I think other car manufacturers will catch up in the EV market and Tesla will remain fairly niche. I just can’t see them becoming a mainstream brand with their problems with production, build quality, spare parts availability etc.
Their current market cap is bigger than Ford which is quite simply ridiculous
I agree with this.
They will suffer from first mover disadvantages.
The established vehicle firms will ramp up product releases once the build process is matured and there are plenty of qualified engineers in the labour market for them to hire.
I still remember circa 2010 head of a mainstream motor company (possibly GM) took out a front page broadsheet ad, proclaiming that individual cars are dead for metropolitan cities - and sustainable public transport is the future. But that is for another thread I guess.
Would you believe that Monzo and Revolut have bigger market caps than Greggs? Greggs’ turnover is about £1bn and net profit after tax about £60m. Monzo’s revenue is only £2m (that is million, not billion. That is before any other expenses).
Granted, comparing private and public companies is not the best thing to do, but the concept of growth remains. The former two will either get bankrupt or have an explosive growth, whereas Greggs will just slowly go up (or down).
Similarly, Ford’s revenue growth in the past 10 years has been circa 1% per year, whereas Tesla’s was over 160%. Tesla has significant outliers (659% and 387%) but even a 10-years’ median is 73%. Even taking the last four years into account, Ford is 2% and Tesla 57% YoY.
And Tesla is not only a car manufacturer but also a renewable energy company too, that’s product diversification. And just like any other startup, it will either fail miserably sooner or later or fly to the moon.
I personally would not have it as a majority of the portfolio, but everything is better than crypto so a little “let’s hope it will fly to the moon” will not hurt even if the former scenario happens eventually.
There have been really significant changes in AI or machine learning in the last few years - pattern recognition is performing really well for many web companies, speech recognition has been completely solved, AlphaGo defeated one of the best human Go players, reinforcement learning is making progress too based on large datasets (which Tesla cleverly has been building).
Advanced general AI is not in any way required for road vehicles, which have clear signage, delineated routes, and a set of rules of the road to guide them. Their problems so far seem to be recognising certain edge conditions like a lorry parked sideways across a road - these are not going to be a problem for long and could be solved in multiple ways (hardware, software). Autonomous cars do not need a very sophisticated model of the world and it is improving all the time.
I think the important factor for autonomous vehicles is that they only have to beat humans. They are already safer than a normal human driver in most conditions and that is only going to improve. If you think about it is kind of crazy that we allow humans with minimal training unsupervised control of large multi-ton machines which travel at high speeds next to unprotected pedestrians.
If you take humans out of the loop, things will be a lot safer, as the actions of the other vehicles are more predictable - if roads were clear of human drivers machine drivers could take over tomorrow. I expect at some point if sales of semi-autonomous cars keep growing we’ll hit a tipping point where it becomes sensible to exclude human drivers in larger global cities, and at that point its game over for human control of vehicles.
I don’t hold any Tesla as yet, by my thoughts are here (threads were moved around) - they are ahead on multiple fronts and are behaving much more like a software company than an established manufacturer - Ford and GM for example are hobbled by their past and their reliance on ICE cars for current revenue. This is important in the long run. Musk is unstable but also good at choosing brilliant people and motivating them (as evidenced by SpaceX for example), I think they’d do fine without him soon, once past their cash crunch and a good team are in place.
I do think they are overvalued at present for their likely future revenue though, unless they manage to become a dominant global player very quickly.
All the analysis is interesting. But the biggest thing stopping me investing in Tesla is Elon Musk. No doubt a genius and a world leader, but he has the annoying habit of pressing the self destruct button - usually on Twitter. For me he is too much of a risk to buy any more than a handful of shares.
I understand the argument but I think it’s a slight simplification. On average current AI might be better (as AI drivers won’t get tired, be drunk, distracted etc). However if they perform poorly under certain conditions - e.g. driving through towns with lots of people and unclear road markings - and are less safe in those instances. I think they’ll be rejected by the public. I hope I’m wrong though
The edge conditions are the hard part, It’s straight forward to get a car to follow a road and recognise signs. Dealing with a weird situation that the car has never seen before, and possibly never came up in testing is a huge problem
He also has a pretty dim view of Elon Musk if you reads some of his other entries, (again possibly a bit too pessimistic, but it’s interesting to get opposing views)
I watched a quite impressive video a while ago where an autonomous Nissan made almost a whole journey with no input, but it did get stuck behind a stopped bus at one point and the driver had to take over. Solving these kinds of problems my end up being far harder than getting to the stage we are at now
It has not turned out to be a huge problem in current real life testing. I’m not particularly enamoured of Musk and agree he’s pretty flaky but do think self-driving cars in the next few decades are a certainty IMO. The data already shows that they are significantly safer than humans in the conditions they drive in (admittedly probably the easier set of conditions, and with human supervision), and they are improving.
The article cited excludes human DUI from the accidents it considers - humans are unreliable and drive in bad conditions when they shouldn’t (sleep deprivation, alcohol, drugs), and always will. I find it interesting as a counterpoint but not really convincing I’m afraid due to the manipulated figures. I expect supervised driving to continue for another decade or so but at some point after all that data is collected machine drivers are going to far exceed normal human drivers in every situation. Most humans are simply not very good at driving and would gladly do something else during that time. The article ends with:
Will the future bring AV’s safer than humans that everyone loves to talk about? Probably yes
So I think he’s agreeing with AV taking over, it’s just a question of time.
Re being stuck behind a bus, that’s an interesting example, because the machine seemed to consider overtaking unsafe, whereas the human was willing to take that risk. That’s really a matter of debate and many would be happy for the machine to take a safe slow approach.
In the next few decades sure, Some people have been saying they are nearly here and they’ll be here in the next few years, Elon Musk is one of the ones who I think is guilty of exaggerating how close we are
I think even though they may be 90% there the last 10% will be the hardest
The wider picture as several have said is that while there’s good progress, the deployment phase is going to be terribly long for various reasons [1]. So you could say it’s starting now (Waymo taxis in Chandler) but there will be many years of self-driving having to hand over [2] to human drivers and many years of self-driving cars alongside human-driven ones. All of which adds more complexity.
But it’s not clear yet that Tesla’s bet that cameras-with-ML get good enough quicker than lidar sensors get cheap will come off.
[1] linking to myself, so shameless
[2] and handover is an alarming prospect but humans aren’t good at doing alert-and-primed-for-action for durations of a whole journey
Audi is meant to have Level 3 autonomy but they’ve not activated it because of legal concerns, interestingly Level 3 is where the liability shifts from humans to machine, so not too surprising that car makers are reluctant.
I’m not sure we’ll ever be able to cover all the edge cases so you will need Artificial General Intelligence (AGI). Earlier on today I walked through a narrow street, with a double parked workman’s truck opposite a closed footpath near a corner, so you had people trying to walk past that point while cars were coming the other way.
In terms of AI overall, I don’t think we are close to Artificial General Intelligence (AGI). Most AI is trying to create a computer brain using the human brain as the starting point, but the human brain is complex, and not properly understood by us. Also computers do not really function like brains. There was a article on FT a few weeks ago by Henry Marsh ( a neurosurgeon ) which makes interesting reading Can man ever build a mind?
On Elon Musk, he’s one of few that often warns people on AI. I don’t recall the figures but the amount of focus on making AI safe is nearly negligible when compared to building AI, it’s not something that can be tacked on at the end, so expect a lot of hurdles.
I could be wrong but I’m very skeptical that Audi’s got here first. A quick Google hasn’t produced any articles explaining how they’ve trained their AI & I haven’t heard about them adding the necessary sensors to their existing range of vehicles. I’d expect that to have been mentioned in a story like this, for example. So I’m not sure how they’d have beaten Tesla & co. to it?
Perhaps they mean that they have the technical components, that would be necessary for Level 3 autonomy, ready to go but don’t actually have the capability to fully enable it yet? They’re calling it a pilot here which suggests that it may not be 100% reliable (& could be significantly less reliable) at this stage.
You would expect Tesla, or Google to get there first but the incumbents would definitely be working on something as well. The incumbents are probably a bit more cautious, or don’t need those headlines for funding so they don’t speak about things as much.
Audi has always been big on tech/gadgets, so on some level I wouldn’t be surprised if they had something that’s available for the premium edition of a premium car.
Toyota is also doing big things in this space, their Lexus brand was the first commercial car that could parallel park itself.
Coming to this late but Tesla’s advantage is technology (for now at least) - they still have the most efficient drive train and proprietory cell chemistry - but more significantly cost. They have come through the valley of pain and are able to sell as many of their vehicles as they want and grow as quickly as possible. Their competitors are now developing decent cars but very very few are able to produce them at scale and/or make money off them. Until they figure this out how they wont be able to compete. It’s one thing for GM to sell 30,000 Bolts a year at a $3000 loss on each one; quite another to sell 300,000.
Tesla’s tech lead could disappear quickly (although I dont think it will) but their cost advantages give them at least 5 years on everyone else in my view.
SEC news is out today, as is the disappointing but unsurprising Q1 delivery results. Stock is already down ~10%.
I personally wasn’t so sure on the demand struggle thesis but it seems Tesla is having serious issues here. c. 3000 cars in inventory and 10k+ being shipped. Q1 and probably Q2 profitability will evaporate given the restructuring and price cuts. S and X need urgent revamps now imho otherwise they are going to be to reliant on rushing out Energy + Truck contracts for cash flow.
V quick summary on the Full Self Driving Computer:
14nm chip being fab’d by Samsung, designed in house
Fixed inputs - video only all Teslas being built right now are being fitted with the FSDC
144 TOPS, 21x times as powerful as the world’s next most powerful autonomous drive unit (NVIDIA’s)
Already started work on a gen 2 FSD which is “perhaps 3x as powerful”
Extremely Muskian statements of the day:
Lidar is “a fool’s errand”
Every Tesla vehicle since a few months ago now only limited from FSD by software
Now moving on the Software discussion and eventually demos.
