Tesla and Uber Unlikely to Survive...

[RobDickinson]

EddieSteadyGo said:

RobDickinson said:

Elon has pretty much bet the company on that, designing and building your own chip has cost them $1bn plus I suspect with the rest, 150 AI hires this year etc It looks like it is going OK so far but its still a gamble

This "full stack" approach from Tesla looks quite arrogant to me. And it adds a lot of risk. And it doesn't maximise the opportunity.

So Tesla want to build the chip, the software, the main hardware, and then sell the car. So like a full stack.

If they make a mistake, or it turns out someone else finds a much better solution (which is highly possible when you are at the cutting edge of tech), they risk the whole thing.

Whereas if they continued using strategic partners, like their previous approach with Nvidia, they can share the investment cost with other companies and help target their own money into areas like the software, which should be their USP. They won't control the full IP in this way, but being realistic they aren't going to beat Nvidia or Intel or AMD over the long term at making chips.

They are doing that where they can, Tesla make their own car seats, pretty much the only car company that does (there are 4 major car seat suppliers).

They also have had mobileeye (intel) and nvidia involved in their self driving program, and have chosen (with that knowledge) to go their own way.

The guy heading up the hardware is a recognised industry leader, the new chip looks very good from a perf/power point of view and the low batch level is what they are after.

I basically dont think nvidia could supply tesla with hardware that matches what they need (doesnt mean it wont be fine for someone else), but the current nvidia board is 140x batch level and 500w ( more powerful but fragmented also).

[gangzoom]

EddieSteadyGo said:

They won't control the full IP in this way, but being realistic they aren't going to beat Nvidia or Intel or AMD over the long term at making chips.

FSD isnt just about CPU power or even sensory hardware. Look up how many Terflops of CPU time DeepMind has access too, IBM have committed to neural nets etc. CPU time is not the limiting factor.

Its the software that matters almost more. All our computers/phones/games consoles are useless bricks without the right software.

I just cannot see how Tesla have got the software talent to compete with the biggest names in tech. Building an EV is really easy compared to what they are trying to do with software.

To bet your entire company on the promise of FSD is really really brave/stupid.

[RobDickinson]

Burwood said:

Few will care. As I have said before, it's the overall package not the range or winning a 0-60

People keep saying this but the main thing anyone looks at with an EV is range, and the majority of people will cross shop between brands.

Taycan 217 miles
Etron 204 miles
Tesla 370 miles

Its not a few its damned near double.

[kambites]

RobDickinson said:

People keep saying this but the main thing anyone looks at with an EV is range, and the majority of people will cross shop between brands.

The way I look at it when considering what to replace our family car with is that I have a required range which is a hard cut-off - all I care is which side of that number a car falls, how far it falls there is irrelevant. So I'll draw up my shortlist by looking at cars with sufficient range but beyond that range will not play a part in my buying decision.

Right now, many EVs fall the wrong side of that cut-off (which for me is 200 miles in the worst conditions we ever get in the UK, so probably 250-300 miles WLTP), but give it a few years and I think pretty much all family sized EVs will have configurations which can achieve that.

Edited by kambites on Thursday 25th April 10:06

[RobDickinson]

kambites said:

The way I look at it when considering what to replace our family car with is that I have a required range which is a hard cut-off - all I care is which side of that number a car falls, how far it falls there is irrelevant. So I'll draw up my shortlist by looking at cars with sufficient range but beyond that range will not play a part in my buying decision.

Yep sometimes enough is enough - personally I need one with 500 km of realistic range (least in summer), which is why I'd likely go for the 135kwh rivian not the 185 etc.


I think in general 200 miles is OK for bumbling about but I have doubts the etron or taycan will make German autobahn users very happy.

[EddieSteadyGo]

gangzoom said:

FSD isnt just about CPU power or even sensory hardware. Look up how many Terflops of CPU time DeepMind has access too, IBM have committed to neural nets etc. CPU time is not the limiting factor.

Its the software that matters almost more. All our computers/phones/games consoles are useless bricks without the right software.

I just cannot see how Tesla have got the software talent to compete with the biggest names in tech. Building an EV is really easy compared to what they are trying to do with software.

To bet your entire company on the promise of FSD is really really brave/stupid.

The chip thing is like Microsoft launching Windows 3.1, and saying we won't sell it to anyone unless they also buy a Microsoft made PC.

I highly doubt Microsoft would have become one of the world's largest companies if they took that approach.

On the FSD, I doubt they have the required amount of money to invest over the next decade to invest into software development. But this becomes even harder when you divert precious resources into chip development.

[anonymous-user]

Musk hinted on the call that he had some interesting insights into customer preference between 220 range and 240 range. And that range matters more than you might intuitively think. Hence the focus on increasing range.

[Burwood]

RobDickinson said:

kambites said:

The way I look at it when considering what to replace our family car with is that I have a required range which is a hard cut-off - all I care is which side of that number a car falls, how far it falls there is irrelevant. So I'll draw up my shortlist by looking at cars with sufficient range but beyond that range will not play a part in my buying decision.

Yep sometimes enough is enough - personally I need one with 500 km of realistic range (least in summer), which is why I'd likely go for the 135kwh rivian not the 185 etc.


I think in general 200 miles is OK for bumbling about but I have doubts the etron or taycan will make German autobahn users very happy.

where does the 200m range come from. Porsche claim 305 and i would agree that it's pie in the sky but I'm not sure about 200 real world. Id expect 250.

[gangzoom]

RobDickinson said:

The guy heading up the hardware is a recognised industry leader, the new chip looks very good from a perf/power point of view and the low batch level is what they are after.

CPU time is not the limiting factor, Deep Blue was able to calculate 200,000,000 calculations per second in 1997, IBM was supplying racks with up to near 500 Teraflops to Universities in 2007 - So IBM had better hardware over 10 years ago!!

I would imagine the CPU time Tesal/Mobileye/Navidia are quoting is like doing 1+1 for a company like IBM these days. Yet if FSD was just a matter of CPU time why hasn't IBM already launched their own FSD equipment??

Tesla have made real progress with the AP software, and I was ready to 'gamble' £5k for the FSD software. But the vision Musk is now selling is an eutopian future that barely exists in SciFi books, we have to wait and see how it plays out, but I would say there a good 90% chance Tesla will crash and burn on the FSD promise.

[anonymous-user]

gangzoom said:

the vision Musk is now selling is an eutopian future that barely exists in SciFi books, we have to wait and see how it plays out, but I would say there a good 90% chance Tesla will crash and burn on the FSD promise.

I'd agree Musk is massively exagerating progress and underrepresenting competition. But waymo has working taxis in arizona which work pretty well already. Recent research papers confirm vision can emulate lidar pretty well. 3 years is a reasonable timeline for what musk is describing.

[RobDickinson]

Burwood said:

RobDickinson said:

kambites said:

The way I look at it when considering what to replace our family car with is that I have a required range which is a hard cut-off - all I care is which side of that number a car falls, how far it falls there is irrelevant. So I'll draw up my shortlist by looking at cars with sufficient range but beyond that range will not play a part in my buying decision.

Yep sometimes enough is enough - personally I need one with 500 km of realistic range (least in summer), which is why I'd likely go for the 135kwh rivian not the 185 etc.


I think in general 200 miles is OK for bumbling about but I have doubts the etron or taycan will make German autobahn users very happy.

where does the 200m range come from. Porsche claim 305 and i would agree that it's pie in the sky but I'm not sure about 200 real world. Id expect 250.

Porsche claim 311miles NEDC, to get EPA range divide by 1.43 (with 11% accuracy) So could be 240 or 199 or something. Lands at 217 miles

[RobDickinson]

gangzoom said:

RobDickinson said:

The guy heading up the hardware is a recognised industry leader, the new chip looks very good from a perf/power point of view and the low batch level is what they are after.

CPU time is not the limiting factor, Deep Blue was able to calculate 200,000,000 calculations per second in 1997, IBM was supplying racks with up to near 500 Teraflops to Universities in 2007 - So IBM had better hardware over 10 years ago!!

I think you would struggle to get Deep Blue in the glove box of a car. And I am guessing it uses more than 100w.

its certainly not all about CPU performance, but you need a certain amount and it has to be workable in a production EV and not use up all the battery

[Burwood]

Pricing for the VW ID confirmed at Euro 29,999 before any subsidy (48Kwh). No word yet on the launch edition 62Kwh version. The range quoted is 200m for the lower output battery, again take with a pinch of salt. 170 maybe.

VW claim they will make a loss of 3k on every one sold until 2025. Although that is an accounting loss most likely linked to the amortisation of the billions in R&D

So looks like with govt grant £25K to £28K (+3 for the bigger battery?) plus a few options 28-31 gbp? Not cheap but within the sweet spot. I expect most cars to sell in basic trim as I would think it's all about the utility

[gangzoom]

sambucket said:

I'd agree Musk is massively exagerating progress and underrepresenting competition. But waymo has working taxis in arizona which work pretty well already. Recent research papers confirm vision can emulate lidar pretty well. 3 years is a reasonable timeline for what musk is describing.

I agree, Waymo must be getting somewhere, especially with their rideshare App now live on the App store.

I hope Tesla can pull if off, but its a might big task, not only do they have to build the cars (easy bit), build the best CPU hardware, and beat the best software companies around.

Essentially become the best in everything tech, not saying its impossible, but a very big risk. No one has done it before.

Oh looks like IBM is now at the Petrflop level for CPU power, so interms of CPU power the Tesla FSD is a basic calculator in comparison.

https://www.ibm.com/ibm/history/ibm100/us/en/icons...

[gangzoom]

RobDickinson said:

I think you would struggle to get Deep Blue in the glove box of a car. And I am guessing it uses more than 100w.

its certainly not all about CPU performance, but you need a certain amount and it has to be workable in a production EV and not use up all the battery

Hardware shrinks we all know that, the fact is even with a boot/car full of CPUs no one has demonstrated FSD. The hardware isnt the bottleneck, and hasn't been for the best part of a decade.

The performance by Musk was a good one as many people see the hardware and link it to delivery. But the software is the real key.

[Tuna]

gangzoom said:

Why are people even bothering with looking at production/sales, with FSD in a $35k product, Tesla will take over personal transportation. Undercut all Taxi services, public transport, and current car ownership models.....

..and if they invent the teleporter, it'll be even better. That's the whole point - genuine FSD at this stage isn't just a small incremental step - it's a miraculous leap. It may never be possible for some traditional journey types (door to door presents some huge challenges), and even for special cases (long distance lorries) the technical and regulatory hurdles are massive.

It's a very big if... and the consequences will not necessarily be what you expect. Market forces don't support products making unusually high margins, so FSD cars will be far from prized possessions. They'll be snotters run into the ground 24/7 to generate cash. Maintenance, battery lifespan, reliability will be the issues. A hose-down interior wouldn't hurt either. That's a country mile away from the affluent middle class eco-warrirors who're buying Teslas right now.

Short term you'd expect everyone to pile into the market expecting to make millions, then undercut each other as it turns out institutional investors don't need anything better than single digit returns, then the harsh realities of long term maintenance will kick in and half of the owners will be wiped out. Then someone figures out how to put FSD in a $7000 wipe clean ICE vehicle and everyone moves to the significantly lower capital investment to get the best return.

FSD pretty much wipes out the advantages of EVs - if you're not driving you don't need a high power, high torque engine, so why not put in a bulletproof 40hp ICE engine that does 100mpg+ and will run for a million miles without maintenance?

That's ignoring the issues of regulation - which are going to be massive, and social - you can get someone to send round $35,000 worth of car with no-one guarding it? What could possibly go wrong?

Predicting the future is never a case of extrapolating our current usage patterns.

[anonymous-user]

Tuna said:

That's ignoring the issues of regulation - which are going to be massive, and social - you can get someone to send round $35,000 worth of car with no-one guarding it? What could possibly go wrong?

Tesla insurance seems interesting. All that data on driving habits....Tesla refusing to release any autopilot safety data, because they want to retain information arbitrage advantage for initial sales.

I think though that big change is that Musk's promises don't have any weight any more. Tesla has limited time now to make a genuine case for being a disrupter.

This data can only hide so long.

Edited by anonymous-user on Thursday 25th April 14:41

[DonkeyApple]

PMacanGTS said:

No one really cares about range when you get over 250 miles. No one drives further than that without needing to take a piss.

What you’re really saying is that only people over 50 are able to afford them.

[Tuna]

sambucket said:

I think though that big change is that Musk's promises don't have any weight any more. Tesla has limited time now to make a genuine case for being a disrupter.

That was always the case for 2019 though - even without the FSD distraction, he's moved into the phase where his company needs to be showing results for the decades and billions of investment put into it. The Model 3 is a watershed moment when he can't just point at early adopters and guess at demand.

The fact that FSD is being given such prominence suggests that the 3 has not 'settled the argument' of whether Tesla can succeed. But we now have a lot of people looking at their watches and asking when that question is going to be answered. You can't be in research mode forever.

[DonkeyApple]

sambucket said:

Musk hinted on the call that he had some interesting insights into customer preference between 220 range and 240 range. And that range matters more than you might intuitively think. Hence the focus on increasing range.

Different markets have different geography. For a very simplistic approach just look at the distances between points of civilisation in say California v the UK. Two similar economies with similar demographics and consumer spending powers but enormously different data re car usage.

So in the UK you will see an earlier tail off in the power of range and it being replaced sooner by other key factors such as brand image and financing costs etc.