Y Combinator's Graham Doesn't See "Bubble" in Technology | Hacker News Reader

Y Combinator's Graham Doesn't See "Bubble" in Technology

youtube.com

163 pointsanswerly76 comments

My favorite line in that, after the interviewer asked PG what the big successes for YC were and he answered AirBnB and Dropbox:

Q: "But aren't you forgetting Heroku? They just sold for $220 million."

A: "Oh sure, Heroku was a success... but you couldn't buy Dropbox or AirBnB right now for $220 million."

A better question would have been "what are the big failures of YC" - smaller list to iterate through.

Not the most convincing spokesperson for that issue.

"CEOs of all major banks don't see a mortgage bubble!"

Sigh. I click on comments and the top comment is two line DH2 responding to the title of the post. Not HN at its best.

Actually I did say that valuations are high, and they'll probably fall in the future. But there is a difference between the ordinary rise and fall of prices and a bubble.

How would you define a bubble then?

The standard definition is "trade in high volumes at prices that are considerably at variance from intrinsic values".

"The valuations are high, and they'll probably fall in the future" implies (by the way I construe it) "considerably at variance from intrinsic values". It would be hard to argue that there is not high-volume trade in the tech industry.

Individual prices rising and falling due to changes in intrinsic value, for example, due to information about new competing technologies that could not have been anticipated before, or unanticipated information about things much of the tech industry is dependent on, like energy prices, can change the intrinsic valuations.

But prices rising above the intrinsic value because of high investor confidence and investment dollars from one company cycling around the industry and buoying up the existing players is a bubble. It might not be as big a bubble as the last bubble, but it is still a bubble.

You choose one of several definitions you find on the web and call it the "standard definition," as if that made it a sort of axiom. Then you interpret its vague "considerably" as synonymous with "at all." But then you're one step away from your own reductio ad absurdum, because you can now prove that any random (upward) price fluctuation is a bubble, and that is just not how people use the word. Stock prices rise and fall with investor confidence, but not every rise is a bubble. The term "bubble" is reserved for really extreme cases.

Incidentally, I've found that when the conversation in a forum descends to arguments about the meanings of the words being used, it never rises back up to talking about ideas. So if you don't mind, I'm done with this one.

> The term "bubble" is reserved for really extreme cases.

Maybe so, but it doesn't tend to get tossed around until cases start to get extreme. There are a lot of rumblings that there's a bubble building here. Can you think of any examples in recent history when the rumblings of a bubble turned out to be hogwash?

That said, recent history also suggest that the rumblings of bubbliciousness tend to precede the ultimate burst by a good couple of years and the lead up to the burst has been where the most frantic run up takes place, so when bubble rumblings start, that might actually be the best time to get into a market... :-)

In this case, I don't think talking about definition is inevitably useless.

Defining could be determining characteristics of a bubble and figuring out how to recognize one. It's consequences (bursting) are, I think, not under dispute.

One analogy is inflation vs hyperinflation.

Inflation is usually driven by money supply. Print more money or expand supply some other way and you get some inflation.

Hyperinflation is driven by hyperinflation. That is, prices are rising so fast, that they are forced even higher. People rush to buy goods, commodities to get rid of rapidly devaluing money as fast as possible. This drives up prices. Governments need to print more all the time because what they printed last month isn't enough this month.

Financial bubbles are similar and different to overpricing in a similar way. They drive themselves.

"The valuations are high, and they'll probably fall in the future" implies (by the way I construe it) "considerably at variance from intrinsic values."

At least for Angel rounds and series A, IMHO the valuations are finally where they should be. I agree that they will probably go down, but not because they are intrinsically overvalued; rather, I think the investors will eventually just come up with new ways to tip the balance of power back toward themselves.

There definitely are a bunch of intrinsically overvalued startups that are raising a few hundred thousand bucks in angel rounds, but I suspect that we'll just see a gradual return toward using metrics when investing rather than some huge collapse.

You did reject their premise and then agree with their facts.

To someone with a 20 second attention span (or someone spreading 20 seconds' attention over 5 minutes of interview) that sounds like talking your book. Soviet Union has no problems. There are no little green men.

I don't know what you're obliged to say or not say, but you might get better results by reframing: "If this is a bubble, then bubbles happen every five years."

At least as I read http://www.paulgraham.com/disagree.html, it is an ontology for disagreements.

Seiji's comment is not phrased as a disagreement, but rather a statement about how people will perceive your views. When the way people perceive a statement is important, predictions about how the statement will be perceived is as or more important than the truth of falsity of the statement.

In terms of your hierarchy; I don't think that the take-home message from your hierarchy should be 'only ever say things that could be classified as DH6'. Discourse from authority or lack thereof has no place in scholarly correspondence of any form; everything should be based on the evidence available. However, non-experts often don't have the background or patience to evaluate claims about a certain field well enough to distinguish good reasoning from crank reasoning.

Arguments from authority are therefore sometimes necessary to convince people on things they aren't experts in, and an argument against this authority can be an appropriate response (e.g. arguments like 'this person is saying things at odds with what 99.9% of scientists in the relevant field say, and are making claims that are outside their area of expertise', or 'this person has a serious conflict of interest in saying that').

The YouTube video doesn't work for me with either gnash or Adobe SWF player (YouTube normally works, I don't know why this one doesn't - maybe geographic restrictions) so I can't respond to the content of the video. The title is certainly an appeal to authority - "PG says this".

For the uninitiated: http://www.paulgraham.com/disagree.html

I think you meant DH1...?

Exactly. Saying there's no bubble is a far cry from saying that valuations are high and will go down/correct/whatever.

The digital media industry is a bit more mature now and the likelihood of a bubble is much less now. I imagine it will have it's ups and downs (as you said) just like any other industry. Also, a big difference between now and the late 90s is that companies are actually making money! The promise is starting to pay off.

The suggestion here is that YC's view is due to them not being impartial. I think a more likely explanation is that they are indifferent - their investment system has continued to allow them to 'buy low', even through-out this so called bubble, and AFAIK they only sell at maturity (aquisition / IPO), so the timing of the sales are not bubble timing driven.

If you see YC attempt to liquidate lots of their positions at once, then maybe your suggestion might ring true...

Buy low is contingent on sell high. Reaching sell high requires funding to live long enough for an exit. Constantly saying "there is no bubble" helps investors feel better about a three month old company claiming a $10MM+ valuation. "No bubble" also helps acquirers justify paying > $20MM for three guys sitting in their underwear typing on a computer for a few years.

(disclaimer tags: conspiracy, not aligned with reality, condescending, wrong numbers.)

In general, good investments should remain good investments even if there are no buyers. I'd bet this is true for startups as well. Overvaluations help, but aren't necessary.

I remember seeing a comment by pg that if he really thought there was a bubble and was cynically exploiting that fact, he just wouldn't say anything at all.

I think, from the perspective of pg and YC, there really is no bubble, at least with angel investments. With YC, high quality startups building things people want are raising enough money to bring products to market.

The real bubble is on the high end with Facebook's valuation jumping billions of dollars every week, even though they're obviously not creating billions of dollars of value in a week. That's the definition of speculation.

"... I remember seeing a comment by pg that if he really thought there was a bubble and was cynically exploiting that fact, he just wouldn't say anything at all. ..."

That was probably straight out of "How to Win Friends and Influence People". If there is a bubble, what you see is large transfers of money flow from people who want to risk capital to make profit. Does Milner, DST really understand "social media"? Probably not. There was a dig by Annie Lowrey a writing an article on Warren Buffet buying into railways in Slate that hints at this [0]. Is this a wise move by YCombinator? In the short term yes. In the longer term I'm not so sure. Why? If you want to understand the nature of an investor, follow the source and providence of the cash.

    Yuri Milner > DST > Alisher Usmanov

The Milner investment trail leads straight back to Uzbeck Russia and old-school Oligarchy, something the Economist has noted. [1] Randomness didn't play a part in the success of DST as a company. [2] I can't believe pg isn't aware of this [3] and if you look carefully, the Milner deal is at arms-length of ycombinator. A deal good for all parties at the expense of appearing "murky".

[0] Annie Lowrey, Slate, "Working on the Railroad: All the big returns were in Singapore or Hong Kong or Silicon Valley, they said, so they put their money into hot social-media start-ups and emerging economies." ~ http://news.ycombinator.com/item?id=2277702

[1] "A bigger problem for DST may be that some see it as Russian—and thus “murky”." ~ http://www.economist.com/node/16539424?story_id=16539424

[2] "According to the 2010 edition of Forbes magazine, the oligarch is one of Russia's richest men, with a fortune estimated at US$19.9 billion, and the world's 100th richest person." ~ http://en.wikipedia.org/wiki/Alisher_Usmanov

[3] "There are a lot of ways to get rich... There are plenty of other ways to get money, including chance, speculation, marriage, inheritance, theft, extortion, fraud, monopoly, graft, lobbying, counterfeiting, and prospecting. Most of the greatest fortunes have probably involved several of these." ~ http://paulgraham.com/wealth.html

Just because there's speculation doesn't make it a bubble. Look at oil prices. They've been supremely inflated by commodities and futures traders, but they'll remain inflated as long as there's sufficient trading volume to keep them up. If they ever deflate, it's not like it will bankrupt the oil industry.

A tech bubble is only a problem if its bursting would destroy sufficient wealth so as to impact the viability of future innovation. Taking facebook as the canonical example, popping that bubble is only a problem if a lot of the wrong people lose their money. GS losing their money doesn't really hurt anybody but GS and their clients. FB's VCs have all gotten their money back by now, so they're fine.

In order for the bubble that everyone here fears (startups) to burst, you would have to have the widescale destruction of wealth through the failure of many highly-priced startups. It would have to remove enough investors from the market to make money "sticky", that is, to the point where it stops flowing. If you look at the silly valuations out there, Facebook isn't going anywhere and isn't losing money, so they've got time to figure things out. Groupon makes disgusting amounts of money. Twitter... well, they're dispensible and their valuation isn't too ridiculous.

Even the "service not business" startups aren't taking much money out of the ecosystem. We might be in for an adjustment, but it shouldn't get to the point where good startups have trouble finding money.

Yeah, I came over here to point out that he's talking his book. :)

Speaking as someone who does not live in the Valley, if this is a bubble it's a great deal smaller than the 1999-2000 one, in the intangibles at least. Buzz may be up in the Valley but it's nonexistent out where I am.

I do think there may be a rush to declare bubble. You know, a lot of the promises that powered the first bubble are still true. The Internet really is going to revolutionize every business. Opportunities really are everywhere. It just was and is going to take a bit longer than initially expected, and 1999 infrastructure really couldn't support it. (Remember, in 1999, your top-of-the-line server chip is a Pentium III Xeon, built on a 250nm die, at 600MHz or so, and let's not even talk about the price of one of these. Or how your non-very-tech-savvy customers are supposed to get to your very expensive server.)

> The Internet really is going to revolutionize every business.

No it's not. How is the internet going to revolutionize construction work or trucking? The internet may improve aspects of the business, but it won't bring out any revolution.

Do you really think how your houses foundation is laid will be changed by the internet? Do you think how it's framed will be changed by the internet? Sided, shingled, insulated, drywalled, carpeted, tiled, plumbed, wired? If you genuinely think the internet will revolutionize these industries you're wholly detached from reality and then we might as well be in a bubble.

I suspect you'd find bigger changes that you might think when it comes to coordination and project complexity.

But I really should have said "computers" and not the internet. In about 5 to 10 years, robotics are going to "revolutionize" construction, which in this case sort of translates to "devestate" in much the same way manufacturing has been "devestated" by high efficiency. But somebody in the Valley is going to make a crapload of money on that.

The internet changed airlines, but airplanes still fly the same way.

Change isn't revolutionizing. Change is change. Revolutionizing is changing virtually every aspect of a business. The internet didn't change how airplanes fly, are refuelled or maintained. It changes how the seats are sold.

Indeed. But since selling seats is now much cheaper, other aspects changed too to squeeze costs. Though you could argue that changing regulation did as much if not more to effect [sic] the rise of low cost airlines as the internet did.

Most of the impact of the internet seems to be about making things more efficient.

For example, in trucking, GPS/navigation combined with automatic planning of routes, and notification where the trucks are, allows making delivery more efficient and predictable.

It's subjective wether you can call this revolutionary or evolutionary. I'm also tending toward the second. These days, it's all about automating things we were already doing and replacing human workers. Not about really big changes...

Then again, the change from horses to automobiles didn't happen in a day either. There might be big changes in progress that we're not even aware of.

>How is the internet going to revolutionize construction work or trucking?

If you include the smartphone technology than there are big changes coming to construction and trucking. For instance, organizing fleets of trucks using real time gps tracking.

Construction will benefit even more. With an iPhone and the right apps they can co-ordinate in the field in new ways. GPS tagged photos have a lot of uses on construction sites. An electrician can easily document where every wire is. That's a huge advantage if you want to make changes later.

Plus things like real-time ordering of parts.

What's the economic benefit for an electrician to document where every wire is? In house construction the electrician is contracted and will never be the one who'll wire you a new plug 10 years down the road when the house is already on it's 2nd owner. For commercial the wiring is already in conduit and done to blueprints. Why would a commercial electrician have incentive to document where every wire is, when there's already blueprints saying where every wire is.

Just an FYI I work in construction, and have done many jobs. I've worked full-time as an electrician and currently as a siding installer (moved country). On the side or through renovation work I've done: plumbing/gas, drywall/plastering, baton-insulation, roofing, framing.

Beyond a foreman's job the internet and smartphone devices are useless. Again a sole-operator (how I worked as an electrician) would gain benefit as you're foremaning your own jobs, which is where real-time ordering would be beneficial. I work with a company now and my being able to real-time order is irrelevant as we have a warehouse manager and the companies only deliver once a week to us. Anything beyond our siding materials, we get from home depot (AKA insta-delivery).

As an electrician I'd have rather had info on why a previous electrician had wired it that way. That takes more figuring out than the 5 minutes tracing the wires (unless you're chasing through a drywall ceiling, FYI that's where I first picked up plastering skills).

I always think this is funny. Most people didn't see the housing crisis coming before it was terribly obvious to everyone. Crying bubble doesn't make you look like you're any good at predicting anything. The only danger of a bubble has to do with debt.

Most startups are funded by savings and while that savings can be wiped out no one is hurt after that hit. Businesses fail all the time and only when massive amounts debt is involved is anyone else (outside the people directly involved) affected.

A bubble would start to scare me if people started taking out massive loans to startup companies that had no hope to make money. I sometimes wonder about the quality of the companies that are able to raise money now but I'm not worried about them taking it. The investors will learn from their mistakes and hopefully we won't have as many daily deal sites for me to deal with. In the meantime I'm using Yipit to help me who also raised a decent amount of money I guess. Oh well.

Are you sure that most VC comes from savings?

Wouldn't it have to? You can't take out a loan for a VC venture for less than 5% considering the extreme risk involved in a poorly run fund. I'd imagine debt financing would be cost prohibitive even if you found a bank crazy enough to loan you the money.

Are you sure that banks don't somehow have a hand in these funds. I don't know -- just guessing that easy money has something to do with all of the cash getting tossed around right now.

I'd interpret "most startups are funded by savings" as meaning private savings rather than either VC or Angel investment.

PG has all these amazing quips.

> "If you have big plans initially, you are probably Webvan" > "The valuation of an early-stage startup is the % chance they will be big. i.e. a $10M valuation ~= 1% chance they will get to a $1B valuation".

Also, am I the only one that picked up that AirBnB and Dropbox clearly have a valuation higher than $250M. I wonder what they consider their valuation to be right now.

I think the only valid answer to this is a tweet from http://twitter.com/hackernewstips:

> Today, @PaulG claimed there is no tech bubble. In unrelated news, @AdKeeperInc raised $40 mil. in funding for "Delicious, for banner ads".

Which is exactly what AdKeeper is, btw.

I don't remember the company, but I got a demo of this exact thing in the last bubble (around '99 or '00). The one I saw failed utterly, but maybe this time, they'll figure it out.

Heh, I ran into Scott Kurnit and he asked me to take a look at it. I guess that's why.

http://about.adkeeper.com/team/leadership-team/scott-kurnit/

I think he knows what he is doing.

ad hominem

This is one of the classic signs it's a bubble right? (when people in the industry claim it's not)

Isn't the biggest and most obvious sign when people outside of the industry get into the industry?

As in when everybody's grandma had started investing in startups in 1999 and in the real estate bubble when everybody's grandma was suddenly buying big houses and selling them with a markup 2 months later.

Wait until they start making TV shows about how to make money quick off of it.

That's actually a pretty good sign it's saturated, which is probably a necessary condition for a bubble.

You mean like AngelList?

True, but by that measure you can't win, because when even the insiders who profit from the situation are calling it a bubble, the signs are even worse!

OK, so when is it not a bubble?

I suppose if everyone says it's a bubble, it's not one, but how often does that happen?

It's easy to be negative and "predict" things - once in a while you might even get it right, then the world will notice you as a visionary :)

The press is shouting "bubble". People in the industry, afraid of a domino effect are claiming it is not. News at 11.

If you want signs, you need to look for them elsewhere. Like the huge evaluations for companies that fail to produce profits; in spite of their enormous size. Also, are these evaluations growing exponentially?

I'm don't know much about the inner-workings of an economy, but most articles I've seen on the subject are very shallow and totally void of meaningful data.

I appreciate your humor. Unfortunately there are so many who actually reason using the most common of fallacies: reversing the direction of the arrow of implication. A->B != B->A.

That is such a worthless, and wrong, thing to say. It's like saying that since denial is a sign of grief, if you say you're not grieving therefore you are.

PG - your enthusiasm is inspiring.

Let's talk about burst scenarios. I'm not worried about the tens of startups raising $500K at inflated valuations. It doesn't really matter.

What I'm worried about is someone like Zynga having a bad day. Imagine their next game doesn't quite take off as much. Boom: their valuation crashes. Since they are so linked to Facebook, I'd expect Facebook to crash as well. And once Facebook is not so hot, then Twitter, with abysmal revenue, will crash too.

The question is: what happens after that? Angels stop pouring money in startups, Silicon Valley goes into nuclear winter like 2003, or not?

No one really knows if there's a bubble before it bursts. They are easier to see in hindsight.

On the one hand everyone in the world is getting a smartphone since the pricing for chips has come down so much this year. That will make the market for mobile computing/commerce/you-name-it huge.

On the other hand, Facebook hasn't really nailed their business model. Yes, it's pulling in a lot of cash, but not enough to justify its valuation. Assume Facebook IPOs in 2012 and because everyone except newborns is on the network, the price gets bid sky-high, but they fail to grow revenue. Investors get scared and pull out and take the rest of the tech market with it. Then Facebook grows credits to the be the biggest payment platform in the world and the whole market goes up again.

Like PG said, market valuations are a sine wave. They are educated guesses, no one knows anything for sure.

Here's more on Facebook's valuation: http://community.nasdaq.com/News/2011-03/how-to-justify-face...

No one really knows if there's a bubble before it bursts

This is commonly quoted but not true. Everyone who had any clue saw the housing bubble collapse coming a mile away. I was warning friends buying houses in 2006 that they were buying at the peak of a bubble. And everyone on Wall St knew exactly what was coming down the pipe.

You have a point, but no one knew when it was coming (hence why I said you can only see bubbles in hindsight). I should rephrase and specify that no one can tell when a bubble will burst. And as long as a bubble keeps growing, you can still make money buying into it.

In video it's explained like "valuations are high, but it's not a bubble".

Would that mean their growth will stop at certain level for a longer period? I guess realization will need to meet expectation at some level. Expectation for Facebook are sure high, but I doubt it can go go for trillion valuation without multi^2 billion revenue.

Would love to hear more thoughts from Paul.

Another example from the video is that a 10M valuation for a startup isn't really saying the startup is worth 10 million, but rather that it has a 1% chance of reaching 1 Billion.

One can actually look at Facebook the same way. One could argue that Facebook is competing with Google at being the primary way people discover information online, and at being a driver of incoming traffic. So one could say that Facebook's valuation with the GS investment reflects a belief that Facebook has around a 25% chance of becoming as valuable as Google.

Would you ask the barber if you need a haircut?

Might want to rethink your position in bonds. Much more to worry about there than in startups.

Ha, his shirt is same color as this website bar.

(also just realized it's near prison jumpsuit orange, lol)

The tech biz has had boom and bust cycles for as long as I have been alive. While it's true that some of the bubbles are minor ones we should all be aware of the fact that the bubbles will always pop — it's not a question of "if" but "when" (and "who" does it impact and my favorite "why").

If you accept this simple fact of life you'll always do better when a gold rush does occur. Also the great thing about throwing away your rose colored glasses is that it makes you stay in the field for the next boom.

Yours truly a survivor of the original dot.bomb fiasco who got his first break in the era of HyperCard....

In this video the way he talks(volume, stress, etc.) is really similar to Mark Zuckerberg.

When someone says "There is a Bubble". Which means they are feared, got mass hysteria, jealous of early investors in future hot companies & may actually have some valid point."

Valid point is always about 20% of all over hype.

The companies that are being talked about are not public yet - Facebook, Twitter etc. so would a tradational bubble really apply to this?

Why not? A bubble occurs when people over invest due to mass hysteria. Whether the investments occur through a public exchange, or through second markets the psychology is still the same.

Facebook may have a valuation of $60bn, but actually the ammount of cash invested is only $2.34bn [http://www.crunchbase.com/company/facebook] so that would be what was lost if it collapsed. Same for Twitter, actual investments are $360m not the $7bn it maybe valued at. Though I guess it's unknown how much has been traded through secondary markets.

My point was, what effect would it have on the NASDAQ or any stock market? And would it mean that anyone investing in bonds really lost anything?

"Discovered SPAM filtering algorithm" ?

PG wrote an influential article "The Plan For Spam" some years ago (early 2000s because it was old when I entered industry) describing spam filtering using a naive Bayesean filter, a technique which was not totally unknown but which was subsequently widely adopted. It works pretty well, especially since you can chain it with other things like (probably the biggest one) IP based reputation. One popular implementation, PoPFile, was by fellow HNer John Graham-Cumming. For years I had an unfortunate hash collision and thought they were the same people.

Spam researcher hat off.

Didn't Paul Graham actually describe something more involved than a naive Bayesean filter?

Feel free to Google it and post your own synopsis, but insofar as you can do lossless compression to three words, I think those are the right three words.

How about dropping the `naive' part?

If I'm reading his essay correctly, it pretty directly proposes the kind of Bayesian inference called "naive Bayes", i.e. which makes an assumption of independence of the features (in this case words), and calculates the total probability of an email being spam by simple multiplication of the per-feature probabilities.

Yes, that's true.

I was more thinking of his weird pre- and post-processing like >>> When new mail arrives, it is scanned into tokens, and the most interesting fifteen tokens, where interesting is measured by how far their spam probability is from a neutral .5, are used to calculate the probability that the mail is spam. <<<

Yeah, I agree he has some interesting pragmatic tweaks on it. I suppose he was proposing "[naive Bayesian] filtering" but not necessarily "naive [Bayesian filtering]".

I never knew Paul Graham had an American accent :(

Reply on news.ycombinator.com