Hacker News Reader

Comment by CM30 | Hacker News Reader

I don't think this is all that surprising to be honest. Google is (by and large) a software development and services company, not a car manufacturer. If they make their own autonomous vehicles, that means they then have to handle everything from safety testing to customer support and sales in addition to the self driving car technology itself.

By working with existing car manufacturers, they can focus on their part of the deal (the technology) and let someone who knows about designing, building and selling cars do the rest of the job.

View on HN

Working on self-driving cars, much like working on advanced robots, requires tight integration between the software, sensors, and motors. These aspects cannot really be developed independently, especially if you want to end up with an optimized system that could be a product.

Safety is a part of this, too.

Google has ~10 years of experience in self-driving cars. They also have acquired many robotics companies. Surely, they know this.

[edit: reorder]

But the reality of the product/market is that it will be a transitionary stage between assisted self-driving cars -> full robotic cars. There have been a number of complex highly technical products that have components developed independently and integrated at the end (fighter jets, space shuttles, etc).

It's possible Google will want to acquire one of the car companies to full do vertical integration. But that is not entirely necessary right now.

And given the whole complex regulatory and distribution constraints in the marketplace for automobiles it's good to have a partnership with a car company who has years of experience dealing with those issues.

> Google has ~10 years of experience in self-driving cars. They also have acquired many robotics companies. Surely, they know this.

Really I think Google loses steam very easily when key people leave the team. I don't believe they have the mgmt infrastructure to keep the ball rolling when people get bored, run into hurdles, etc....

20% time and individual project ownership at Google X have helped to make Google a famous employer, but I do get the sense that the drawbacks are becoming more visible as the company grows and people move on.

Some of this is undoubtedly belt-tightening, but it's a lot easier to ax a specific project when it had a few fervent advocates who are no longer around. I'd be surprised if this wasn't a factor in the high project churn Google is known for.

I agree this move isn't surprising, but disagree with the assertion that Google is a software development and services company.

Google is first and foremost an Internet Search Advertising company, so far it has been the thing keeping the lights on. I expect that if it had felt it could make the transition into self driving car company it would have. But it helps to remember a little bit of history here.

Google was, for the most part, the very first company of significant size to invest in self driving cars. They did that at the behest of Sebastian Thrun who was working there. It made some big headlines and it certainly added to Google's "glow" as a place where the future was happening. And then it ran into the sludge that is Google's inability to actually make "products" that aren't a web page. And as it did people started slipping away, first Sebastian, then Chris, then a big exit to what was Otto, Etc.

While Google was sitting in that quagmire of their inability to execute on a product vision, Tesla came in, announced auto pilot, deployed autopilot, updated it to do many of the things Google could actually do, announced self drive, deployed the hardware for self drive. Basically all executing on a product strategy while Google argued internally about whether or not their "car" should have a steering wheel or not.

What I've read in the news so far it sounds like Google is spinning off what is left of that group as Waymo (that seems like it is a sort of PIP for groups at Google) and trying to sell it off to someone.

I left in 2010 and the self driving car group was about a year old (and I tried to join it but they turned me down :-() Here it is 2016 so 7 years of development. A number of actual vehicles were produced, a number of cars were modified, I'm guessing there were about 150 people in the group at its peak. Just doing back of the envelope conversations for that group I'm guessing Google invested at least $300M in the effort. And now its scrap.

I appreciate that it is a really hard problem, and I appreciate that it was pretty far off Google's expertise spectrum. I also appreciate that they take "big bets" even when they fail. But given where Tesla is, and where they are, I worry that even if one of their big bets is a "winner", Google won't be able to execute on it. Just like Xerox couldn't execute on the personal computer and GM on electric cars. Once you get to be too big to win, what do you do?

Yes, perhaps, but the counter to this is that they could "benefit from the closed system of vertical integration" that seems to be all the rage these days. I could envision a scenario where they partner with existing manufacturers for a couple years, then "courageously" break off and produce their own vehicle in order to "control the process".

That would be one hell of an ecosystem to bootstrap, even starting with a partnership with an existing company, even for Google. Building "a" car isn't that hard. Building a car that conforms with all regulations, has had all the testing done on it, can be manufactured at any useful scale, and, oh, also is something that people want to buy, is not a problem that should be underestimated.

(Yes, props to Tesla. They did it. Well... mostly, at least, but it's at least a reasonable guess they're on a success trajectory. It's still a hard problem.)

It would probably be easier for Google to manufacture their own computing hardware from scratch, and as far as I know, they don't do that. Yes, they get it customized because they are a big customer, and they've at least dabbled in phone hardware, but even that wasn't "from scratch". I've got a 2013 Nexus tablet, and it's got all sorts of things in it that aren't from Google.

As big as Google may be, they are not infinitely big, and they have the paradoxical problem of being too successful to take on every little business. Large profitable businesses can't economically afford to take on new businesses that are significantly less profitable than their current businesses, because the rational business decision then is to reallocate the resources being given to the unprofitable sub-business and give them to the profitable sub-business to make more profits. [1] The car industry is very hit-driven due to the need to sell enough cars to overcome massive fixed costs, but even at in their most profitable years I'm not sure the car companies came even close to Google's profitability. I'm not taking the time to look that claim up, because I'm sure that even if they did, they never did it year after year after year the way Google has.

[1] Obviously I'm simplifying a complex thing here. I understand that's a complicated topic, but details would be an unimportant-to-my-point waste of time for all.

Large profitable businesses can't economically afford to take on new businesses that are significantly less profitable than their current businesses

Whoa there, isn't that completely backwards? Who can afford to invest in new businesses if not large, profitable companies?

because the rational business decision then is to reallocate the resources being given to the unprofitable sub-business and give them to the profitable sub-business to make more profits. [...] Obviously I'm simplifying a complex thing here. I understand that's a complicated topic, but details would be an unimportant-to-my-point waste of time for all.

(Edit to add: I think that apparent paradox is easily resolved by thinking in terms of long-term versus short-term investment.)

I agree there are complications, but fundamentally, big profitable companies are the ones which have spare resources to invest.

One problem is company culture and skill sets, and the difficulty of ramping up in a new area, but that's why big tech companies try things like internal "startup incubators" with lots of autonomy. Amazon in particular seems really effective at expanding into new areas.

It's not clear to me that Amazon actually has one business line that is both huge and so massively profitable that it prevents them from spreading out. They are huge, but they don't make the profits that Google, Microsoft, or Apple does.

I wouldn't be surprised to see AWS get spun out as it continues to grow in the next 3-10 years, because unless the trajectory of that subbusiness changes, it threatens to become the business that is too profitable to put resources anywhere else.

But Amazon has explicitly chosen to keep its margins tight, and aggressively invest for expansion. In a sense they choose not to make huge profits. They're famous for it!

I still don't buy your general point. It's true that some large companies have run into trouble because they were afraid of damaging their core product. But that's a risk everyone's aware of now -- the whole "innovator's dilemma" thing.

What's an example of a large successful company that was thereby prevented from spreading out? I don't mean companies that failed due to complacency, or overly rigid business practices, but where their size actually worked against them.

(Hmm, possibly we really mean the same thing, we're just describing it in different ways...)

> Building "a" car isn't that hard.

Tesla deserves plenty of accolades just for getting to the point where they are producing multiple models of vehicle that conform to or exceed modern standards of quality .

However, Tthey have only managed to build "a" luxury sedan, "a" crossover SUV, and will soon be building "a" budget-luxury sedan. The real challenge is in attaining and sustaining the pace of development maintained by the rest of the industry. The automotive industry at large keeps a rolling cycle of refreshes and replacements on a roughly 3/6 year timetable (refreshed after 3 years, replaced after 6). Some niche models (Jeep Wrangler, Mazda Miata) stay in production far longer than that average but a 3/6 cycle is also a bit too long in ultra-competitive categories such as the compact crossover segment.

This is where I see Tesla beginning to struggle as their entirely bootstrapped operation has struggled to meet deadlines. That is cause for concern as it places their current lineup at a competitive disadvantage as their competition has gone through at least 1 refresh, if not a full replacement, since the design of their most recent model (Model X) was set in stone.

The Model S design was largely set in stone by 2009, 2010 at the latest and the Model X design was mostly locked in by 2012, 2013 at the latest. It took 3 years for both of those designs to actually reach showroom floors (2012 and 2016 respectively) so while the clock hasn't run out in terms of production life, the designs themselves are beginning to become dates. I don't just mean "design" as in styling, I mean it as the all-encompassing architecture of the vehicles.

While Tesla's offerings have several unique attributes which cannot be found in any of their closest competitors, one can only push an old design for so long before it becomes completely unappealing to consumers, even when sold at break even prices. I dare say that such a scenario may be playing out with the Model S as its sales had slumped leading up to the 2016 refresh that boosted sales, yet the bump still fell short of that model's best sales quarter which occurred back in 2015. The fact is that the Model S was designed to achieve the minimum level of refinement expected of a 2010-era, $65,000 luxury sedan with the silent, powerful electric powertrain being the plan to make up for its shortcomings in the areas of ride quality and interior fit and finish. Now that ~7 years have passed, the high end, mid-size sedan offerings from mass market brands have caught up to the Model S in many areas.

The Model S may smoke a Nissan Maxima in straight line and the Maxima can't hold a candle to the silence of the Tesla's powertrain but it also costs ~$35,000. In my opinion, it also has a better finished interior and more up-to-date styling than the Model S. We can argue all day over these points but you have to admit that manufacturers like Nissan are at least getting within striking distance in certain areas with cars costing half the price. Tesla needs to seriously step up their development game lest they be left behind with woefully dates products that simply cannot be discounted enough to remain appealing.

You aren't simplifying, you're misrepresenting. A company wants to invest in projects with an expected return exceeding it's cost of capital for the project. It doesn't matter if you lower your overall rate of return.

To give a concrete example, imagine I have a hotel in New York generating profits of $50 mn/year. If it required an investment of $100 mn to build hotels, making another one in Atlanta that generates profits of $10 mn per year would still be sensible if these exceed the financing costs. My return on (book) assets would fall from 50% - 30% but my profits would rise from $45-50 mn.

Tax, risk and use of management time can complicate the issue but your point isn't first-order correct...

Your example doesn't apply because my point applies to different businesses, not just "investments" in general. A hotelier buying or building another hotel is just expanding, not investing in different businesses. A hotelier that is making 10% RoI in the hotel industry would be ill-advised to take on a car rental business that makes 3% RoI, because that car rental business won't be free, and they should be investing that money where they can make 10% RoI.

Google, paradoxically, by having such a good RoI on ads has a hard time taking on other businesses. This is a big part of why they spun so much stuff off; if they stay within Google, business forces would tend to starve them despite the abundance of resources that at first glance seem to be available. It's why you see companies so often cut product lines that are profitable, but not as profitable as something else... or, to put that another way, this isn't something I'm hypothesizing about how it might happen because of the theory, it's a thing that happens all the time. I've seen at least two major passes of it at the company I work at.

Ok so if a business a) has limited management bandwidth and many promising options or b) has a particularly cyclical/risky business model that results in a high cost of borrowing then you're right.

These are somewhat linked to profitability as you say, but a safe slow-growth company generating massive profits (eg Coca-Cola) doesn't face this.

I've got a 2013 Nexus tablet, and it's got all sorts of things in it that aren't from Google.

The majority of components in a car are not manufactured by the car maker.

Google will want ad revenue from the captive audience in the self driving car. Control the self driving car ecosystem in the long term and avoid the cost of manufacturing and compliance in the short. I like the assumption that they will "come back" into the space, because it's a great play.

I have the same response to people who claim Tesla will take over ridehsaring once they have a self-driving car (as opposed to just selling to existing ones that have already built a customer base and worked out the domain specific problems).

Because clearly Uber would succeed over a manufacturer who actually makes self driving car. Aren't there already almost 8-10 ridesharing options available in Austin after Uber and Lyft vacated?

Uber and Lyft are the WebVan of this decade.

Just like GM prefers to run taxi services rather than just focus on making cars and selling them to th highest bidders (including taxi companies)?

Uber profit margin: 0.

GM/Ford Profit Margin: 9-12%

Tesla Profit Margin: 22.8%

Go where the margin is.

Which, by those very figures, would be in selling cars to ridesharing services, not competing with them.

Why would Musk and Tesla sell cars to ridesharing services when they can collect a piece of the revenue that they share with Tesla owners who participate on the Tesla Network rideshare service? Why should Tesla allow Uber or Lyft to profit off of its extensive autopilot work? They shouldn't, and they won't.

The same reason GM lets car rental agencies "profit off" its development of new features: power seats, Onstar, media center, voice nav, better steering, heated seats, etc.

Which is to say, it doesn't, if you measure it right: they charge as much as they can for a car with features x/y/z; once you impute out who is making how much profit from what, you find that GM is profiting from the Onstar, and Avis is profiting from the rental service (matching cars with temporary-use customers). To the extent that Avis can rent cars at a premium with Onstar, that profit is mostly eaten by the extra price.

At no point does GM think, "Merely sell our cars to rental agencies? There go the Onstar profits!"

In just the same way, if Tesla sold cars to Uber/Lyft, they would still be the ones profiting from the SDC premium.

As for Tesla owners participating in the rideshare network: they can probably get a lot steadier renting of the car's off-hours if they have access to an established platform rather than waiting for riders to add another app that only gets rides from the few spare Teslas in the area.

> As for Tesla owners participating in the rideshare network: they can probably get a lot steadier renting of the car's off-hours if they have access to an established platform rather than waiting for riders to add another app that only gets rides from the few spare Teslas in the area.

I could go out today and borrow ~$100K to purchase vehicles and put them on Uber, but I'm not; I'm waiting for Tesla network to leverage up.

That is Tesla's stated intent, so it isn't much of a leap to think they'll try to do that.

I suspect Google got cold feet when they started considering the liability issues once this goes mainstream. Even their millions of miles road tests were done at low speeds near their campus.

They're focused on moonshots. They would have managed the liability like any other company for the prospect of transforming transportation as we know it.

A lot of Google's testing was done on US 101 and other interstates at highway speeds, insomuch as there wasn't a traffic jam at the time.

Highspeed testing is important. Also the complexity of the traffic is. Some say that the high speed of the German Autobahn is what made German premium cars so technically advanced. As soon as you get faster than 130-150 km/h most cars will let you know they weren't built for this. BMW, Daimler or Audi make 230 km/h ++ feel safe and sound (which I often think they shouldn't). A Tesla for example can't perform autonomous driving here as easily because the variances in speed require sensors that detect cars in further distance, quicker and in more detail.

A quite ride at speed only really means it's a luxury car. Toyota, Honda, Nissan, Ford, et. al have had plenty of experience making cars that can go fast, and they also make luxury cars.

In any passenger car, getting in a wreck at 100 km/h or above is more than likely a fatal event. "5 Star" safety ratings and the EU equivalent are only for 30~55 km/h crashes. Granted, even highway crashes involve some braking before impact, but the faster you go the more difficult it is to slow down enough to a 'safe' speed before impact.

Besides, the sensors on cars are more than capable of driving at 150 KM/H. While they cannot "see" as far as a human, they require much less time to process what they do see.

> Besides, the sensors on cars are more than capable of driving at 150 KM/H. While they cannot "see" as far as a human, they require much less time to process what they do see.

And then you get something like the adaptive cruise control on my BMW.

The bloody thing controls like a human just learning to drive - only focusing on what's right in front of it, instead of the general road situation, and making correspondingly jerky and drastic (albeit safe) maneuvers.

I often have to look ahead for it, and turn it off when I see a situation brewing that will entice it to drive like a moron. Then I handle that situation (typically by doing essentially nothing, e.g. just coasting for a few seconds), and turn it back on.

From what I understand their vehicle was limited to 25mph, how did they do the highway speed testing?

Google uses Lexus cars for highway-speed testing.

That's the Koala-type. The first iteration were modified Lexus', and self-driving in highways is much easier than in cities, so it's attacked first.

I suspect Google never planned to be a car seller, they had their own vehicle because they wanted less baggage and friction while de veloping/proving the technology. Google will occasionally make concrete products where they see it beneficial to their software-baaed service focus, but they really don't seem to want to be in that kind of business except peripherally.

Just like the did with smartphones. Maybe that's why most Android phones aren't that good.

Galaxy series phones are fantastic.

at exploding

The Galanxy Note 7 was the one with the problem, the flagship phablet, not the flagship phone.

Fascinating. Neither of my Edge 7s have blown up yet. Should I request a refund?

> Maybe that's why most Android phones aren't that good.

I'm assuming you're talking about the phones themselves and not Android as an OS, but your statement is pretty sketchy. I'm no expert, but most Android phones that I'm aware of in the same price range as the iPhone are up-to-par or have superior hardware. Many carriers sell cheap Android phones as well, that are "not as good" because iPhones are expensive. So yeah, 90% of Android phones may be "not as good" as an iPhone, but Apple isn't selling new devices to the under-$500 market, and of course the cheaper phones aren't as good.

The problem is few car companies want to partner with Google because Google asks to collect too much data from all of their customers - not that I blame the car makers.

Do you work in the car industry? This seems like a very inside thing to both know and casually drop on Hacker News.

car companies don't particularly seem to mind data collection if you look at newer "connected" features.

Reply on news.ycombinator.com