SoftBank’s next bet: $940M into autonomous delivery startup Nuro
techcrunch.com
techcrunch.com
They said that softbank has almost had a "chilling effect" on their VC group - especially around investments that are truly capital intensive.
Their strategy has shifted away from Series A and towards Series B / C where they can either co-invest with SoftBank, or if wait and see where Softbank invests before they pick a horse in a category they like so they at least have that info in as part of their investment calculus.
EDITED last paragraph for clarity
I mean, softbank investment is competitive advantage in itself, just because of it's sheer amount.
Yes. If they have a good idea, large market, great product, good marketing and outstanding team.
If I could only make a few big or medium size investments I'd probably follow the market leaders as well.
For instance, if you invest in 200 startups at $5m each, you’re more likely to get better returns, than betting almost $1b on a single company. There’s higher risk with the one company (probably).
Further, and to the point on efficiency. There is diminishing returns to the effectiveness of investment, in most cases. You can get to MVP for most companies with very little, if any capital. Then the product in most cases should sell itself (needing less investment).
At $1b you’re either artificially propping the market, blundering around for an idea, or have a high capital intensive industry like building rockets or railroad tracks.
I understand they may be building autonomous vehicles, but certainly the winner of that race is going to Win using technology, not manufacturing the vehicle. Perhaps I’m wrong, idk.
It's also a lot less likely to have outsized returns (50x?), but at those levels of capital, almost no investment is able to.
Just fyi though: I have worked in markets for about ten years...I have seen this over and over. I studied economic history...I have read about this over and over. And the subject almost never makes it out alive. Usually, this is psychological. Most people get involved with clear heads but that clarity doesn't survive the rush of the bull market/greed...they want more, they start to believe their own bullshit, they end up fully invested at the top...they get carried out of the market.
The overhead of sourcing and dealing with 200 investments is not small: they've only managed to invest $50 billion of their fund. Their minimum check size is $100 million. [1]
[1] https://www.axios.com/softbank-vision-fund-1517515371-c269cc...
It's Inadequate Equilibria by Eliezer Yudkowsky: https://equilibriabook.com/inadequacy-and-modesty/
Yudkowsky agreed with a small group of economists, and disagreed with the Japanese equivalent of the Federal Reserve. You might think, "The Japanese Federal Reserve has every incentive and resource to be correct about growth policies, and this small group of economists has no insight they don't have," but the story is more complicated than that.
That doesn't apply here.
Of course, an appeal to authority is not a valid method of obtaining new knowledge (an authority would be able to formulate the actual arguments that support it). It is a useful method to defer judgment, however, such as when the authority cannot be expected to come in and explain their reasoning in a Hacker News comment.
I have been through it myself more times than I care to mention. Everyone assumes that the other person has some knowledge they don't know about or has some secret "edge"...they never do.
Amd after ten years or so, I learned that people do weird things for one simple reason: they are fucking stupid. That is it. They won't learn from mistakes. They are overconfident. They are greedy. Intelligence doesn't matter. Money doesn't matter. Whatever, they are just incurably dumb.
And btw: organisations are never well optimised for stuff that matters. Finance is the prime example of this. If I wanted to set fire to $100bn, I would do exactly what SoftBank are doing right now.
I know that everyone here believes in technology and continuing progress into infinity. This is true in some areas, it is not true in finance.
It reminds me of a response I got in a theological debate: “you really think you’re smarter than pastor ABC?” No, I simply agree with the pastors that disagree with him!
SVF is a vehicle to deal with excess liquidity from central banks, who have pushed everyone out on the yield curve by flattening the yield curve. The primary way to flatten the yield curve is by purchasing short to long dated government/municipal bonds, and short to long dated corporate bonds, until none of that looks like an attractive investment (overbidding on bonds makes their ROI lower). These purchases are done with newly created money, which is now in the hands of the prior owners (financial institutions) and also the issuers (governments, mostly investment grade corporations).
Because society itself is not comfortable purchasing riskier assets, at the speed of which the addressable investment landscape demands it, SVF collects this capital and deploys it in riskier private equity. Keeping the money in circulation better than without SVF existing. This behavior is the point of why the central banks dampened the yield curve with new money, to force people to make the money circulate instead of sitting so passively in bank accounts or holding government bonds.
SVF is a reflection of reality, which required the vision of independent people and the assertiveness to form the deals and do it. This is completely independent of acumen deploying large amounts of capital, and should be seen as a natural progression of capital formation. SVF's shouldn't be seen as an outlier in deal size 10 years from now, or 20 years from now, if the money supply is able to continue its expansion at a similar same pace. People will get used to it, ideally wages increase or the benefits of the deal flow reaches a broader population.
I don't know how much it takes to build a safe autonomous car - and nobody else does either as it is uncharted territory. Google and a few others have made progress but they have limits which they are still trying to figure out.
There are fewer entities that can place bets like that, which means that it is unexplored and likely has more opportunities. Seed funds with $20mm are a dime a dozen.
With this being said the failures are much more catastrophic when such investments are done blindly, without fundamental market / technology forces (e.g take a look at Theranos or Better Place). I believe this is what you are alluding to.
Good investments in this $100mm + category are precisely justified in terms of demand and precisely justified in proof of technology execution -- they are largely for scaling technology to fit a known demand.
A fundamental paradigm shift happened in the past decade which Peter Thiel describes as "moving from the world of bits to the world of atoms" ... this necessitates larger capex to win. With software the scaling costs are constant, with atoms -- they become linear.
This paradigm shift was driven by mobile technology and AI ... so the world is completely different from the days when spreading 10 $1mm bets was a solid strategy. One of the fundamental problems with applying past intelligence is that it was built on top of a different world.
There's a bunch of these things. Starship Technologies has a little stroller-sized vehicle that's occasionally seen driving on Redwood City sidewalks. Nuro's is sub-car sized and can operate on roads. Slowly. It can't unload itself, so the customer has to go to it.
Remember Amazon Prime Air, drone delivery from five years ago? What happened with that?
You might be right, but you should have pretty high uncertainty about that claim.
If driverless tech reduces the delivery cost down 10x, the average consumer might be very willing to walk out to the street. Some might even do it in the rain (umbrella, etc).
I'm not trying to fault the driver for not locating my unit, not do I enjoy sounding like an entitled bourgeois snoot, but it definitely makes the delivery experience less-than-complete. I imagine the last 20-meter problem will be a much higher barrier for those who are experiencing mobility-related challenges, and it is indeed a real problem to be solved.
The biggest challenge I see is discovering and navigating to the appropriate place for the vehicle to wait for a customer to come down and pick up their order. Just thinking about my building, there’s a turnaround with a ton of non-obvious cues as to where taxis and Ubers are supposed to wait, and it’s shared with a public parking garage and a restaurant with a valet stand. An AV is pretty much guaranteed to get it wrong, and Nuro needs a way for door staff or valets to tell it where to go.
except pedestrians and other drivers. so, in fact you do.
If I were Nuro, and if it isn't already factored into the design, I would purposefully design the vehicle so that the Nuro would be squished and minimize likelihood any other vehicle involved in a crash would sustain too much damage.
Yes, in the same degree. It is a street vehicle. They even published their safety manifesto some months ago: https://static1.squarespace.com/static/57bcb0e02994ca36c2ee7...
However you may meet standards and regulations, and pass all certifications, but the issues about autonomous vehicles present big unknowns. We already have accidents and fatalities related to use or misuse of UAVs. Even small sidewalk delivery vehicles may injure or kill a baby in a stroller because of incorrect detection or malfunction, or start a major accident when other drivers need to swerve to avoid them because they entered a road or stopped without notice.
Nuro is testing in Arizona like other self-driving vehicles companies. Some people are getting increasingly annoyed dealing with their behaviors and have started retaliating. With Waymo's cars, for example: https://www.azcentral.com/story/money/business/tech/2018/12/...
I guess I must really be missing the point of all this, as a bunch of people seem to want to throw a lot of money and effort at the 'problem'.
Once you create a taxi or delivery service where you have no labor cost, then your unit economics greatly improve and you have a killer business.
If you have a $10 Uber ride, ~$7.50 or so of that goes to the driver with $2.50 to Uber. With an autonomous taxi model Uber could theoretically halve fares, in this case to $5, and take in the entire $5 for themselves.
The general point is that the savings might not be enough to offset the higher capital costs from autonomous vehicles + development spend.
Where I live the streets are covered with snow and ice, and many paths and routes are difficult to navigate even on foot. No delivery vehicle is going to cope with this.
Is anyone else getting really tired of hearing this kind of rhetoric? This is a company that, if successful in its mission, is going to help eliminate a whole bunch of jobs (delivery drivers, auxiliary grocery chain workers, etc). Yes, it's gonna provide a cool upside - I can have robots deliver my toilet paper to me without having to leave the comfort of my home! But do we really need to describe that as "fundamentally improving our daily lives"?
On the other hand I'm sceptical that any company can compete with Waymo's experience or Tesla's amount of data.
Tesla's business model of giving away sensors for free gave it an enormous advantage in acquiring data (which is crucial for deep learning ).
- It saves you time.
- It reduces your need for expensive appliances and electric bills.
- It reduces your need for an expensive car.
- It reduces food waste/spoilage.
- It allows cheaper access to delivery for all, especially the disabled/elderly.
Consumers don't care one bit that their groceries were delivered 'autonomously'.
The 'advantage' of the 'self driving' in the end must be cost - and frankly a significant reduction in cost - or it won't pan out.
In this scenario, there still has to be a picker, a checkout, a loader, and some human overhead on managing a live fleet of cars. The 'savings' will be on the cost of the driver - and that's it. There's also a deficit in experience related to the fact it might not work well for apartment buildings and can't unload etc..
After all the fuss, ops and overhead, this tech has to take the average grocery+delivery bill down by a fairly quantifiable amount or else it won't make sense to use it.
It will happen eventually, it must may not be this company at this valuation.
With Instacart, you as a grocery customer decide it’s worth a significant premium to have groceries delivered to your door, with full service. With Nuro, the initial customer is the grocery store itself; to you as the grocery customer, it’s stores with Nuro versus stores without Nuro.
With a service like Nuro, the stores probably hire an extra person or two per shift so some of the floor staff can be assigned picking and loading, spreading out the labor costs. As you say, getting the cost down is critical, and if Nuro succeeds, I could imagine autonomous delivery eventually being priced into part of the standard operating costs for a grocery store.
Innovation and automation lead to jobs being lost that were based on repetitive, unskilled tasks. However that has been going on for thousands of years and it puts to the burden on the labor force to gain more skills. We can go down the hole of UBI, leisure, specialization, etc. but we shouldn't stop the pace of innovation just because we are worried if society can pick up the slack.
They've launched a PR stunt, not an unmanned delivery service.
[1] https://techcrunch.com/2019/02/07/amazon-sequoia-invest-in-s...
If we take just barely good enough as the status quo we can expect over the next several years, then Nuro is well positioned. You can actually build a revenue generating business with just barely good enough autonomy doing unmanned delivery.
Aurora is good because they acknowledge that building capable autonomous vehicles is still an applied science problem. There is still invention that needs to be done.
> I am one of the founding team members of the Google self-driving car project.
https://www.linkedin.com/in/jiajun/
> I built and then led the computer vision, machine learning, behavior prediction, and scene understanding teams for Google's self-driving car project.
I'm assuming you'd have to be home in order get them and there would be an app that tells you an estimated delivery time and location of the vehicle.
Edit: People will steal anything if the risk/effort/reward numbers make sense. People commit felony crimes against pizza guys for $50 and a few pizza all the time. I can see people stealing groceries if the risk and effort dips low enough.
FWIW I used to work at a Softbank funded company. It was a fucking shitshow where the solution to every problem was to throw money at it, rather than actually try to fix the underlying problem. People are still writing articles about how this company is the future.
Lucky for me I have nothing invested and can just enjoy the show from the sidelines!
It's one thing for FB to buy a massive social network for billions: they can effectively calculate how much it's worth because they know how it would monetize.
But this is getting bonkers because there are so, so many things that can go wrong. Do consumers give a rats if their groceries are delivered by a person or by a robot? Will they respond well to the fact they have to walk outside? How do they deliver to apartments? Regulations? Safety? Theft? Service costs? Operational performance? And all of this assumes the 'self driving' part just 'works'.
I get the market is 'monstrous' ... but this is a lot of money one would expect when all of the above is taken care of.
Of course, they will. When savings will be passed to you.
https://www.reuters.com/article/us-nuro-funding-softbank/dri...