How SoftBank’s costly bet on the ‘internet of things’ backfired at Arm | Hacker News Reader

How SoftBank’s costly bet on the ‘internet of things’ backfired at Arm

ft.com

60 pointshacman86 comments

Can anyone tell me an example where Softbank has been spectacularly correct in the past few years?

We all know the failures - are there any major success in their portfolio? Or is it all meh?

Pretty much a bunch of tech-illiterate people buying the latest buzzword without due diligence. Lighting money on fire. Even the successes were bought at the all time high prices (Bytedance) or mismanaged (Arm) and brought negative returns.

Take a look at these comical slides from Softbank back when they took over WeWork after it collapsed: https://news.crunchbase.com/news/the-best-slides-from-softba...

I honestly can't tell how the Softbank vision fund is doing. News articles on it report record breaking profits and losses. The two vision funds were recently-ish reported as being worth $150b, but I can't tell what that translates to in annualized returns, or how that looks relative to the NASDAQ.

The CEO Masayoshi Son lost spectacularly during the dot com crash.

https://en.wikipedia.org/wiki/Masayoshi_Son

> s of October 2021, Bloomberg Billionaires Index estimated Son's net worth at US$23.1 billion, making him the second richest man in Japan and 68th richest person in the world,[1] despite having the distinction of losing the most money in history (approximately $70bn during the dot com crash of 2000).[4]

After that the big success came with investment in Ali Baba.

https://en.wikipedia.org/wiki/SoftBank_Group

> In October 1999, SoftBank became a holding company.[21] In 2000, SoftBank made its most successful investment – $20 million to a then-fledgling Chinese Internet venture called Alibaba.[22] This investment turned into $60 billion when Alibaba went public in September 2014.[23][24]

Depends on what you mean "past few years", and also VisionFund is huge as has a finger in a lot of pies. https://visionfund.com/portfolio

If we talk news worthy stuff, ByteDance could be the biggest name on that list ?

They are the main investor in goPuff which will win the instant needs market.

Really? You think gopuff is going to win that space? Why?

The “This Week in Startups” podcast from last week, episode 1378, interviewed the CEO.

They are hiring their own drivers and building their own delivery hubs, meaning products you order are coming from their hubs, not from CVS or Walmart. They have more control over efficiency.

They even acquired BevMo and are converting those into hubs.

Wasn't GoPuff in the news lately for spending tens of thousands of dollars per week on Instacart? The magical cycle of VC money...

I'm still skeptical that anyone will win that space. The core problem (instant delivery logistics in a dense urban area) is just too expensive for the average consumer to bear. Right now every player is VC subsidized, but what happens when that money dries up?

Why do you think it’s (eventually) too expensive? Do you think distribution is too expensive?

No, the labor is too expensive. Ultimately you are paying someone for a couple hours of work, plus spending on vehicle deprecation, gas, parking, storage, all to deliver someone a packet of chips.

The core problem is the whole business model is a solution looking for a problem. The number of people who actually need this is tiny, the rest are just being lazy and as soon as it gets expensive they’ll stop being lazy. The only way around this is to ‘disrupt’ local shops out of business which I’m sure is the plan and afaict that’s a net negative for everyone apart from them.

How do they scale demand for peak times with a fixed workforce? It seems trivially easy for say Uber to start hubs if they wanted to and they have a huge existing base.

But it comes down to execution and it will be interesting to see the space in another 5 years.

They own the hubs but the drivers are technically contractors like all the others.

Basically they won big with Alibaba and everything else is... not great.

So probably about the same as most VC companies.

Softbank stake in Alibaba is worth north of $80 billion.

https://visionfund.com/portfolio

ByteDance, Grab, Ola, Uber, Opendoor, Slack, DoorDash, Didi, Coupang were/are all great bets.

According to CrunchBase (https://news.crunchbase.com/news/softbank-vision-fund-strate...), as of March 31, 2021 Vision Fund 1 was worth $146.5 billion from $86.2 billion in initial investment. I don't know enough about the space to judge whether that is considered good enough or not.

Yahoo, Alibaba?

That did not happen "in the past few years"

Question was "are there any major success in their portfolio?", time period was not specified.

Maybe you want to read the comment again?

Sorry. Overlooked the first part of the question.

This is still underperforming the sp500 who went up 87% in the same 5 years.

It's not as simple a comparison. We don't know the annual return % since all the money wasn't invested on day 1 but at various different points over 5 years. And the majority of it is still tied up in companies that haven't exited yet, so the value is mostly theoretical.

Sp500 is meant to be the benchmark. This is very unfair to present worst numbers than the benchmark and present it as a win even if numbers can be off.

Vision Fund's minimum term is 12 years, and it will likely go on for a lot longer than that. Comparing its very early years to a stock market bull run is pointless. Like 40 out of 400-500 total investments have even seen an exit yet.

I mean, 90% of large cap investors have underperformed the sp500. It's not about optimal returns, it's about diversification and portfolio risk. The Saudi's want some of their money in things that aren't correlated with oil, and pretty much everything in the physical word is. Technology is one of the few things that may even be inversely correlated.

https://www.cnbc.com/2020/09/18/stock-picking-has-a-terrible...

That's assuming that the Saudis are being rational with their money. Given that they are currently trying to build a city that's a straight line[0], and a "shape based" floating octagon city[1], that assumption might not be correct. It's entirely likely that the real answer has less to do with Softbank's performance and their need to diversify, and more to do with MBS liking Masayoshi Son.

0 - https://www.architecturaldigest.com/story/saudi-arabia-build...

1 - https://markets.businessinsider.com/news/stocks/saudi-arabia...

The line one actually looks brilliant. It has the appeal of letting any new growth easily happen and keep transport options simple. Though most likely there will be uneven development even on this design.

It's genuinely the dumbest city design I have ever seen, and it makes transport an absolute nightmare assuming you're interacting with real technology and not magic.

Cities are generally round for a good reason. You put the important stuff in the middle and your average transit trip remains as low as possible. Straight lines do the opposite, as you guarantee that each new resident is in literally the worst possible position for transit possible, and you over tax the center’s transit in order to move people to the ends of the line.

Oh, and consider the cost for things like sewage and water. They’re a bit component of a city’s cost, and here they’re laid out as inefficiently as possible.

May be. But think not about the utilization or optimization for a moment and focus on the maintainability and buildability aspects. Something that looks this regular probably be both built and maintained through pure automation. I mean self driving is super simple in this layout (as they mention their transport is automated). Compare the same to the ever expanding cities that quickly run out of central space and end up having a premium center and poor suburbs.

Obviously line is not the only thing that can provide these benefits. How about a large square or circle with the center purely reserved for green space. Anything beats the shitty cities we have today where people stay outside because they cant afford the rent and have to travel towards the center for work.

Investors in Vision Fund likely have the goal of returns for this investment that are not similar to the S&P 500 or other public equity indices. I'd further bet that 15-20% annualized is considered a very good result for this part of those investors' portfolios.

sp500 has both better returns and better diversification so less overall risks. And I am sure Vision Fund has higher expenses ratio than most of sp500 indexes. The only ones winning here is the execs of the fund.

You aren't understanding how capital and risk allocation work at a portfolio level. The people putting money in the SoftBank funds ALREADY have hundreds of billions invested in public equities. They are looking to diversify their portfolio across different asset classes. Your comment isn't relevant.

maybe it's just me but the tone of this comment reads unnecessarily hostile

This is correct: an uncorrected return stream that underperforms market beta is often still highly desirable / sought after.

Probably meant *correlated

No I meant uncorrelated

so they'd rather plow it in a fund that buys options on exactly the equities that make up the index? or blow it on bets like oyo, wework et al? all the while expense ratio blowing through the roof.

My brother was a buy side banker at softbank, we all know the kind of debauchery that went on in there.

The pandemic killed a lot of good bets at Soft Bank. The rogue wave was not kind to anyone.

If you went 5 years back, I don’t think you would have predicted that sp500 will be yielding 87%. At that point in time, SoftBank would have looked pretty reasonable. So there is a survivorship bias when people compare with sp500.

Now, let’s look at next 5 years. Would you be willing to bet that sp500 continues its streak? Would all macro-economic coincidences and government actions continue in same manner for next 5 years?

> According to CrunchBase (https://news.crunchbase.com/news/softbank-vision-fund-strate...), as of March 31, 2021 Vision Fund 1 was worth $146.5 billion from $86.2 billion in initial investment. I don't know enough about the space to judge whether that is considered good enough or not.

Also Crunchbase is imperfect data so those numbers could be WAY off for all we know.

For growth stock ... doesn't seem very impressive.

But consider how big they are, probably not too bad?

you're welcome[1] [1] https://archive.is/miQXP

Indeed I am.

Whose welcome?

My, thank you!

What I find particularly interesting about the ARM acquisition is that it was always hard to see how it was going to continue to grow. The industry is littered with IP companies that get squished by their big customers (see also: Imagination Technologies). It's very difficult to keep a competitive advantage with IP because it'll always leak and people will always catch up, and since all your customers know that the marginal cost of your IP is 0, it's very difficult to extract marginal profit. So there's massive downward pressure on your margins, meanwhile, if your customer is using your IP at a massive scale, they'll either want to pay you a fixed cost for the IP or it becomes cheaper for them to either acquire you or poach all your engineers and replace you (which they naturally want to do anyway since they like owning the IP and having exclusive use). So the natural tendency in this market is for IP companies to pop up with a competitive advantage, grow by selling IP to larger businesses until they finally reach a scale where one of the bigger customers decide to eat their lunch. ARM is well past that point. The only real way to mitigate this is to vertically integrate and become your own biggest customer (like Nvidia) but even then you're going to face challenges (Large cloud providers moving to explore manufacturing their own chips, TPUs etc)

This just seems like many of Softbank's acquisitions where they went in to the deal with a really weird strategy for success and then 5 years later it turns out they aren't smarter than everyone else in the room. It's also not like ARM was some early stage start up at that point, they weren't taking some long bet knowing it had high risk.

lol the cortex a7 is still the go to low power arm cpu.

how a vast company with all the funding in the world can "bet on iot" but fail so miserably to make any improvement or change at all for so long is something i don't think I'll ever understand.

dont get me started on how impossible to purchase most chips are, how inaccessible/non-existant the docs are, how vendored the frak up the screwball drivers are. what a devolved old world 80's style trashfire of computing arm has been.

open source is finally building up real traction bringing light to this dark mushroom hole, Google is too exhausted & the public opinion wearing thin from unsupportable shortlived high tech devices, but what a slow slow painful march it's been dragging arm & such a broken industry into a respectable place.

The Cortex-M4 is the goto low-power CPU.

Depends on your audience. M4 is decidedly midrange to me. Modern low power would probably be an M0, to others it still means 8 bit AVR.

Normally I'd say M0+ is the king and offers good debug facilities over the M0 despite the slightly larger die size.

However, @rurban is probably correct purely because Ambiq is building their portfolio around the M4 (and M33), and they seem to be working hardest in pursuing and patenting IoT layouts/technologies that reduce energy usage.

I love AVR for teaching/learning, but the fact that you can* get a more powerful, larger memory Cortex-M at a much lower cost than, say, an Atmega... I don't know why anyone would choose AVR for a new design outside of niche uses.

(* - pre-2020)

I love it all. Cortex-M0, M3, M4, M7, and A7 all together span several full orders of magnitude in performance. Every one of them has their niche. Unfortunately for ARM, all of those designs are many years old and they are still nearly optimal in their niches. ARMv8-M parts are starting to show up, but they aren't exactly revolutionary compared to their predecessors.

they're all a decade old. it's absurd that we havent found any architectural wins.

i have no idea whether there'a just scant dew wins to be had, or whether this complete lack of innovation is a company asleep at the wheel.

do you also love the cortex-a53 for application processing? because it too is another dinosaur relic that we seemingly cant eacape from. a55 seems to have made a like 12% difference.

As an EE/embedded dev the M4s are approaching a mature technology for the bog standard MCU driven products I work on. You can get a 180MHz processor with 1MB of RAM and 2MB of flash. For reading sensors and doing some lightweight application layer stuff I don’t know what more I would even ask for. Current product is not power constrained so not paying much attention to that.

Cortex a7 is whats used on some raspberry pi boards, kind of overpowered for IoT I would think. Maybe as a hub device. Unless you go Juicero kind of cracy.

interesting how high this comment got then how hard it fell. +6 to +2. for my posts thats a notable swing

amusingly the critics are saying that a7 isnt the right choice, but are citing an equally aged, unmoving set of microcontrollers as what should be used. these chips at least uaually have docs, i feel like.

personally i think decent low power application processors are as much if not more of the challenge than microcontrollers to iot's developmemt. everything is a smart speaker, smart camera, has screens. watches run apps, clocks run apps. arm failing this industry miserably is quite the story. but yes, datacenters & the cloud are eattinf the world. tech is innovating ul, to bigger & bigger more expensive offerings, & the cut-throat consumer tech offerings are dying. it's a market intel bounced out of, amd commands a huge price in, and which sees nearly no wins. m3/m4+a7 forever and ever; maybe someday risc-v will shake things up.

> When Son spearheaded the $31bn purchase of Arm, he saw it as a wager on the future of the entire technology industry, which was crystallising at that time around the IoT concept. He proceeded to push the executive team firmly on the course to designing chips for this future of machine connectivity.

Son did not push anyone, ARM was IoT crazy before it was purchased: http://web.archive.org/web/20151006082751/http://www.arm.com...

In the end if Softbank did nothing, they would have alot more to riding the baba cloud to fortunes.

Makes you wonder if the work you're doing is at all productive or really is not.

Are there any opinions that harm has come to ARM while it's been under the ownership of SoftBank? What might have happened to them if they had remained independent all these years? Might SoftBank have been the perfect shelter all this while, or might ARM have ascended to even greater heights if they had gone it alone?

How it 'backfired', if it is going to do an IPO? Post IPO its market value could be much higher than acquisition price of $31B.

Failure to sell to Nvidia at 2X the acquisition price is not a failure, it was an honest attempt to make a profit.

Correct. Here "backfired" is the investment did not make 10x return

Well, 10x was never a target, they were happy with 2x return from Nvidia acquisition.

Absolutely wrong. They settled for the Nvidia acquisition after they saw things were not going as well as they planned after years and years. Now it looks like they will have to settle for even less.

Analysts do not expect ARM to IPO at over $32B (https://www.reuters.com/markets/asia/arm-ipo-marks-sober-end...). That is why Softbank agreed to the acquisition offer in the first place, otherwise they would have just gone for an IPO directly.

ARM will IPO well over $32B.

If you need to avoid the FT paywall, here is the story republished through the FT partnership with Arstechnica. HN does not allow this link to be submitted. Instead HN robots mark [dupe] and redirect to the paywall. https://arstechnica.com/gadgets/2022/02/how-softbanks-costly...

I always thought IoT was a stupid idea. From day 1 it sounded pointless and unprofitable.

It is not a stupid idea. Most companies cannot make too much money from designing/selling the hardware. On the other hand there is money to be made from the data collected from these IoTs and selling device management and updates as a service.

ARM is just not the right company to make that happen. They are fantastic at what they do but not software services, big data, etc

Do we really want every household item squirreling information back to the manufacturer about our personal habits? Exactly who stands to benefit from that? Consumers will lose a fortune and shun the technology.

Of course no one “wants” that. But let’s rewind 20 years ago and ask: “do we really want our telephones squirreling info back to the manufacturer about our personal habits?” Sounds equally absurd.

If it only has benefits for the maker, it sounds pretty stupid...

It's a great idea ... for industrial processes and machines. Unfortunately this is not where the hype is, nor where continual profits are. Industrial customers have a nasty habit (from the seller's perspective) of buying stable stuff that they can run for decades, which tends to rule out planned obsolescence based profits.

There are few to no compelling use cases for consumers creating a true network of internet connected things, because they don't have distributed physical things. If all your stuff is nearby, and isn't in large quantities, then the internet, or computer networking in general, is not the path of least resistance to control/monitor/utilize it to some greater application. Controlling X with your phone is novel at first, but its nearly always an incremental improvement on controlling X with physical interfaces. Its not a killer app by any definition. It's businesses that have lots of stuff in lots of places to track and control. In these scenarios, a microcontroller with a radio is not just an incrementally better solution competing with an existing alternative, but the only feasible solution.

I know, that's why I was talking about industrial use cases....

My sincerest apologies for agreeing with you and expanding on your points.

You can do IoT correct and you can do it wrong. Most decide that you need to pull out your phone or register a device before you can use it. No one wants to download and app to be able to operate a juicer. However a coffee machine for example could be extended with optional features not hindering the core functionality.

I contracted for an IoT company that built hardware and software for the manufacturing industry, it was very profitable. I agree that the consumer side “smart toaster” part is garbage though.

Energy efficiency in computing will absolutely be needed on 5 years as climate change worsens and arm is well placed to deliver.

This article is garbage.

Dude bought ARM at 31bn and was going to flip it for 66bn until anti-trust regulators intervened.

SoftBank did nothing wrong, other than accept the risk of regulatory intervention.

He pushed the company in the wrong direction. The company could've been worth a lot more if it hadn't been investing in IoT nonsense.

Sure, the valuation went from 33 to 66bn, but could've done a lot better. Just because you made money doesn't mean you didn't do something stupid.

This is only true in retrospect. When IoT was hot, everyone was eating up that story. A lot of people are still eating up IoT story because of rise in 5G.

Also, competing in server market is very very hard. If they had exclusively focused only on server market for past 5 years, they probably might still have been just minor player because of massive Intel legacy that needs to be overcome but no one knows how.

He made a bad decision in pushing ARM to chase IoT. Instead he could have pushed more in traditional areas - servers, data centers, laptops and desktops. Some of that did happen but more could have been done if they had not been distracted by IoT.

During the same time that ARM went from 31 to a possible 66 (dependent on a quite absurd acquirer). Nvidia went from 15Bn to 600Bn.

Reply on news.ycombinator.com