price
In their own words
Ballmer has acknowledged mobile as his greatest regret, saying the thing he would redo is the transition to hardware and devices. He has not disowned the Nokia purchase in the same terms.
Two distinct errors that are easy to merge. The first was reading a new product against the incumbent's criteria — price and keyboard — rather than against what customers were about to want. The second was trying to buy back a position with capital once the platform window had already closed. The dataset's own network-effects reasoning explains why the second could not work: you cannot purchase an ecosystem that has already chosen.
Telling Apple to shut down
1997Asked at a Gartner symposium what he would do about Apple, Dell said he would 'shut it down and give the money back to the shareholders.' Apple returned to profit the following year and became the most valuable company in the world.
Their response
Dell has said the answer was largely misconstrued — that he meant he was not a CEO for hire, not that Apple was worthless. He has declined to apologise for the call.
Never publicly conceded as an error — recorded here because the absence of an admission is itself part of the record.
Included because the interesting failure is a judgment error, not a financial one: the most successful direct-sales operator of his era could not see that the axis of competition was about to move from distribution efficiency to integrated design. Being excellent at the current game is a poor predictor of recognising the next one — and note that the public clarification reframes the quote rather than concedes the read.
'Move fast and break things', then unbreaking it
2009–2014Facebook's engineering culture explicitly traded stability for speed. In May 2014 Zuckerberg replaced the motto with 'move fast with stable infrastructure', saying the constant fixing of bugs was slowing the company more than caution would have.
In their own words
He framed the change as a practical trade rather than a moral one: shipping fast stopped being fast once the breakage had to be repaired.
A rare documented case of a founder reversing the cultural slogan that defined him, on operational grounds. The useful reading is that a maxim which works at one scale can become the constraint at the next — the same speed heuristic that got Facebook built was, by 2014, the thing slowing it down.
Left side is documented and sourced. Right side is our reading of what it teaches — the same fact/interpretation split used on every profile.
Read this section carefully
These are the mistakes that survived. Every error here was recoverable, which is precisely why the person who made it is still on a rich list. The instructive failures — the ones that ended companies and careers — belong to people this dataset will never contain.
There is a genre of business writing that treats a billionaire’s failure as proof that failure leads to success. It does not. The base rate of expensive mistakes among people who never recovered is far higher. What these episodes actually show is that a large enough balance sheet converts a catastrophic error into a survivable one.
Note also what the pattern is not: several of these are not bold experiments but ordinary lapses — spite, misreading a market, defending a prior position. That is the more useful observation.