Bloomberg's Inside Look At Tesla's Model 3 Factory (bloomberg.com)

An anonymous reader shares an excerpt from an exclusive inside look at Tesla's Model 3 factory in Fremont, California: On the Model 3 body line on a Tuesday afternoon in early June, everything is still. Tesla is just coming off a week of downtime during which workers added a new production line, improved ventilation after a fire in the paint shop, and overhauled machines across the factory. But even after the changes, there are kinks to work out. Suddenly, dozens of robots snap into frenzied action, picking up door panels, welding window pillars, taking measurements, and on and on. This robotic dance is a visceral representation of what Tesla chief executive Elon Musk has dubbed "Alien Dreadnought," a code name for the factory that evokes an early 20th century warship, but with extraterrestrials.

The stakes couldn't be higher for Tesla, which is sprinting to produce the Model 3 in quantities great enough to turn a profit. But so far, the plant's choreography has been choppy. The flow at the factory in Fremont, California, is constantly interrupted while robots and humans are trained, retrained, or swapped out. If Tesla can't make this dance work, it will be remembered as a lesson in the dangers of irrational exuberance for automation. Success, on the other hand, could transform the car industry.

Comments

I think they'll succeed even if they fail

[Kjella]

Take Amazon, it had the most ridiculous negative P/E ever. At the height of the dotcom boom I got a free $10 that I used to buy a $4 book with $6 shipping (international). How the fuckity fuck fuck to you make money on that? By staying the course with enough VC money that eventually you'll surface. That's kinda what I'm thinking about driverless cars at the moment, even if it doesn't make sense... even if they never recover their investment... there's so much money and so many corporations behind it that it'll happen. When I'm a senior citizen in 25+ years I'll have a self driving car. I'm probably not going to Mars, but that's okay. And the singularity is not happening. Immortality certainly not. You have a life, enjoy it.

Re:I think they'll succeed even if they fail

[Pinky's Brain]

Amazon stands on top of its market, regardless if EV and self driving cars take off Tesla will only have a small part of the pie ... Amazon's money was better spent.

Re: I think they'll succeed even if they fail

[Bing Tsher E]

'Driverless cars' is a ridiculous moving target. Maybe Tesla should stick to their original mission: electric cars with steering wheels.

Re:I think they'll succeed even if they fail

[nojayuk]

Take Amazon, it had the most ridiculous negative P/E ever.

Amazon was growing out a lot of infrastructure, expanding their customer base and developing AWS among other things as it ran astounding deficits for a decade or more. The people lending them money could see where it was going, there was a roadmap to profitability. They could have failed but they're a success story now. Tesla has a single factory, a lot of debt and a track record of not meeting the Boss's production promises quarter on quarter.

If Musk is indeed overnighting at the plant, looking over everyone's shoulders while they try and fix what's wrong then they've got even more problems. Imagine coding under pressure and then the CEO sets up camp in your office...

Re:I think they'll succeed even if they fail

[sfcat]

Amazon stands on top of its market, regardless if EV and self driving cars take off Tesla will only have a small part of the pie ... Amazon's money was better spent.

Keep telling yourself that. Where are all those batteries for those other car companies EVs coming from again? You can't just put a bunch of laptop batteries into an EV and sell them for a profit (anymore). You need to use a type of battery cell with a cost below a certain price point (about $150 per Kwh I believe) and nobody else currently is doing that. Hell, there aren't even plans to make a non-Tesla factory to build such batteries. Telsa will have at least a 5 year head start on the rest of the car companies and by that time, it will already be over. GM loses money on every Volt and Bolt sold and that won't change unless they start making their own batteries at scale. GM makes a small number of EV batteries at very high cost currently and GM has no current plans to scale up to the point where they can make those cars economically.

Also, you are betting that people want to buy cars from car dealerships (really, never bought a new car have you).

Finally, you are betting that car companies will entirely change their basic culture and start making EVs. Go to any car company's HQ and talk to the employees and the real problem will come into view. These folks simply don't want to change. And by the time they figure out how completely fucked they are, it will be entirely too late. It happened with every other type of business when a new business model becomes viable. The old guard gets crushed. It happened with Walmart vs Amazon. It happened with MS vs Google. It happened dozens of times before that in many industries including many industrial businesses. It will happen again, its one of the few patterns in life that happens with such high certainty.

Re:I think they'll succeed even if they fail

[EnsilZah]

Tesla in fact has three factories, for whatever that's worth.
And I'm sure the people lending them money see where they're going as well, mass market electric cars, trucks, a global charging network, self driving, residential solar, residential and commercial energy storage.
There is a roadmap for profitability around Q3-Q4 of this year.

And sure Musk sets optimistic and ambitious goals timelines that he usually doesn't achieve, but the goals are eventually achieved.
And while the ramp up for Model 3 production is slower than expected, the news I see from other car companies is 'We'll have our Tesla-Killer any moment now, in 2020', 'We're starting mass production of battery cells... in 2024', 'By 2028 half the cars we produce will be electric...'

Re:I think they'll succeed even if they fail

[michelcolman]

The people lending them money could see where it was going, there was a roadmap to profitability.

Tesla has a road to profitability too. They expect to actually be profitable in the third and fourth quarter (before they start sinking money into Model Y development and go negative again, probably). Same story with Model S and X: each time, they had one or two profitable quarters before they started pouring money into the next model. Model 3 should have a very healthy profit margin, and it looks like Elon is extremely confident they are going to crush expectations in only a couple of months from now.

Imagine coding under pressure and then the CEO sets up camp in your office...

If that CEO was a pointy-haired boss, that would be bad. But if the CEO spends 80% of his time doing actual engineering and helping you solve the problems you encounter instead of telling you to work harder, I imagine it actually creates the kind of atmosphere where everyone wants to work as hard as they can without the boss even asking. Makes them feel they are working with him rather than for him.

Re:I think they'll succeed even if they fail

[michelcolman]

And in two years we'll have a car with almost the range and almost the power of today's Teslas, which of course won't improve in the mean time.

Re:I think they'll succeed even if they fail

[drinkypoo]

GM loses money on every Volt and Bolt sold and that won't change unless they start making their own batteries at scale. GM makes a small number of EV batteries at very high cost currently and GM has no current plans to scale up to the point where they can make those cars economically.

GM is engaged in bet-hedging. They appear to believe that MIC interest in hydrogen-powered vehicles is going to make them a viable option going forward. If people buy H2 vehicles, they don't have to make good batteries. They can just buy small ones from someone else (H2 vehicles need a battery, too.)

Re:I think they'll succeed even if they fail

[Hognoxious]

Tesla has a road to profitability too.

Better hope there's not a fire truck on it.

Re:I think they'll succeed even if they fail

[im_thatoneguy]

Imagine coding under pressure and then the CEO sets up camp in your office...

If he's committing bug fixes and answering questions immediately great. The worst is when you're trying to fix something but you can't get a decision made for 8 hours.

Re:I think they'll succeed even if they fail

[torkus]

Except H2 tech is at least a generation behind battery EVs, probably 2 or 3.

The distribution network? Equally so. Plus H2 just outright sucks to work with. Electricity? We've got a pretty robust distribution network there (yah, has it's flaws, but EVs can actually help balance that)

As far as I'm aware, there's never been a single production H2-based vehicle produced and sold for anything even remotely near breakeven price.

Re:I think they'll succeed even if they fail

[drinkypoo]

Except H2 tech is at least a generation behind battery EVs, probably 2 or 3.

I'd say two. GM, Honda, and Toyota are all headed for the next generation in their next product cycle, though.

The distribution network? Equally so. Plus H2 just outright sucks to work with. Electricity? We've got a pretty robust distribution network there (yah, has it's flaws, but EVs can actually help balance that)

Well, let's talk about those flaws. Most long-distance links are still poorly monitored, and the grid is not exactly suffering from a surplus of capacity, but you can have more hydrogen as quickly as you can cram it into trucks or train cars.

As far as I'm aware, there's never been a single production H2-based vehicle produced and sold for anything even remotely near breakeven price.

Nope. The claim is that the next generation will be profitable. That's why GM is still sitting it out, and focusing on military applications which will be profitable much sooner. Still, Californians can lease a Honda Clarity right now.

Bloomberg? Why?

[whoever57]

Bloomberg has published a slew of negative stories about Tesla, so why give them an exclusive like this? Is it some attempt to get Bloomberg to change its obvious stance?

To support my assertion of Bloomberg's negative stance, I point readers to Bloomberg's Model 3 tracker, which has been and remains consistently low in its estimates, without acknowledging that its model is producing wrong numbers. Musk announced that Tesla is producing 3500 cars per week, while Bloomberg's estimate is 2560.

Re:Bloomberg? Why?

[AlanObject]

Is it some attempt to get Bloomberg to change its obvious stance?

That would be my guess and if I were running PR for Tesla that is what I would do.

My housekeeper's husband works at the plant (we live about 3 miles away) and he tells me it has become pretty regular for Musk to be spending overnight at the factory. Whatever else you might say about him he seems to be taking it seriously.

Re:Bloomberg? Why?

[crow]

Tesla says it hit 500 in one day, not that it was sustaining that, so the Bloomberg analysis is not inconsistent. Bloomberg also says that their model lags behind sudden changes. Overall, I think their tracker is fair, given the caveats they include.

Re:Bloomberg? Why?

[Rei]

The exact quote is: "Despite our production difficulties, all parts of the Model 3 production system have demonstrated a 500 car per day capability, or, a 3500 car per week capability, and we just did a big set of upgrades and are spooling up the production lines again. I think it's quite likely that we will achieve a 5000 cars a week by the end of the month."

This was with 2 GA lines (the biggest bottleneck). They just built a third.

I don't think there's any malice in the Bloomberg tracker. I think it's just hindered by too long of an averaging period on their algorithm. It averages in planned line downtimes in with line production rates, which you don't want.

Re:Bloomberg? Why?

[Rei]

You're confused. The "pulling people off shifts" was first off due to planned S/X downtime, but more importantly, that was not this quarter. That was at the end of last quarter, and to achieve 2020 vehicles per week, not 3500. And more to the point, they not only maintained that rate for several weeks after the other lines went back up (all the way up until the Model 3 lines' next scheduled downtime), but increased it.

Re:Bloomberg? Why?

[LynnwoodRooster]

Demonstrating a capability and actually USING the capability are two different things. It's like looking at the nameplate capacity of, say Ivanpah Solar Power Facility as 392 MW, and then checking and finding the best it's ever done for a month was 132 MW (98,849 MWh for the month). If you read Musk's claim - you will see he does not, in fact, state what you wish him to state. He says it COULD run at 500 units per day, or 3500 per week, but he does NOT say they actually run at that rate.

Re:Bloomberg? Why?

[LynnwoodRooster]

Look at your own quote posted above; they have NOT achieved 3500 cars in a week. Musk stated that all parts of the assembly line have demonstrated the ability to do 500 cars per day; that does not mean it has done 500 per day. Nor does it mean it can do 3500 per week.

Re: Bloomberg? Why?

[kenh]

Right, it's a republican thing to:

Take salt shakers of restaurant tables;
Limit soft drink sizes;
And push for anti-gun laws.

You're right, Bloomberg is a raging conservative.

Re: Bloomberg? Why?

[kenh]

500 / 24 hours gives you just over 20 cars/hr.

20 cars/hr gives you one car every three minutes.

That each area of the factory can demonstrate a capability to do it's work on one car inside three minutes "demonstrates" the "capability" of doing 500 cars/day.

Re: Bloomberg? Why?

[sfcat]

Right, it's a republican thing to:

Take salt shakers of restaurant tables; Limit soft drink sizes; And push for anti-gun laws.

You're right, Bloomberg is a raging conservative.

Its a greed thing. Tesla is the most shorted stock in history. $12 billion is bet against Tesla. Most of those positions were initiated a few weeks/months ago when TSLA was trading in a range between $280-305. Recently the price rose to $325 (closed at ~318 today I think) and many short positions were stopped out and a brief short squeeze happened but many traders re-entered new shorts at higher entry prices and defended against a total rout (for now). If the price pops again (because Elon tweets that 5000 model 3s were produced that week), those shorts are fucked and several billion dollars will be lost by hedge funds (and gained by those, like me, long on Tesla). This is the battle being carried out right now. Its mostly retail investors in CA taking on the hedge funds in Conn and NY and the retail side is winning big right now (which rarely happens). If too much more good news comes out about Tesla, it will be all over and those shorts will be squeezed entirely out of the stock (by their risk managers). This might be an effect of our current political divisions, but its entirely outside of politics and an interesting case of the market deciding something against some very big moneyed interests. Its the rare case where the "masses" have more money than a few billionaires and are bullying them around. Its also the rare case where engineering knowledge is far more valuable than accounting knowledge in the pricing of a stock.

Re:Bloomberg? Why?

[sfcat]

The exact quote is: "Despite our production difficulties, all parts of the Model 3 production system have demonstrated a 500 car per day capability, or, a 3500 car per week capability, and we just did a big set of upgrades and are spooling up the production lines again. I think it's quite likely that we will achieve a 5000 cars a week by the end of the month."

This was with 2 GA lines (the biggest bottleneck). They just built a third.

I don't think there's any malice in the Bloomberg tracker. I think it's just hindered by too long of an averaging period on their algorithm. It averages in planned line downtimes in with line production rates, which you don't want.

Its also hindered by the fact its only counting US deliveries. For reasons having to do with preserving the tax credits for more customers, Tesla has only been delivering cars to Canada and foreign markets (left side steering only) lately. This has basically hidden the real model 3 production numbers for the last couple of weeks and those numbers won't become accurate again till sometime in mid-July. Until then, we are all just guessing.

Re: Bloomberg? Why?

[kenh]

All of which proves/supports the notion that Bloomberg is a Republican?

Re:Bloomberg? Why?

[lgw]

Bloomberg has published a slew of negative stories about Tesla,

The financial press exists to call bullshit on corporate numbers.

. Musk announced that Tesla is producing 3500 cars per week, while Bloomberg's estimate is 2560.

Corporate propaganda or investigative journalism: which is more credible? I wonder.

Quality Control

[tgibson]

I don't follow Tesla closely; I found it an interesting read. Having just rid myself of a lemon (not Tesla), I was impressed with this excerpt:

Tesla says it has 47 robots deployed in scanning stations throughout the body line. They measure 1,900 points in every Model 3 to match them to design specs—with a precision of 0.15 millimeters. Torque measurements are also automatically recorded for every bolt that’s fastened. During the final test drives on the track, sound recorders measure squeaks, rattles and wind and road noise that a test driver might miss. All of this data is stored with each car’s unique Vehicle Identification Number, or VIN, so service centers can trace any issue back to a root cause in the factory. The idea is that Tesla will be able to improve its cars, even after they're in a customer’s driveway.

I hope they succeed and force other carmakers to follow suit.

Re:Quality Control

[ole_timer]

nice to see that quality and metrics at work!

Re:Quality Control

[greenwow]

Well that's ISO-9001. You don't fix quality problems. You only document them.

The manufacturing company I used to work that I did machine controls for for was certified based on customer contract requirements rather than an actual improvement to the quality. It was fine producing crap as long as it was documented.

> measuring defects

The company I work for now does a ton of QA and create reports, but for most of them there's nothing actionable from that information. Don't measure something you can't use. It's a waste of time.

Re: Quality Control

[Bing Tsher E]

It goes against all the modern quality initiatives. You don't test quality into your product, you design it into your product.

Re:Quality Control

[Dutch Gun]

I'm not sure I agree. This seems more like the manufacturing world equivalent of automated unit and integration testing of software. Essentially, if anything starts going wrong due to a minor change you make, you're made aware of the situation immediately and can correct it.

Essentially, it's a lot of time and effort that on first blush does absolutely nothing, because in the optimal case, nothing ever happens. But the end result is that you can make minor changes or corrections with more assurance that you're not negatively affecting the quality of the final product.

Re:Quality Control

[hackertourist]

What makes you think other manufacturers aren't already doing this? Dimension scanning and torque figure records are SOP. Don't know about audio recordings of the test drive. Those are going to be pretty hard to match to a specific defect, though.

Re: Quality Control

[Anonymous Coward]

Yes but no! You still need measure or else you're flying blind.

Re:Quality Control

[EnsilZah]

SpaceX managed to track down a faulty strut by acoustic triangulation of a few milliseconds of data from some accelerators, I'm sure it would be much easier for Tesla in a controlled environment where the vehicle doesn't explode.

Re:Quality Control

[lgw]

They are measuring defects instead of eliminating them by design.

"Boss, boss, my new design eliminates all defects!"
"Interesting claim. What numbers do you have to back that up?"
"Well, ummm, ...."
"Try measuring defects for a while, old design and then new. Tell me if it gets better."

Tesla is not unique here. Automotive is HARD

[rahvin112]

Automotive factory spinup is HARD. Your typical car takes over 3-4 years to spin up to full production and they don't make major changes to any model for around 5 years to maximize the CapEX spending this requires. Telsa is new to the game but many of the people they've hired aren't. They did try to do to much automation but otherwise they are experiencing the same growing pains every automotive manufacturer does. Vehicles are hard to build, they have thousands of parts with tight tolerances.

For example, every time Toyota or GM changes a car model they'll design that change 3-4 years before it goes into production and they'll spend a year or two on a test production line refining the production and making part changes to accommodate the best work flows before the new production line is rolled out to a full production facility. Tesla as a new automaker doesn't have this option (they don't even have a test factory), they are doing the testing while building production cars and it's a painful and expensive process that will have fits and starts. Just like the model S they will change that cars subtly as time goes on and they refine production making part changes and swap outs to improve production flow and errors.

Re:Tesla is not unique here. Automotive is HARD

[drinkypoo]

Maybe its just the economy 3-wheel car concept that is destined to fail though. Before Elio there was Aptera. See: https://www.greencarreports.co...

And before Aptera there was the Corbin Sparrow.

Three-wheeled vehicles are inherently stupid, because they are inherently unstable. If you lose a tire, your vehicle becomes difficult to control. If you lose a wheel, the vehicle becomes completely uncontrollable. This is something that actually happens in the really real world, so it's irresponsible not to plan for it.

I used to have a Ford truck with a rear track literally inches narrower in the rear. Even that was annoying. Now imagine trying to dodge potholes with a third wheel in the center of the vehicle. GLWT!

Re: Transform the industry

[Bing Tsher E]

It's a shame Toyota can't just buy Tesla and inject their experienced staff into Tesla to get the job done. S.V. geeks can buy all the gear, but they can't instantly know what to do with it.

Re: They said this about the AI's

[Bing Tsher E]

Playing go is just abstract logic. It's entirely software. Building cars doesn't scale out in any way similar.

Profit vs. achievement

[RonLillycrop]

It strikes me that somehow Elon Musk has managed to chase principled objectives completely independent of financial performance and in a way that has completely fooled the people with the control of money to parting with what they have. Here's hoping that the ruse persists since the actual impact goes way beyond making people some money.

Re:Profit vs. achievement

[EnsilZah]

My impression is he's well aware of financial realities, he just wants to make the greatest impact in the shortest amount of time and is willing to take as much risk as the market will bear for accelerating short-medium term growth.

Re: Transform the industry

[RonLillycrop]

I've had some direct experience dealing with both Toyota and Tesla and their approach to automation. 100% completely incompatible. Toyota: study something methodically, systematically and comprehensively and get better over time. Win. Tesla: Just do it. Fail. Think about it a very tiny bit and then Just do it again. Win. Both good at certain things but a marriage would be the worst thing for both.

Re: Transform the industry

[drinkypoo]

It's a shame Toyota can't just buy Tesla and inject their experienced staff into Tesla to get the job done.

Which job? Making boring boxes that aren't actually more reliable than anyone else's cars any more? If you compare Toyota's cars in literally any space to their competitors', the other vehicle is cheaper, just as reliable, and not just more fun to drive but easier (the two are in fact much the same thing; handling is at least nine-tenths of what makes a car fun, and better handling means more driving ease.)