Self-driving tech startup values crash 81% in 2 years

It's not a great time for self-driving tech startups. The ones that managed an IPO in the past two years have lost a combined 81 percent of their market value, dropping from a total of nearly $51 billion offering time to a mere $9 billion this month. Crunchbase, which tracks startup valuations and investments, said in its …

[VoiceOfTruth]

Overvalued by people who don't know what they were valuing

We had the dotcom bubble. This is the AI and ML bubble.

Remember the words of Scott McNealy about the madness surrounding Sun's market valuation at peak. This is a partial quote, you can search for the rest: "At 10 times revenues, to give you a 10-year payback, I have to pay you 100% of revenues for 10 straight years in dividends. That assumes I can get that by my shareholders. That assumes I have zero cost of goods sold, which is very hard for a computer company. That assumes zero expenses, which is really hard with 39,000 employees..."

A quick way to grab some lucre these days is to put the letters AI or ML on something. Somebody gullible out there will buy it. That doesn't mean that all AI and ML things are essentially worthless snake oil, but a lot of things simply do not measure up.

[Yet Another Anonymous coward]

Re: Overvalued by people who don't know what they were valuing

>We had the dotcom bubble. This is the AI and ML bubble.

And yet we now have ubiquitous internet and smart phones, instant virtual finance services and online delivery of every product we use .

[Sorry that handle is already taken.]

Re: Overvalued by people who don't know what they were valuing

If I understand you correctly, the dot com bubble was characterised by hype and mania for internet-related investments regardless of any supporting business cases. That some online technologies became ubiquitous doesn't change that.

[John Brown (no body)]

Re: Overvalued by people who don't know what they were valuing

Replying to a quote out of context is why got downvoted. You missed the last sentence: "That doesn't mean that all AI and ML things are essentially worthless snake oil, but a lot of things simply do not measure up."

[Yet Another Anonymous coward]

Re: Overvalued by people who don't know what they were valuing

> Sun's market valuation at peak

"At peak" for a HW company in a niche market.

Would you have paid 10x earnings for ARM in 1993?

[Sorry that handle is already taken.]

Re: Overvalued by people who don't know what they were valuing

If there was a suggestion at the time that it was sensible, sure. Otherwise it's gambling.

[John Smith 19]

A VC and their cash are soon parted.....

I guess (subconsciously) people are thinking "Humans don't need active sensors to drive"

And they're right.

That said it seems like a quick fix to help the problem.

[Sorry that handle is already taken.]

Re: A VC and their cash are soon parted.....

Except in this case the VCs made out like bandits. These numbers are post-IPO – the VCs have already taken their cash and run, leaving the public holding the bags.

[Michael Wojcik]

Re: A VC and their cash are soon parted.....

Exactly. The VC calculation for a startup isn't "will this company be successful?", but "will this company have a profitable IPO?". (Or, alternatively, "will I be able to hand this hot potato off to the next guy who figures he's not at the end of the chain?".)

[Anonymous Coward]

Re: A VC and their cash are soon parted.....

Most VC's don't even do that. They rely on probability - if I invest in enough startups I only need one of them to be a unicorn to make out like a bandit.

[Kev18999]

FSD is hard, all those Waymo cars still need a team of IT pros to monitor it and control it and Tesla is far from activating FSD beta for more people.

[OhForF']

Companies value

Just a year after its debut, Embark is valued at less than the cash reserves it had at the end of its last quarter. Shares are down a whopping 97 percent from their debut price

If the C-Suite at Embark believes their company is worth the debut price the logical thing to do would be using those cash reserves to buy back as many shares as they can get hold of.

Wonder why this doesn't happen ...

[Claptrap314]

Re: Companies value

That only works if you don't actually need cash reserves--that means be profitable after taxes, depreciation, interest & the like & be confident that your market isn't going to dry up in a recession.

Not saying that they should not buy SOME shares back with SOME of the reserves.

[Anonymous Coward]

Coincidentally?

SMBC's strip for today

Initial Fruit Offering

[bazza]

Expensive Snake Oil

So the markets have lost $42billion to glitzy PowerPoint and hype.

I know that there is little enthusiasm for intervention in markets, but situations like this are nuts.

Economies need investors to be savvy about what is good and what is bad. Self driving cars have clearly always been a bad idea unlikely to succeed, yet that has not stopped Persons B, hoping to get rich quick, being persuaded to part with $42billion and give it to Persons A in return for some really very glossy brochures.

It's legalised pump and dump, with "technology risk" being the thing that makes it all okay. If it were a financial scheme people were being encouraged to invest in, there'd be accusations of fraud and court cases

Sure, caveat emptor and all that, but that's $42billion not invested in other ventures more likely to succeed. This kind of thing keeps happening, and it's money not leading to economic growth. It's become too easy to fool investors with techno glitz. The tech industry has got a few such tech frauds on its hands, and sooner or later governments are going have to intervene if the markets to keep the crazy ideas at bay.

Glad I've not wasted any time or money on it.

[StrangerHereMyself]

Prediction

Musk's prediction is aimed at LiDar but I'm convinced the whole Autonomous Driving sector will collapse in the next couple of years. Unless an enormous breakthrough presents itself (like: a general intelligence that can drive at least as well as a human, under all circumstances).

[Sorry that handle is already taken.]

Re: Prediction

Time was, Ford believed it would have a fleet of MAAS AVs operating in 2017...

Back then, in the mid-2010s, I was very optimistic about the potential of the technology. I'm not so optimistic any more, as increasing numbers of critical problems turn out to be intractable.

If you were to graph hype, promises and predictions about (C)AV adoption over time, it would probably look like that famous graph of Itanic sales forecasts over time.

[StrangerHereMyself]

Re: Prediction

Quantum computers and nuclear fusion will be taking the same route.

[badflorist]

Re: Prediction

LiDAR can do it but it is a crazy amount of resources. You can find point clouds right now that are good enough, especially if 2 LiDAR devices are used but, it's MASSIVE compute and memory requirements. Like way back when a computer could do your taxes but... you couldn't fit the computer in your house :-/

I can't find it but some biologist or scientist used it to map what was essentially an Ant hill. 1 device mapped the hill and general ants but when it was time for the 2nd device to map the activities of the individual ants, the device that could do it was basically a mini supercomputer... so the project ended there.

[StrangerHereMyself]

Re: Prediction

Yesterday I was driving around town and a few scattered leaves blew across the road. I considered how an AI would manage navigating these and the following thoughts came to mind: its "sensitivity"would be lowered by its engineers to avoid the car from stopping or braking for every leaf that crossed its path, resulting in the fatal accident in Las Vegas where a woman was run over by an AV (autonomous vehicle) or it would brake or swerve every time it picked up a leaf on its LiDAR.

To me this seems an unworkable situation because these AV's have no real sense of the world (including the physics of leaves, leaflets or plastic bottles that blow across roads).

[badflorist]

Re: Prediction

There's also the size of your leaves that matter. For instance, what if car is trained to ignore leaves but, someone dresses up in a giant leaf costume for Halloween and decides to cross the street? tin_foil_hat_on()... what if someone reprogrammed/hacked the image recognition and the car started targeting people that looked like Elvis? Sounds crazy but programming it to target humans is equally as difficult as to dodge them (it might even be easier :-/ ).

[StrangerHereMyself]

Re: Prediction

To me it seems to motion physics also come into play. Any human can estimate the way the leaf moves (taking into account the wind speed) that it's weight and impact on the vehicle is negligible. Also, since it's not a living being we can ignore it as well.

No AV takes these things into account. It's either ignored or seen as a hazard, mostly based on pattern recognition in a neural network.

[Binraider]

Considering that we won't allow self-driving trains in most places, in a (relatively) controlled environment that is a railway, self driving cars are just pie in the sky.

A box with two axles, motor and battery; fine. Loading it up with thousands of highly sensitive sensors and silicon to supplement or replace a human? Not so fine, not least because of the resulting cost of the product, and complexity in maintaining it all doesn't make sense for the majority of owners.

Get back to me when you can come up with cheap sensors and general purpose AI that doesn't need a shedload of power to run. EV is fine. Tesla is not fine.

[2342354]

not quite

The article seems to be a little ethnocentric in its conclusions that Mr Musk's views on LiDar are dispositive. On the one hand no surprise that immature pre-profit companies ipo'ed during the height of a massive bubble would see their valuations crushed... on the other hand there are several mid to premium EVs in China, as well as some luxury cars in Europe that do ship with LiDar sensors on board. Incidentally some Chinese robo vacuums also have them...

Then, the scaling of high precision maps has also been nicely solved by swarm based data acquisition technology such as used by MobileEye.... so this article has a whiff of one-sided Musk fanboyism to it that seems unbecoming of the Register.