Transportation 2.0 update

Roughly three years ago I wrote some thoughts about the upcoming revolution in transportation, electrification and autonomy. Time for a quick update.

In the electrification front, sales of battery electric vehicles (BEV) in Europe (EU + EFTA + UK) + US have been blooming1,

BEV sales in units

I have excluded China and hybrid sales, because in both situations basically anything with a wheel and a battery can (and will) be considered an electric vehicle, which distorts the data to the optimist side.

Bottom line in the electrification sales in Europe+USA have reached 3.2 millions units in 2023, a 164% increase from 2021.

There is some near term concern with interest rates, sold out media to big oil and legacy auto (pumping hybrids and denigrating electric cars), and looming price wars with China… but despise all this, the long game is playing out very well, and in my mind it’s already won due to Wright’s law and the learning curve effect. The learning curve model posits that for each doubling of the total quantity of items produced, costs decrease by a fixed proportion (observed experience curve effects for various industries ranged from 10% to 25%).

Every battle is won before it’s ever fought

Sun Tzu

For instance, from 2020 with 800k units (Europe + USA data) and a cumulative production trough 2023 of 6400k units, we can calculate with a conservative 10% learning slope curve a 27.1% cost improvement over these last 3 years. In the real world EV prices actually have been dropping and the specs getting better each year. It’s kind of comparing apples to oranges, to compare a battery electric vehicle (EV) to an internal combustion engine (ICE) car, but I believe point of sale price parity is just around the corner, and the trend will be cheaper EVs than ICE in the (near) future.

It’s important to notice ICE industry doesn’t benefit pretty much from scale anymore, Earth currently has circa 1.5 billion cars2 in the world and a yearly 75 million production (an average 20 years renew of the fleet) so it would take at least 40 years to double the cumulative production… I’m convinced we will experience a curve discontinuity, when existing processes become obsolete by technological change.

Actually in the beginning of the ICE industry we can visualize Wright’s law working hard in Ford model T yearly price/units produced.

Also has expected, battery tech is progressing fast. At pack level (more relevant than cell level) the energy density trend is pretty self explanatory and there is still plenty of room for progress, while combustion engine is at the end of the technical development line.

Tesla Battery Pack Energy Density Wh/kg

Data points source3

YET… I reiterate, the true revolution it’s not the disruption of ICE, but it’s brewing in the autonomous vehicle (AV) front. And in the world of silicon chips and code lines, the frenetic Moore’s law is king. Dictating an expected doubling of capacity for the same cost every two years, hence a compound annual growth rate of 41%, Moore’s law predicts 180% better systems from 2021 to 2023 at the same cost (300% better by this year end, 21x better by decade’s end at the same cost… that’s the power of the compounding effect).

And boy there are big practical developments in the computers world, forget about machines playing chess (or go) at super human level, forget about image and text recognition. Artificial intelligence (AI) large language models (LLM) bots, being ChatGPT the most known to the general public, can now pass the Turing test!

Proposed by Alan Turing, it tests for intelligence in a computer, requiring that a human being should be unable to distinguish the machine from another human being by using the replies to questions put to both.

Also AI can now easily generate images and videos according to the user prompt. Let’s say, “draw me a picture of a dirt bike popping a wheelie on Mars with several spaceships in the sky” and after a couple of seconds… booom there you go….

So, as expected lots of progress going on, but can a machine drive a car safely in the real world with other road users around?

And the short answer is: YES IT CAN. Google’s Waymo currently operates daily commercial robotaxi services in parts of Phoenix, San Francisco and Los Angeles.

Google’s Waymo in SF

Also General Motors launched Cruise, a robotaxi service, but after several incidents, like cars freezing causing gridlocks, and crashes involving pedestrians and third party vehicles resulted in the suspension of the license to operate without a safety driver, and GM opted to halt it’s AV efforts “while undertakes a comprehensive safety review”4. Once again Mary Led and it matters.

Joe Biden notable remarks in 2021 Detroit Evs conference. Since, in 2023 Tesla sold 1.81 million (1.81 x 1000000) EVs, and GM sold 66 thousand EVs (66 x 1000) or about 3,6% of Tesla EVs sales…..

Also in the legacy auto world, Ford and Volkswagen, shut down Argo AI5 their joint effort to develop self-driving cars without achieving nothing relevant.

Meanwhile Mercedes has launched Drive Pilot, the first (and so far only) SAE Level 3 system (you can take hands out of wheel and eyes out of road) approved for sale in Europe and the US. Highly publicized It’s a mere marketing gimmick, because it only works on specific pre-mapped highways, during daytime, in good weather, and up to 40 mph in traffic jam situations (following a car).

Really what to expect from legacy auto in AV? As in 2024 they still struggle with simple software stuff like over the air updates and infotainment systems.I really don’t expect nothing but marketing.

From the tech world, Mobileye an Israeli spinoff from chip giant Intel focused only in AV technology, is currently valued at 26 billions, and it develops AV hardware and software to auto makers. Mobileye actually was the original provider of Tesla first autopilot, but after a deadly crash Tesla decided to pursuit AV on it own6. Also worth mention is Comma.ai, creation of George Hotz (the Iphone and Playstation jail brake hacker) that develops Openpilot, an open source (find it on github) effort that provides advanced driver assist features to several car makers trough purpose built hardware designed to be installed in the car.

There are clearly two different approaches to solve the AV problem, the legacy auto (and Waymo) using highly detailed and up-to-date maps combined with a suite of sensors like high precision GPS, lidar, radar, 5G connectivity. And the tech world going for cameras (vision only) and AI solution.

While the first approach, can bear fruit much quicker in geofenced applications like Waymo or Mercedes Drive Pilot it’s non practical or economical to scale, it’s a nightmare and expensive to produce and keep updated 3D maps at millimeter precision, and all those sensors come with a hefty cost, so (at least by now) it’s a non economic and non scalable solution. A general AV solution is needed, and in my mind this will be achieved with cameras and AI neural nets, you have to teach a F*g machine to see and react in real time to the world circumstances and changes.

$TSLA

So, if you read the original Transportation 2.0 post back in the day and decided to invest some hard earned money in $TSLA (260.20 close at the posting day) you are now a certified bag holder carrying a 7% loss.

General sentiment is not great, inflation, interest rates, waging wars are taking a toll in the markets generally, and Tesla specifically is also victim of a strong negative narrative from main stream media, and some US government harassment specially since Elon Musk bought Twitter.

So, is this the end? Better to sell all and move along? CNBC, Bloomberg, CNN heck most of the main stream media and Zuck social networks are spoon feeding the viewers with the same message, Elon Musk is bad, EVs suck, Tesla is doomed, FSD is a public danger, Cybertruck rusts overnight, the electric small/affordable car will never be produced, Tesla is doomed. investing in Tesla is crazy, the Chinese are coming, the end of world is upon us…

They must be right? Right? When, in doubt zoom out (and filter signal from noise).

Tesla # Cars Delivered per Year
Tesla GAAP Net Income Per Second (in USD)

I personally feel some 2018 vibes, when the company was between 2 growth waves, the first wave was the proof of concept with luxury segment Roadster, Model S and model X (tens of thousands units production), and the second wave was the EV mass production with Model 3 and Model Y (millions of units production). Back in the day, just like today, main stream media (with the best compliments of legacy auto and big oil advertising dollars) was pumping the don’t buy a Tesla and don’t touch TSLA stock narrative, they were soon to be bankrupt they say, DOOMED they say.

Tesla is doomed…

fate loves irony as Bob Lutz was a former vice chair of GM bailed in 2009 at a cost in excess of 10 billion USD to the tax payer.

I love it, it shows the fear and desperation of the bears.

Tesla FSD solution is now in 12.5 supervised, meaning it can do all the trips for you without touching the controls, but it requires a person at the driver seat supervising and being ready to intervene and override any error of the system. It’s not geofenced, it works well in rain or night conditions, with or without traffic (leading cars). A general solution for all, that is driving more and more miles.

Is it perfect? No, it’s a ongoing effort, and as time goes by It’s actually learning to drive better and better. Many FSD users report most of trips now are zero interventions, and it’s pretty easy to find Youtube videos of the car driving itself on FSD flawlessly for hours (the videos are getting quite boring).

We also know that in AI systems more data and more compute are always better. This is a big advantage to Tesla, because it’s the only auto maker with 6 million + vehicles with data collection and reporting capability. Talking about more compute, Tesla is developing a huge supercomputer cluster, code name Cortex, in Giga Texas scheduled to be operational in a few months, powered by a system up to 500MW specifically designed to train AI models.

Tesla has the biggest fans

So, when will the Tesla FSD be unsupervised? When will it drive around without no driver? Well my best guess (yes, it’s a guess) it will be available for customers in the next couple of years. There is a massive effort going, results are coming in. Back in 2021 it could drive for a couple of minutes without disengagements, now we are in the days mark and pretty sure weeks without disengagements are coming in. Tesla is getting closer and closer to the holy grail.

Pretty soon in 10/10 Tesla is doing the robotaxi unveil in the Warner Bros studios, and for sure we will have juicy updates and more visibility about the current state and future developments.

Other toughts amd considerations ex manufacturing and FSD

Biden administration is clearly anti Elon Musk, maybe because Tesla is not a unionized company, maybe because he’s a billionaire, maybe because he leans more to the center right, maybe because he bought and freed Twitter from left bias, maybe because he is a free thinker. The bottom point is that if Kamala wins the November USA presidential election, the federal harassment will continue, so let’s hope for a Trump win and a free market normal operation.


The media bias, as most of the media is left wing biased and on top of that Tesla doesn’t advertise (so they don’t get their money) it’s a daily stream of false information against Tesla that can eventually take it’s toll. The good point is that someone said any advertising is good advertising, and the facts speak for themselves and most persons with a functional brain can dismantle the lies and the bias really fast.

and….

so…. probably the Norwegians are just stupid or the news is stupid?

The TSLA stock chart is very promising with a massive symmetrical triangle ready to pop.

Symmetrical triangles represent a pause in the prevailing trend as bulls and bears reach an equilibrium. However, once the price breaks out decisively from the triangle, it often signals the start of a new trend or continuation of the prior trend. The direction of the breakout, whether above the upper trend line or below the lower trend line, tells you which side has gained the upper hand

Yesterday the Federal Reserve reduced the target for its key lending rate by 0.5 percentage points, to the range of 4.75%-5%. The first drop in 4 years signaling the beginning of an easing cycle that is historically very positive for stocks.

So if you are holding the bag since the 2021 post, be excited about the future and hold, there is a bright future ahead, and your patience can (and will) be very well rewarded (nfa).

Personally I wouldn’t bet against Tesla and Elon Musk.

  1. Data sources: European Automobile Manufacturers Association, Kelley Blue Book ↩︎
  2. Data source: fair estimate from automotive industry research firm Hedges & Company. ↩︎
  3. https://www.youtube.com/watch?v=iWOD55f2mEk by https://x.com/LimitingThe ↩︎
  4. GM 2023 Annual report filing 10-k ↩︎
  5. Ford 2023 Annual report filing 10-k ↩︎
  6. Tesla and Mobileye disagree on lack of emergency braking in deadly Autopilot crash ↩︎