I think you underestimate how many companies do chip design. That part of it is actually quite unremarkable.

And while that transitional Nvidia hardware obviously isn't ideal, it's still necessary to run Autopilot today. And it is what is allowing Tesla to gather mountains of real world training data for its machine learning algorithms.

I'm not confident Tesla will pull off their vision of a driverless taxi business, but if they do pull it off, the business fundamentals promise astronomical revenues. Nobody questions that Uber and Lyft are viable businesses. Tesla's plan is to be in their space but with cheaper fuel, cheaper labor and a greater revenue share.

there is a line of reasoning, rarely articulated, that EVs will never reach the price points it would take to fully 100% replace ICE cars in the private/personal market.

li-ion batteries have done an excellent job walking down the learning/cost curve, but we can all see the way it bends and how much headroom is left in the chemistry. we'll likely hit $100/kwh in the next few years, but even just $80/kwh is likely 5+ years away after that. the model 3 will never start with a $2.

so if you're going to bottom out at around 3x the price it would take to fully replace the existing market... then you have to figure out a way to get 3x+ the value out of the asset you're delivering.

that means some combination of lasting 3x longer (~500k miles) and/or being in use 3x more (~600hrs/year). the former is viably unlocked by the EV engineering advantages, but w/r/t the latter, individual drivers have no intention of tripling their car usage. so it simply has to be turned into some kind of shared asset.

imho self driving makes this a nirvanna, so I'm glad they're pursuing it, but its not actually required. tesla could add a basic uber/lyft feature set to their existing app and then when the last 0.01% of self driving winds up taking years longer than elon said it would he can blame the regulators and just make it a regular human-driver ride sharing app.

remember the mission is the end of carbon emissions, so any plan for "lets just get good market share and margins in the luxury segment" is a plan to fail.

I think that view neglects the costs of the fossil fuel infrastructure. At some point gas stations are going to start closing for lack of customers. That time is a while off yet, but all of the existing stations are not going to be able to be supported by say, half as many ICE cars. Convenience will go down and at some point prices may go up as the existing infrastructure is paid for by fewer stations and less fuel. At the same time electrics will continue getting better, cheaper and more convenient. This could provide an extra push.

I agree with your points on a longer time horizon, but elon needs to solve for model3 demand now (this year).

EVs will not drive the kind of oil distribution market retraction you're describing until they're at least 10% of cars on the road, which would be years and years after they're 10% of new cars sold.

So while I agree the affect you describe will help, we don't have time to wait for it.

I think the time horizon for a step change in oil distribution is a lot shorter than people expect, precisely because it will be more likely a severe step change (ie, a snowball/crash) than an easy transition. Oil distribution is extremely complex, amazingly baroque, and has only a small number of players (they use a shell game [Shell Oil pun intended] of a large number of different "station" brands, but are really only a few companies left). When instability hits, it is likely to hit hard and look almost immediate (look at 70s Oil crises, for example).

I don't yet have a guess what percentage of the car fleet needs to be EV for that to happen, but that number is also not the only potential disruption that may happen to cause it (again, we have the 70s Oil crises as examples).

(Not that causing an Oil crisis is necessarily the right way to combat climate change, but in the question of chicken-and-egg between ICE and EV, people sure have a short memory for how volatile Oil is.)

It is currently manufacturered in Fremont California. Nothing in Fremont is cheap, it is no different than manufacturing in Manhattan, since employees need to pay for super high housing costs and deal with high costs of doing business.

Manufacturing in China could bring the cost down significantly, with lower wages, taxes, material costs, etc.

i don't know why this comment is being downvoted. He's absolutely right about every word. Fremont is not a place to be building cars. If you're gonna build in the US, Tesla could have chosen almost any other city in the US and gotten labor cost that was 33% cheaper.

Software companies can afford the higher labor cost in the bay area because they tend to have really large margin when they win. Car companies have very very low margins, even when they win, so they really need to pay attention to things like labor cost. This is a huge mistep for Tesla and shows poor judgement on Elon Musk or whoever decided to do it in Fremont.

Tesla over estimated the benefits of automation. Maybe they thought that labour costs wouldn't be that much of a factor anyway.

Musk long ago was talking about his calculations about the physical limits to assembly speed. “the output is going to be volume times density times velocity.” https://electrek.co/2016/09/15/elon-musk-confident-that-tesl...

A genius no doubt. I will see you all on Mars.

They bought the old NUMMI factory that used to make GM and Toyota cars. It was active right up until they bought it in 2010 (they announced the sale a month after NUMMI closed) so my guess is they were able to accelerate early progress because of that purchase.

In the long run, I'd agree, the expense vs. reward calculation doesn't make sense now. But GM and Toyota successfully used that plant for something like 30-50 years before that and there was a whole community of auto workers living around it, so I'm not sure I'd say that was poor judgment for them at all.

Nobody expected housing prices to go up 100%+ over the last few years. 2010-2015ish was a real estate boom period for Valley workers if anything. You had post-recession housing fire sales; AirBnB wasn't big enough yet to incentivize empty houses; Google, Apple and FB didn't quite own the entire Menlo Park->Sunnyvale corridor yet; etc. Good times.

> EVs will never reach the price points it would take to fully 100% replace ICE cars in the private/personal market

The Chevy Bolt and to a lesser extent the Model 3 have already hit the "Corolla/Civic/Sonata/et al" sedan sweet spot for new sales. The mean in new car sales has always been around and just above $30k. Certainly the mean in total car sales is closer to $20k, but that hugely because of the secondary market and fast depreciation (used car sales). It's still too early to tell what sort of impact models like the Bolt and Model 3 may have on used sales (and obviously production numbers are still a factor, especially with Tesla's production and delivery woes).

Chinese manufacturers have been exploring the "bottom" of the EV market much more effectively than the US and EU markets simply because they don't need the carrot of luxury to attract early adopters in quite the same way. If Chinese EV companies are to be believed, hitting below that magic USD$20k mean is possible with new car sales alone due to EV supply chain efficiencies that ICE supply chains can't match.

> self driving makes this a nirvana

I feel self-driving is a red herring/impossible goal like the 60s AI boom all over again, and we're all going to look back on this in a decade or so and wonder what everyone was thinking.

Also, ride sharing isn't likely the answer in current America. It's enough of a culture shift to get gas-lovers/petrolheads to enjoy/love/trust EVs. It's even more of an ask to remove ownership from the question and to force them to only ever rent a car. We might not have time to take only one hurdle at a time, but we may have to.

> The Chevy Bolt and to a lesser extent the Model 3 have already hit the "Corolla/Civic/Sonata/et al" sedan sweet spot for new sales

GM sold 18,000 Bolts in 2018 [1]. Over the same period, Toyota sold 280,000 Corolla Sedans [2].

[1] https://www.theverge.com/2019/1/3/18166619/gm-ev-tax-credit-...

[2] http://carsalesbase.com/us-car-sales-data/toyota/toyota-coro...

The implication in the further context from what you quoted was sweet spot for price for new sales (not actual numbers of sales).

I've read that the wiring harness of a modern ICE car costs more than the transmission.

Do you think it's possible that all we need is ~$100/KWh, after which due to diminishing returns, find better ROI improving other areas like a dramatically cheaper wiring harness?

Yes but its not enough and its not soon enough. Tesla needs to stoke demand this quarter. Humanity needs to be driving toward zero emissions yesterday.

IMHO privately owned electric vehicles will only ever have been a historical oddity of the transition between personally owned cars and shared/pooled cars.

> Keep it simple stupid - focus on making great high-value EVs, and keep dropping their prices in a profitable manner.

Problem is, self-driving is an essential part of their image of high-tech luxury. Their marketing strategy depends on being the opposote of KISS. Making the best cars they can while also being electric is the only way to get the volume necessary to make electric affordable.

I don't care about self-driving tech in the least, for me it is about as likely for us to get level 5 autonomy as us getting nuclear fusion soon. Just make a good cheap electric car that works. The new Tesla chip info I've heard promising it next year sounds like a complete lie/classic Elon Musk fabrication.

https://www.inc.com/geoffrey-james/elon-musks-dumb-lie-about...