He said there must be an allowance for failure when you try something new.
"There must also be an allowance for failure because if you are trying something new necessarily that there's some chance it will not work."
Why are your timelines so often late?
Q844 answers2014-2026
Short answer
He says his prediction batting average is very high, with a 'Cassandra effect', though timing may be off.
All 44 answers, oldest to newest
2014 · 1 answer
2015 · 1 answer
He said weekly production rate had grown 100% year over year and Tesla would try to double it again from start to end of the year.
"we're going to try to increase It Again by 100% year-over-year in terms of the the weekly production rate"
2016 · 1 answer
He set a goal of a full-autonomy drive from LA to Times Square by the end of next year without a single touch.
"dropping you off in Times Square in New York and having the car go and park itself by the end of next year without the need for a single touch"
2018 · 4 answers
He said anyone who closed the design has a mental checklist scrolling of all the ways it can fail — thousands of things can go wrong and everything has to go right.
"all the things that can can can break"
He said they tried to cancel the Falcon Heavy program three times at SpaceX because it was way harder than they thought. The initial idea was just sticking two first stages on as side boosters.
"we try to cancel the Falcon Heavy program three times at SpaceX"
He felt optimistic about the timeline but said people have told him his timelines historically have been optimistic, so he was trying to recalibrate.
"My timelines historically have been optimistic"
Model X was an exercise in hubris — they got carried away with the art and it probably delayed Model 3 a little.
"it probably delayed us a little bit with the model 3 as well"
2019 · 7 answers
He said with exponentials, small changes in the calendar breakpoint make enormous percentage differences.
"But when making estimates against an exponential, small changes in the calendar breakpoint have enormous percentage differences."
Explains the middle of the manufacturing S-curve is extremely difficult to predict.
"Extremely difficult to predict middle part of manufacturing S-curve."
He thought very quickly, maybe even toward the end of 2019 and he'd be shocked if not 2020 at the latest, having a human intervene would decrease safety, like elevator operators.
"I'd be shocked if it's not next year at the latest"
He said that next year, for sure, there would be over a million robotaxis on the road.
"next year for sure, we will have over a million robotaxis on the road"
He said if you are going to try something innovative you are in unexplored territory, so the odds that something will go wrong are pretty high. It is only if you try something well understood that there is little chance of failure — but then it will not be interesting. Uncharted territory will result in failures.
"uncharted territory will result in failures"
He acknowledged people were puzzled by his million robotaxis by end of next year prediction.
"A lot of people were puzzled as to, how can I say that we would have like, you know, a million robotaxis by the end of next year."
He said if a design is taking too long, the design is wrong, and therefore the design must be modified to accelerate progress. One of the most fundamental errors in advanced development is sticking to a complicated design and not striving to delete parts and processes.
"if a design is taking too long, the design is wrong"
2020 · 10 answers
He says production almost always follows a herky-jerky S-curve.
"Production pretty much always follows this S curve or it's kind of like a herky jerky S curve."
He says the intermediate portion of the production S-curve is very difficult to predict.
"But that intermediate portion of the S curve is very difficult to predict."
He says he gives the 50th-percentile estimate, so about half his predictions will be wrong.
"If this is normal distribution, I give you the 50th percentile, not the three sigma optimistic or pessimistic. So then that necessarily means at least half my predictions will be wrong and half will be right."
He says absent a massive force majeure event, 40% growth is the likely number.
"I think 40% is the likely number."
He says his estimates are the likely midpoint, not a figure with lots of margin.
"when I give a guess, I give the guess that I think is the likely midpoint, not the point with lots of margin."
He says his predictions may come true late, but they do come true.
"It may have come true late, but it did come true."
He is optimistic in general, and that extends to schedules. He would wish they could do it in two years, which might mean four; he would be surprised if it took more than four.
"I would wish that we could do it in two years I know that might mean for"
He says manufacturing capacity is underestimated at first and later sometimes overestimated, because ramps follow an S-curve.
"generally what I see is the manufacturing capacity is underestimated in the beginning for quite some time. Then it is sometimes overestimated because this is an S curve. It goes exponential to linear to logarithmic."
He says manufacturing follows the S-curve, which people who have not spent time in manufacturing often miss.
"Just in general, manufacturing follows the S curve."
He says new factories typically take 12 to 18 months to reach capacity.
"it will typically take about 12 to 18 months to reach capacity"
2022 · 2 answers
Internally they set the most aggressive timeline they can, because schedules expand to fill the time; it is rare to finish early.
"the most aggressive timeline that we can"
He said self-driving has a lot of false stones: you think you are getting there and then you asymptote.
"I'd say that self-driving is one of those things where"
2023 · 8 answers
He says Tesla guided 1.8 million vehicles instead of its roughly 2 million potential because something unexpected always happens.
"we're saying 1.8 because I don't know, there just always seems to be some freaking force majeure thing that happens somewhere on earth."
He says it is easy to predict the long term but hard to predict the timing in between.
"It's like, I think easy to predict the long term, but hard to predict the"
He says he always downplays the start of production.
"I always like try to downplay the start of production"
He admits he has been over-optimistic, because progress follows a log curve: fast early gains that look linear if you extrapolate them.
"Obviously in the past I've been overly optimistic about this. The reason I've been overly is that the progress tends to sort of look like a log, a log curve"
He notes a 50% compound growth rate cannot continue forever.
"it's not possible to have a compound growth rate of 50% forever or you will exceed the mass of the known universe."
He guesses the Cybertruck output rate will be reached sometime in 2025, not next year.
"I don't think we're going to reach that output rate next year. I think we probably reach it sometime in 2025."
He's naturally, even pathologically, optimistic about time scales; late but always delivers.
"I'm like naturally optimistic about time scales"
He agrees he is pathologically optimistic on schedule.
"I'm pathologically optimistic on schedule"
2024 · 4 answers
From a SpaceX standpoint, not counting regulatory, he said they were ready probably in less than 30 days; with bigger changes, no later than 60 days.
"not counting out regulatory, we're ready probably in less than 30 days"
He admits he's sometimes a little optimistic with schedules but wouldn't do what he does otherwise.
"I realize sometimes I'm perhaps a little optimistic with schedules"
He said for my time-based predictions, I generally aim for the 50% percentile date, which means that half my predictions will be late and half will be early. The press never mentions the predictions that were early, so it see...
"For my time-based predictions, I generally aim for the 50% percentile date, which means that half my predictions will be late and half will be early. The press never mentions the predictions that were early, so it seems like I’m always late. However, it is very rare that a prediction I make does not come true over time."
He says non-autonomous gasoline cars will be like riding a horse.
"non autonomous gasoline vehicles in the future will be like riding a horse"
2025 · 4 answers
He says Optimus timing is hard to predict because Tesla is designing the train, the station and the tracks at the same time.
"For Optimus, we are designing the train and the station and in real time while also building the tracks."
He says the media reports when he is late but never when he is early.
"the media reports on when I'm late, but never reports when I'm early."
He says even the biggest horse advocate could not stop the advent of the automobile.
"you can't stop the advent of the automobile."
He said his schedule estimates are 50th percentile so he should be early half the time, that news only reports when he is late, and that SpaceX will do ~90% of Earth payload to orbit this year.
"My estimates for schedules are at the 50th percentile, which means I should be early half the time and late half the time. Anyone dumb enough to think the news is real will believe I am always late, because the “news” only reports when I am late and never when I am early. What matters is looking at the actual progress of SpaceX, which will do ~90% of all Earth payload to orbit this year, or Tesla, which is worth more than the rest of the auto industry on Earth combined. That cuts through the bullshit."
2026 · 2 answers
He aims for a deadline he thinks is at the 50th percentile, so it'll be late half the time.
"it's the most aggressive deadline I can think of that could be achieved with 50% probability"
He says his prediction batting average is very high, with a 'Cassandra effect', though timing may be off.
"I'm very good at predicting the future"