00:00hi everyone welcome to the a 6 in z
00:01podcast I am sonal and I'm here today
00:03with a special podcast we have on the
00:06heels of announcing our fifth fund for
00:09andreessen horowitz and we thought we'd
00:11talk more broadly about what's changed
00:12between the first fund and now and more
00:15importantly some of the technology
00:16trends and trends we're seeing with
00:17founders and to have that conversation
00:19with us we have our co-founders Marc
00:21Andreessen and Ben Horowitz and our
00:23managing partner Scott Cooper welcome
00:25guys hey okay so let's just kick things
00:28off one of the things that I want to
00:30understand is that it's been since fund
00:321 which is what six seven years ago yeah
00:35seven years ago a lots changed in seven
00:37years and I've actually heard you argue
00:38Marc that things have accelerated in
00:40that time period more so than previous
00:41decades before so what do you guys think
00:44are the biggest shifts now that are
00:46important to us in this newest fund and
00:48what changed in that period like the
00:50biggest things so in fun one when we
00:52started we thought that our timing was
00:55really good despite the fact that I
00:57think the world thought our timing was
00:59really bad and starting a new venture
01:00capital fund and the reason why we
01:02thought that was that there were three
01:04gigantic new platforms hitting all at
01:07the same time which was kind of
01:09unprecedented in the history of
01:11Technology one was mobile the second was
01:14social and the third was cloud and that
01:17really proved out through the course of
01:19the early history that the applications
01:22on top of those particularly mobile and
01:24cloud were just spectacular and I think
01:28we're coming a little bit to the end of
01:30the first phase of the is you know some
01:33of the obvious applications that could
01:36be built on those things and we're
01:37moving into to some new areas yes so let
01:39me go to the foundations so there's
01:42different ways of looking at it the
01:43foundational levels one is Moore's Law
01:45has really flipped and this this
01:47actually has happened I think this
01:48actually has happened over the last
01:49seven or eight years actually almost
01:50exactly over the life of the fund which
01:52is you know for many many years Moore's
01:54law was a process of the chip industry
01:56bringing out a new chip every year and a
01:58half that was twice as fast as the last
02:00one at the same price and that continued
02:02for 4050 years and that's by the way
02:04what resulted in everything from
02:06mainframes many computers pcs and then
02:08and then smartphones about you know
02:10seven eight nine ten years ago that
02:12process actually started
02:13coming to an end the way that it had
02:14worked up until then so chips have kind
02:16of topped out and it was speed of about
02:17three gigahertz and a lot of people have
02:20said they're for like progress in the
02:21tech industry it's gonna stall out
02:22because the chips aren't getting faster
02:24I think what's actually happened is
02:25Moore's law has now flipped the dynamic
02:27now instead of increased performances
02:29reduced cost you now have this dynamic
02:31where every year year and a half the
02:32chip companies come out with a chip
02:33that's just as fast but half the price
02:35and so this is this sort of just this
02:37massive massive deflationary force I
02:40think in the technology world and I
02:41actually also suspect on the economy
02:44we're basically computing is just
02:45becoming free basically what we do in
02:47this businesses we just kind of chart
02:49out the graphs and then just kind of
02:50assume at some point you're gonna get to
02:51the end state and the end state is gonna
02:52be the chips are gonna be free which
02:54means chips will be embedded in
02:55everything you'll be able to use chips
02:56for literally everything and we've never
02:58lived in a world before we where you can
03:00do that so that's the first one second
03:02one is just the obvious implication from
03:04that which is all those chips will be on
03:05the network right so all those chips
03:06will be connected to the Internet
03:07they'll all be on Wi-Fi or mobile
03:09carrier networks or wired networks or
03:10whatever but they'll fundamentally be on
03:12the Internet you know that's that's
03:14something that's not happening at a very
03:16rapid pace and then the third is the
03:18continuation of the piece that I wrote
03:19actually five years ago which was called
03:21suffer eats the world which is basically
03:22just saying if you're gonna live in a
03:24world in which there's gonna be a chip
03:25in every physical object and if you live
03:27in a world in which every physical
03:28object therefore is going to be
03:29networked there's gonna be smart because
03:31it has a chip and it's gonna be
03:32connected to the network then basically
03:34you can then program the world you can
03:36basically write software that applies to
03:37the entire world so you can write
03:39software that all of a sudden applies to
03:40all cars so you can write software that
03:42applies to all you know everything
03:43flying in the sky you can write software
03:45that applies to all buildings so you can
03:47write software the applies to you know
03:49all homes or all businesses or whatever
03:52all factories and so all of a sudden you
03:53can kind of you can program the world
03:55that's really just starting and I think
03:57a lot of there's a number of things that
03:59make the entrepreneurs we're seeing
04:00these days in many ways more interesting
04:02and more aggressive than entrance we've
04:03seen the past and part of it is they
04:04just assume if there's something to be
04:06done in the world there must be a way to
04:07write software to be able to do it
04:08that's at a new level of power or
04:11sophistication it's a new scope of what
04:13the tech industry can do the consequence
04:15of that for us is a fund is that we find
04:17ourselves evaluating business plans and
04:19funding companies that are in markets
04:20where I think seven or eight years ago
04:21we would have never anticipated
04:23operating so mark does that mean that
04:25there's no new innovation in plan
04:27forms themselves and everything all the
04:29innovation will be applications that
04:30ride on that existing infrastructure or
04:32do you think there's also the
04:33opportunity to build a new platform even
04:35giving some those trends I think there
04:37are new platforms and I think there will
04:38be new platforms I just think there will
04:39be different kinds of platforms than
04:40we've had in the past the idea of a
04:42platform in the tech industry as you
04:44know up until up until you know five or
04:4510 years ago was there is a new chip
04:47that has new capabilities is faster and
04:49then therefore you build a new operating
04:50system for it and that might be Windows
04:52or it might be you know maybe iOS or
04:53whatever it is the platforms that we're
04:56seeing getting built these days are
04:57distributed systems so scale out systems
04:59sort of being built on a chip
05:01necessarily with new unique capabilities
05:03they are platforms that are going to
05:04build across lots of chips
05:05and so they're in computer science terms
05:07there's distributed systems the cloud is
05:09one of the first examples right so
05:10anybody who uses AWS can now go on it
05:13can program an application on AWS that
05:15will run across 20,000 computers and
05:17they can run it for an hour and it'll
05:19cost you know 50 bucks and and that's a
05:21kind of platform that did not exist
05:23before and by the way there are many
05:24specific elements to that so for example
05:26we've seen the rise of in that category
05:28seeing the rise of Hadoop and otherwise
05:29a spark for distributed data processing
05:31we've seen in financial technology we've
05:34seen the rise of a Bitcoin of
05:35cryptocurrency which is a literally
05:36distributed platform for you know for
05:38currency and for exchanging value and
05:40now we're seeing the emergence of a
05:42major new platform which is which is AI
05:44machine learning and deep learning which
05:46is inherently the the great thing about
05:48machine learning and deep learning is
05:49they're inherently parallelizable they
05:51can run across many chips and they get
05:52very powerful as you do that and you can
05:55do things in AI today as a consequence
05:58of being able to run across many chips
05:59that you just couldn't even envision
06:01doing five or ten years ago so let's
06:02talk about the rise of the GPU as part
06:04of this next platform chip I may think
06:05the biggest surprise people have had is
06:07that this is the graphical processor
06:09unit which is something that was
06:10developed in the gaming industry for
06:12really high resolution graphics
06:14processing and is now finding I guess
06:16unexpected it was a surprise to us that
06:19it's finding uses in these new platforms
06:21like VR they are deep learning it's
06:23actually interestingly it's a new
06:24application of an old idea back when I
06:26was getting started 30 years ago working
06:27in physics labs if you wanted to run
06:29just a normal program you just you just
06:31buy a normal computer and run the
06:33program but if you wanted to do run a
06:35program many physics simulation said
06:37this had this property where that you
06:39you would want to run a very large
06:41number of calculations in parallel right
06:43now so you could either basically divide
06:44up a problem to simulating anything from
06:46a black hole or two different kinds of
06:48biological simulations you to basically
06:51write these algorithms in a way that
06:52they could run you could basically
06:53parcel the problem into many different
06:55pieces and then run them all in parallel
06:57and there was actually in the old days
06:58there was actually a whole industry of
06:59what we called vector processors which
07:01were literally these kind of sidecar
07:03computers that you would buy and you
07:04would hook up to your main computer and
07:05they would let you run these parallel
07:07problems much faster and so literally 30
07:09years later the GPU is effect it's
07:10basically a vector processor it's
07:12basically a sidecar processor that's
07:13just a long a CPU and runs these
07:15parallel problems much faster and it the
07:17graphics are a natural application of
07:18that but as it turns out graphics aren't
07:20the only application
07:21yeah actually interestingly and I was at
07:23a company making one of these called
07:25silicon graphics and the applications
07:28then were as mark saying a lot of
07:29physics applications computational fluid
07:31dynamics and simulating flight
07:34simulation and all these kinds of things
07:35that are hard physics to calculate when
07:38you go into the virtual world and you're
07:39simulating the physics of the real world
07:41guess what you need the exact same
07:44processor to do it so it's it's super
07:47logical conclusion to what's been going
07:49on but I think we're also in the world
07:51of big data seeing kind of more reasons
07:54to do just lots of math in parallel and
07:57so it's it's an exciting application
07:59yeah you talk about platforms one of the
08:01really interesting hardware platforms is
08:03emerging right now is Nvidia which is a
08:05very well established public chip
08:06company but very successful to your
08:08point doing graphics chips for very long
08:09time has become seemingly overnight it's
08:13really of course the result of years of
08:14work but seemingly overnight has become
08:16the market leader in both not just GPUs
08:18but also in chips being used for AI and
08:20it's it's it's basically extensions of
08:22the GPU technology and and we've see
08:24this overriding theme which is kind of
08:26an amazing thing which is basically
08:28every sharp AI software entrepreneur
08:29comes there comes in here one is now
08:31building on top of Nvidia strips which
08:33is of course a very different outcome
08:34than entrepreneurs of previous years who
08:36would have built other kinds of programs
08:38primarily on top of Intel chips we've
08:39mentioned AI and machine learning a
08:41couple times here and and one of the
08:42interesting things at least I think we
08:44see in the industry is at the same time
08:46we've got startups doing it we also see
08:47some of the very largest
08:48which players investing significantly in
08:50AI and machine learning so certainly
08:52Facebook and Google Apple and others are
08:55obviously building building big
08:56operations how do you think about the
08:58the universe from an investment
08:59perspective what are the kinds of things
09:01that actually lend themselves well to
09:02startup opportunities in the AI space
09:04versus things that actually might make
09:06sense kind of living inside of one of
09:08the larger larger companies like a
09:10Facebook or Google yeah so you know AI
09:12is extremely broad and I think one of
09:14the challenges that people have with it
09:16as they try to paint it as a narrower
09:18thing than it is but it one can think of
09:20it as an entirely new way to write a
09:23computer program and so then it's
09:25applicable to you know the universe of
09:28problems so there are things that
09:31advantage a big company you know if
09:32you're building AI to analyze consumer
09:35Internet data like that's hard to take
09:37Google on it that they do have an awful
09:39lot of data and you know Facebook you
09:42know with AI computing power matters and
09:44and the data set matters having said
09:46that there are a lot of areas where
09:48nobody has any data yet in the areas of
09:50healthcare and the areas of autonomy so
09:55you know there's lots and lots of
09:57opportunities and you know there's also
10:00interesting ideas about well is there a
10:03better user interface than the smart
10:05phone using AI techniques and then what
10:08do you mean when you say there's a
10:09better user interface well yeah if you
10:12think about a smart phone it was kind of
10:13an advanced over what we used to call
10:15the wimp interface windows windows icons
10:21what was it which you know was like a
10:29big advance over the text-based
10:31interface of DOS and then you know the
10:34smartphone with a touch interface it was
10:36more of a direct manipulation was in
10:38advance over that and so you go okay
10:40well but that's not actually what people
10:43do in life right like you're yeah it's
10:47anthropologically it's it's a backward
10:49step in terms of the natural interface
10:51that we've become accustomed to like for
10:54example natural language with AI you get
10:57into a world where things like natural
10:59language and natural gestures and so
11:02become much more plausible so there's
11:04you know potentially an opportunity to
11:06build interfaces for things that that
11:09you couldn't before I mean I think
11:10there's one like really interesting
11:12thing which I'm sure and I know that
11:14Google and Apple and all the giant
11:16companies are very focused on which is
11:17how do you replace the current set of
11:19user interfaces with it but there's
11:21another dimension which is what are all
11:24the applications that you just couldn't
11:26have before because you can build a
11:29workable user interface for it and hey I
11:32seems very promising in those areas
11:34you didn't mention Amazon which is sort
11:36of the stealth player here with echo and
11:38Alexa I mean really they've got an
11:43interesting advantage in that they're
11:44not and you know they're not tied to the
11:47last generation of user interfaces so
11:49that they don't have to pay the strategy
11:51tax for shoehorning in their AI into say
11:56the iPhone and that's that's something
11:57yeah that's worth pointing out there
11:59says there's sort of two kind of classic
12:00rules of thumb in this industry one is
12:02for major new advances especially in
12:04things like interfaces if you don't own
12:05a platform you can't do them and so the
12:07assumption I think had been up until
12:08recently you know that it would have to
12:10be Google or Apple that does these kind
12:11of natural language or interface
12:12advances because they own iOS and
12:14Android the other rule of course is the
12:16exact opposite rule which is the one
12:18that Ben mentioned which is the problem
12:20that big established companies get into
12:21is this what he referred to as the
12:23strategy tax which is basically big
12:25companies with existing agendas have to
12:27sort of fit their next thing into their
12:29existing agenda and they often
12:30compromise it in the process and so it's
12:33sort of this ironic twist of fate that
12:34amazon has all of a sudden taken the
12:36lead from Google and Apple even though
12:37Amazon you know famously flopped with
12:38their phone right which is sort of the
12:40obvious place where you have a voice
12:41interface it didn't matter because they
12:43came out with this new product which was
12:44this basically the speaker the smart
12:45speaker called echo and the fact that
12:47all of a sudden Amazon didn't have a
12:49phone all of a sudden became an
12:50advantage because they could just do the
12:52clean actual breakthrough product
12:53without worrying about tying it into the
12:55existing strategy right and those are
12:57all still big companies those they're
12:58not really hearing where startups can
13:00really play in this space especially
13:02when you are describing this huge data
13:03network effect that all these big
13:05companies have a year ago we would have
13:07probably been sitting here and say that
13:09AI was going to be likely would be a
13:10domain of big companies because of this
13:12sort of this sort of thing of like okay
13:13only big companies can afford the very
13:15engineers that are required to do AI
13:17only big companies can afford the amount
13:19of hardware required to do AI and then
13:21only big companies can get the giant
13:22data sets required to do AI in the last
13:2412 months what we've seen basically is
13:25all three of those changing very fast
13:27and to the advantage of startups we've
13:29seen a lot of AI technologies actually
13:31actually now interestingly standardizing
13:33so going to open source and then the
13:36next step is going to be they're gonna
13:37go to cloud and that we're right on lose
13:38we think we're right on the verge of
13:39that we think all the major cloud
13:40providers are going to be providing AI
13:42as a service and they're gonna really
13:44radically reduce the amount of technical
13:45knowledge you need to apply ai and so
13:47that plays very well to the startups so
13:48there will be like an AWS for AI yeah
13:50exactly and that may be literally AWS or
13:52it may be Google or Microsoft where all
13:53three of them and some you know in some
13:55combination or or it may be other you
13:57know other companies yet to emerge an
13:58example of the open source be like
13:59tensorflow this is a big deal and of
14:01course yes right so Google open sourced
14:03a pretty significant part of how they do
14:05deep learning and that actually now is
14:06something other companies can pick up
14:07and use directly and we see we see
14:08actually a lot of come out only a lot of
14:10companies but like a lot of university a
14:11lot of student projects now just kind of
14:12picked that up and run with it so so
14:14this technology is kind of trickling
14:15down very fast just this past weekend we
14:17had a hackathon and I think most of the
14:20teams had some machine learning AI
14:22component into their hacks and these are
14:25college kids yeah yeah if you're if
14:26you're a Jew you know if you're if
14:27you're a 21 year old junior in college
14:28and you're doing some project it's just
14:31kind of it's becoming rapidly becoming
14:33very obvious that you would have a I'd
14:34be part of it which was very much not
14:35the case even full months ago and that
14:37that's a director point that's direct
14:39consequence of the open sourcing and
14:40kind of this malov spreading out the
14:42second thing was the hardware cost and
14:43there again the cloud AI in the cloud
14:45just the existence of the cloud is
14:46bringing down hardware costs across the
14:48board but AI and the cloud is gonna
14:49bring that down even further and by the
14:50way these trends all slam together so
14:51you get you know what I think in a year
14:53is gonna be very common to these sort of
14:55AI supercomputing chips with AI
14:57algorithms in the cloud available to
14:59anybody for $1 right and so there's
15:01gonna be this massive deflation of
15:02hardware cost on on that side these big
15:05datasets are interesting been made the
15:07case that the Stratus can assemble big
15:08datasets and I think that there are
15:09there are certainly examples of that we
15:11also see another thing happening which
15:12is the newest generation of experts in
15:14deep learning or many of them are
15:15specializing in the idea of deep
15:17learning applied against small datasets
15:18if you talk to those folks what they'll
15:20tell you is oh basically they'll
15:23basically say is primitive and crude
15:24deep learning required big datasets but
15:26the really good stuff doesn't it is
15:28small datasets are fine
15:29and so that that's still very early but
15:31it's extremely enticing it's an
15:32extremely enticing idea because it
15:34really brings a lot of these problems to
15:36your point further into into being
15:38tractable for small companies but
15:40actually one of the things you can do
15:41with these especially with these GPUs is
15:43look you can literally use the same
15:45tools that are used to make video games
15:46and you can create simulated versions of
15:49the real world and then you can actually
15:50let the AI train inside the simulation
15:52and so if you're building a new
15:53self-driving car or a drone or something
15:55like that you can actually create
15:56simulated worlds in which there are
15:58everything from earthquakes to floods to
16:00you know thunderstorms hail storms you
16:03know you can create you know birds you
16:05know this swarms of birds you can
16:06literally simulate the real world
16:08environment and then you can let AI
16:10actually train inside that world and
16:11actually it's funny I actually has no
16:13idea it's training in the virtual world
16:14it's it's learning just the same as if
16:16it were learning in the physical world
16:17and so again for startups with access to
16:19cloud-based AI you could potentially run
16:21basically millions of hours of simulated
16:23training at very low cost and all of a
16:25sudden catch up to big companies
16:26interestingly you know the very famous
16:28AI project that Google did with deep
16:31mind they that whole dataset came from
16:33the game playing itself so you know
16:36there wasn't some dataset that Google
16:37had collected over 20 years it was the
16:40game playing itself so you guys have
16:42both mentioned simulations a few times
16:43why are they so important because I feel
16:45like there was this period like you know
16:46maybe even a decade ago where our
16:47simulations were almost frowned upon as
16:49this promised being that it didn't
16:51really actually deliver in what you
16:53needed to to be able to navigate complex
16:56environments in real life yeah well it's
16:59interesting so was AI was frowned upon
17:01ten years ago saying it was all it
17:04didn't work I mean particularly neural
17:07nets and deep learning were the most
17:09frowned upon area and and there's been
17:11similar kind of breakthroughs for
17:13simulation and first of all so if you
17:15think about the field of data science
17:18and what you do with data there's you
17:20have a giant set of data which is always
17:22historical in nature and you can analyze
17:25that and maybe it's predictive of the
17:27future but oftentimes it's not and you
17:30know we see this in particular in things
17:32like you know really dynamic things
17:33where the past affects the future like
17:35say stock picking or you know where the
17:37weather or other kinds of things where
17:39you know data analysis doesn't get you
17:42sir simulation is a flip side of that
17:44where you can say okay here all the
17:45entities in the world and let's generate
17:48their behavior over time and then their
17:50actual behavior feeds back into the
17:52simulation which is critical you know a
17:55critical component historically that's
17:58been difficult at scale but there have
18:00been some really important breakthroughs
18:03lately particularly from a company that
18:05we're invested in called improbable
18:06which is able to do very large scale
18:09scale out simulation you know using
18:12cloud computing techniques and and some
18:14very important new technology that
18:17they've developed and so you can get a
18:19really complete picture of the world and
18:21as Mark was saying you can actually
18:23generate your own data set rather than
18:26collecting it for certain kinds of
18:27situations so one way to think about it
18:30is it's expensive to make things happen
18:32in the real world like it's expensive to
18:34change things in the real world because
18:35the real world is physical and causing
18:36physical changes to happen I mean
18:38everything from building roads to flying
18:39planes all these things are very
18:40expensive and then things in the real
18:42world changes have serious consequences
18:43right and so if you you know depending
18:45on where you put the dam or where you
18:46put the airport or what your evacuation
18:47plan you have for the city and if
18:49something bad happens likely you know
18:51these these decisions have huge
18:52consequences Thanks you bailout which
18:55banks you don't bail out and so you
18:58always have these consequences and
18:59people who have to make these decisions
19:00are often flying blind because they they
19:02don't have any real sense of what's
19:03gonna happen as a consequence of their
19:04decisions in contrast if you can
19:06simulate a world and if you can even run
19:08an experiment if you can simulate the
19:09real world or some portion of it like
19:11the highway system or the banking system
19:13or whatever and then you can basically
19:14introduce change into that simulation
19:16and you can see what the consequences
19:17are it's very cheap to do that because
19:19Moore's law and the collapse of chips
19:21and the rise of cloud computing all
19:22these other things we've been talking
19:23about all of a sudden make it very cheap
19:24to run these simulations it's much
19:26cheaper to do it in a simulated world
19:28and then there are no consequences you
19:29run a simulation and everything goes you
19:31know wrong and everybody dies or the
19:33entire financial system collapses or
19:34whatever it doesn't matter you just
19:35erase it and you run it again yeah you
19:37have infinite testability well said
19:40there is Elon Musk's simulation quite
19:43direful there's a scenario that we're
19:47I would argue it's bad Leroy as
19:50evidenced by the current political
19:52figures and I do of our button in this
19:53simulation yes and and then you
19:55basically again you you look at more you
19:56look at the progress of Moore's Law in
19:58the rise of these new technologies and
19:59you say ok how about instead of running
20:00one simulation let's run a million
20:01simulation so let's run a billion
20:03simulations and let's try every
20:04conceivable thing we can possibly think
20:05of and let's imagine let's literally
20:07model all potential future states of the
20:09world and then let's decide which one of
20:11those which which path is the one that
20:13leads to the best consequences and so we
20:15can then make these very big real-world
20:17decisions with a lot more foreknowledge
20:18of what of what will unfold afterwards
20:21maybe just to get concrete on some
20:23opportunities what are the other areas
20:25in maybe it's life sciences or what are
20:26some of the other kind of more tangible
20:28areas that you think near-term as you
20:29think about kind of deploying this fund
20:31or beyond over the next five to ten
20:32years it might be interesting for you
20:34know people to think about in the
20:35context of real-world applications of
20:36this technology yeah so spark was saying
20:39we're coming into this era of new
20:41platforms and with the intersection of
20:43health and computer science what we're
20:45saying is really exciting new platforms
20:47around data and around basically you
20:51being able to get much more information
20:52about someone's health from a variety of
20:55of techniques that have been developed
20:57but you know based on the kind of
20:59historic breakthroughs and sequencing
21:01the genome but beyond that as well where
21:05we can get really really powerful data
21:08about people and understand them better
21:11and once you have that data bad people
21:13when you can be predictive of diseases
21:15that they might get or things that are
21:17wrong and you aggregate that into a
21:19platform then you can actually make new
21:21scientific discovery off it as well so
21:24that's one interesting area there is if
21:27you think about the AI platform itself
21:28one of the things about it is the
21:30hardware that's been built for it or
21:33that's been built historically is for a
21:35completely different kind of computer
21:37programming and we've seen Google
21:39already announce a chip to power their
21:41deep learning cloud and you know
21:44similarly there's new breakthroughs and
21:46quantum computing which at least on the
21:49surface look like they may be very
21:50promising for much more powerful deep
21:53learning systems and so forth so there's
21:55a lot of things that are coming out of
21:59you know as we get to chip and
22:00everything the platforms to run and
22:03manage and understand those those chips
22:05are equally as exciting so when you know
22:08one of the themes that's come up through
22:09here is that tech is reaching into
22:12places it never did before I mean every
22:13company's becoming a tech company or
22:15they have tech inside or as Benedict
22:17likes to say attacks outgrowing the tech
22:19the reality is permeating everywhere and
22:21the question I have for us is that we
22:24are founded on this thesis that software
22:26is eating the world that's our premise
22:28and yet we've seem to have been making a
22:30lot of hard investments you know if you
22:32count things like Soylent oculus nutri
22:35box so are we changing our thesis about
22:38hardware as a result of this suffering
22:40in the world no I don't think so I mean
22:42I think that what we see with the
22:44companies that you've named are
22:45interesting so oculus I think we would
22:48all agree that the software component of
22:50oculus is both more complex has many
22:54more people working on it and is kind of
22:56the core of the investment sometimes if
22:57you have a breakthrough technology then
23:00you require new hardware to actually
23:02support it and that's the case there and
23:04I think that Soylent in in neutral box
23:06both of them apply computer science
23:08techniques and information technology to
23:11get people to optimal health and that's
23:12what we're doing there so I think we're
23:15big big believers that you know in the
23:18last hundred years the great
23:20breakthroughs and knowledge have been
23:21the breakthroughs of people like Alan
23:24Turing and Claude Shannon who gave us a
23:27new model of the world and how to
23:29understand it and companies that build
23:32on that fundamental knowledge
23:34breakthrough are what we're about and
23:35will continue to be about that even if
23:38some of them may shift their products in
23:39a box let's talk a little bit about sass
23:44as you've probably seen there's been
23:45actually a bunch of acquisitions in
23:46space recently but what's left to do
23:48there so is the new platform the
23:50Salesforce comms and others of the world
23:52or are they're actually both kind of
23:54vertical applications and or are there
23:56other platforms that actually might
23:57exist over time in that market
23:58so there's sass as the metaphorical in
24:01the cloud version of all the stuff that
24:03we had built over the previous you know
24:0530 40 years so that's like work day
24:09sales for us success
24:12factors you know the kind of big
24:14categories the thing that we believe
24:16that's changed as you go from on-premise
24:19to the cloud is the technology is so
24:21much easier to adopt that we're now
24:23seeing software applications for things
24:26that you just would never do as a
24:28software application because the cost of
24:30as we used to say in the old days
24:32screwing it in and paying the army of
24:35eccentric consultants to get it going
24:37just wasn't worth it for say expense
24:41reporting which you know concur of
24:43course built a really powerful product
24:46in that but like there was no packaged
24:48software for expense reporting in the
24:50same way that that there is now and I
24:52think there's a gigantic number of
24:54categories and everything that you do in
24:56business that can be automated in that
24:58way in addition to that you can scale
25:00down to very very small companies
25:02companies below thousands of employees
25:04never bought Oracle financials it would
25:07have been insane to do so but they're
25:09absolutely buying you know NetSuite and
25:11things like that and then beyond that
25:14you now it becomes economical and very
25:17interesting to build vertical
25:19applications for industry so to build an
25:22application that revolutionizes say the
25:25real estate industry or something like
25:26that or the construction industry is
25:28becoming extremely viable and not just
25:32as a niche business but as a real
25:33venture capital based kind of activity
25:35one of the consequences that will be
25:37interesting to watch play out is that
25:39historically enterprise software has
25:41been described as represented by
25:42companies like Oracle s AP IB M like
25:45that stuff was really only accessible to
25:47the largest companies the top 500
25:49thousand companies in a country and then
25:51in particular only in a handful of
25:53countries those businesses there their
25:55revenue and their customer base of
25:56always been dominated by you know two or
25:58three thousand companies globally that
25:59are these you know these giant
26:00multinational companies that we've all
26:02heard of so big companies have this sort
26:04of inherent advantage versus a lot of
26:06midsize and small companies and then
26:08companies in the US and Western Europe
26:09had this big advantage versus companies
26:11in other parts of the world where the
26:12companies large companies and the large
26:14companies in the US and Western Europe
26:15could just afford to make technology
26:17investments that small and midsize
26:18companies all over the world couldn't
26:19make the sort of changes in in in SAS
26:22that then described they needed an
26:24interesting conclusion which is it may
26:25actually be interesting
26:26for a smaller company or a company not
26:29in the US or Western Europe to be able
26:30to adopt the next generation of sass and
26:33cloud technology it's almost like the
26:34folks who've been able to skip landline
26:36telephones or just go straight to mobile
26:37phones you can just leapfrog the old
26:38stuff because you never had it and you
26:40can just start using the new stuff out
26:41of the box and then the big established
26:43companies might have a harder time
26:44adapting because they've made these
26:45giant investments in the old systems and
26:47it's hard to just jump to the new thing
26:48and so there there may be a power shift
26:50happening from on the one hand large
26:53companies to small and medium companies
26:54that can now more aggressively adopt
26:56technology faster and then from
26:58companies in the US and Western Europe
26:59to companies all over the world that
27:01kind of they can also do the exact same
27:02thing and so at the very least of
27:04leveling the playing field and possibly
27:06even a national shift and balance for
27:07small and midsize companies all over the
27:09world may all of a sudden get a lot more
27:10competitive so you've got kind of
27:11democratization in one point and then to
27:14your point there's one version of
27:15internationalization which is adoption
27:16across international community so how do
27:18you think about then the other aspect of
27:20internationalization which is company
27:21formation should we then expect to see
27:23more new company formation outside the
27:25US partly as a result of some of these
27:27trends and why won't we see or will we
27:29see 50 Silicon Valley's you know over
27:31the next you know twenty thirty four
27:33years and how do you all think about
27:34what the strategy should be b2b those
27:36opportunities that would be probably the
27:38most amazing thing for the world that
27:40could happen and the realm of Business
27:42and Economics so we're we're hoping for
27:46it and certainly building kind of help
27:48trying to build technologies that would
27:50facilitate it and I think the world has
27:52never been kind of more ripe for that
27:55kind of thing having said that look
27:58there are real Network effects
28:00geographical Network effects and Silicon
28:02Valley obviously has the biggest one in
28:04technology and you always have to keep
28:07in mind and this is something that gets
28:09lost is there are no local technology
28:12companies right there there's nobody who
28:14sells you know internet search to
28:17Wyoming that's not like a viable thing
28:20so when you're competing globally
28:23it does matter you know do you have the
28:25best people do you have the best
28:26executives you have the best engineers
28:28do you have access to money like all
28:30these things become real competitive
28:31things so we still are believers in
28:34Silicon Valley and we're very hopeful
28:36that the rest of the world grows and
28:39you know participate in that as well but
28:41that's TBD it's an interesting macro
28:44kind of thing that's happening a lot of
28:45the you know one of the really kind of
28:47negative stories is that there's
28:48basically the world is starved for
28:50innovation and growth one of the data
28:52points you point to on that is there's
28:54now ten trillion dollars of money in
28:57being held in government bonds
28:58governments all over the world trading
29:01at what's called negative yield this is
29:02literally like the equivalent of a
29:03savings account where instead of a bank
29:05paying you interest you have to pay the
29:06bank interest to hold your money and so
29:09there's literally 10 trillion dollars of
29:10capital parked around the world that is
29:12actually losing money as it sits there
29:14which means that people cannot find
29:16enough productive places to deploy
29:18capital the conventional view if you
29:19just pick up the newspaper and read the
29:20economic section the horrible this is
29:21and how it means the world is just start
29:23for growth the optimistic side of it is
29:26there's ten trillion dollars of money
29:27sitting on the sidelines waiting for
29:29something productive to be done with it
29:31what could be productively done with it
29:32right new kinds of health care new kinds
29:33of education right new kinds of new
29:36kinds of consumer products new kinds of
29:37media new kinds of art new kinds of
29:38science you know new kinds of new kinds
29:40of you know self-driving cars new kinds
29:42of housing all these things that need to
29:43be done all over the world and so the
29:45world has never been more ripe for a you
29:48know very large wave of innovation that
29:50would actually be quite easy to finance
29:52a lot of the times you just can't get
29:53things done because you don't know for
29:54money right there's just kind of the
29:55constant state of the world for a very
29:56long time and now ironically we we live
29:58in a world where the opposite is true
29:59there's actually quote-unquote too much
30:02money more money than ideas creativity
30:08human creativity is bottomless and so if
30:10you can get more smart people around the
30:12world educated and with the skills
30:14required to do these things and if you
30:15can get them in environments either
30:16either creating environments to do that
30:18or figure out how to get more the people
30:19from other places in environments where
30:20they can do new things we could do all
30:22kinds of new things globally it's
30:24something that we we hope to contribute
30:25to but I think is is a very big
30:27opportunity for the world so you think
30:28we're getting to the point where it's
30:29kind of geopolitical risk and rule of
30:31law issues that limit adoption or
30:34deployment of some of these new
30:34technologies and other countries outside
30:36the US it sounds like it's less more
30:38it's less so technological advancement
30:41well there's I would say there's bad
30:42news and good news so the bad news is we
30:43frequently have delegations of folks
30:45coming into the valley from all over the
30:47US and all over the world and they
30:48basically come in and its economic
30:50delegations of different kinds of
30:51politicians or whatever
30:53and they come in and they're like okay
30:54what can we do to have her run Silicon
30:55Valley and then you kind of sit down to
30:57them and you kind of go through you know
30:58ABCDEF you know all these things well
31:00you know you want you know what rule of
31:01law you want you know ease of migration
31:03you want ease of trade you want deep
31:04investments it deep investments in
31:06scientific research you want no
31:08non-competes you want fluid labor laws
31:09to let companies very easily both hire
31:11and fire you want the ability for
31:12entrepreneurs to be able to start
31:14companies very quickly you want
31:15bankruptcy laws that make it very easy
31:16to to to move on and start another
31:18company and at some point the visitors
31:21give this trick and look on their face
31:22and they're like whoa at the end of it
31:24they're like okay but like what if we
31:25want Silicon Valley but we can't do any
31:26of those things and so that's that's the
31:29bad news and they can hire Donald Trump
31:30to run this it's Sam it's ironic that we
31:33have this guy running for president who
31:34would seriously move us backwards on a
31:35number of those topics so even we
31:37struggle with with these things right
31:39it's like I would argue the formulas
31:40fairly well known it's just people do
31:41not want to apply it for reasons that
31:43have a lot to do with politics and have
31:45a lot to do with you know with with
31:46other issues the good news is it can be
31:48done and then the other good news is it
31:49is happened and there are very very very
31:52exciting things happening throughout
31:53much of the world there are you know
31:55very active now startup scenes all
31:56through you know South America Brazil
31:57Argentina Buenos Aires amazing things
31:59are happen in India there's all kinds of
32:01startup activity throughout the Middle
32:02East they're startup activity now
32:04throughout Africa there's you know
32:05obviously China's been a gigantic
32:07Korea has all kinds of interesting
32:08things happening so there are lots and
32:11lots of extremely positive early
32:13indications of what's possible in many
32:15places all over the world that's that
32:17there are very big political questions
32:18about whether or not those founders are
32:19gonna be able to operate an environment
32:20that's really to let them succeed to the
32:22level that they should be capable of
32:23doing a big reason that we raise the
32:26fund and are excited about the fund is
32:28it is a backing of our core belief
32:30system here which is we believe in the
32:34creativity and ingenious and
32:37intelligence of human beings and the
32:39entrepreneurs that we see and come to
32:41Silicon Valley and around the world and
32:43we believe that these people absolutely
32:46have the ability to change things and
32:48are changing things and there's plenty
32:50of room to improve the world and there's
32:54plenty of ideas to do so and that's
32:56really what we're about with fun five so
32:59let's talk a little bit about kind of
33:01company building and founders in
33:02particular so you know undoubtedly you
33:06view of what types of founders you
33:07wanted to back when you started the firm
33:09now seven years ago how has that evolved
33:12if it all overtime you know what has
33:13changed either in terms of the types of
33:14founders you see or the types of
33:16qualities you see that actually make
33:17founders successful that's caused you to
33:19either augment or rethink some of the
33:21initial you know foundations for the for
33:23the firm you know I think a lot of the
33:25things we don't had this great advantage
33:27when we started the firm that you know
33:28we ourselves were founders I think that
33:31we've probably gotten I would say more
33:35risk tolerant and our view of founders
33:37over time even though we have this thing
33:42we say at the firm which is we're much
33:44more interested in the magnitude of the
33:46strength than the number of the
33:48weaknesses we always believe that
33:50intellectually I think that some of the
33:52number of weaknesses were fairly
33:55terrifying early on I'm just because you
33:58know you do have a lot of founders with
34:00a very small amount of experience these
34:03days which is also you know part of
34:05their strength in that it's hard to
34:06rewrite the world if you're too steeped
34:08in the world and so I think over time
34:10we've kind of doubled down on that and
34:13really you know that the founders who
34:15have figured out something really
34:17important or who are true geniuses or
34:20have will to power that you know we
34:23can't even contain in the room you know
34:25when they bring those things to the
34:27table whatever is wrong with them
34:29we tend to overlook and work with them
34:31on that and if they're strong enough in
34:33those areas you know the really
34:35interesting thing for us has been those
34:36weaknesses do go away pretty quickly and
34:38that's probably the the biggest learning
34:40is I'd say we went in thinking that
34:42we've gotten even more extreme in our
34:46commitment to that kind of philosophy so
34:49almost in financial terms you're buying
34:51volatility to a certain extent well I
34:53think buying volatility in the sense
34:55that we're buying people have
34:57world-class strengths where we care
34:58about them right and regardless of
35:01whatever else there is volatility in
35:03that but you you can have a different
35:05kind of volatility you know you can have
35:07people have gigantic weaknesses that are
35:09spectacular without having the strengths
35:11and we're not trying to buy that kind of
35:13volatility how do you know though that
35:16they're going to be the ones to actually
35:17build the companies at scale because
35:19seems to be this inflection point where
35:21the very thing that makes you a founder
35:23that's gonna punch through this tough
35:24industry there's also the things that's
35:26pretty much gonna hold you back from
35:27really building your company in a really
35:29meaningful way if you think you can do
35:31everything you know your way and there
35:33seems to be an inherent contradiction in
35:34that I think that that would be right if
35:37founders did not evolve so I think what
35:41and some don't and some don't like like
35:43for some some done and they get stuck
35:44and they and they can't get past that
35:46point but you know it's a real common
35:48characteristic in great founders that
35:50they want to know absolutely everything
35:51about the company and how it works and
35:53you know every knob and every button and
35:57and they want they they really would
36:00like have a strong desire to actually be
36:02able to do every job in that company
36:03themselves if it came down to it but
36:05those kind of founders also have great
36:07ambition and it's very logical and easy
36:11to understand that there's never
36:13actually been a gigantic long you know a
36:16really important long lasting company
36:17that had like five employees that those
36:19just don't exist and so if you're gonna
36:21have to have a bigger company than that
36:23you have to think about the company not
36:28only you know from the scale perspective
36:30but from the perspective the people
36:31working there and how are you gonna get
36:33great people to work with you if you're
36:36literally making every decision in the
36:38company and I think that look not every
36:41founder can let go of that and sometimes
36:42it's a psychological flaw rather than a
36:45desire for greatness and if it's a
36:48psychological flaw that they can't
36:49overcome then you know it's just like
36:51any flaw that any of us have you know we
36:53can't stop eating ice cream or whatever
36:55and you know there's something we can do
36:57at that point like we can give them the
36:58logical explanation but they've got to
37:00fix themselves one of the things that
37:03we've seen even in the short time that
37:04the firm has been in business is
37:05companies staying private longer taking
37:08a longer time to IPO what are some of
37:10the implications of that on the company
37:11building process how do you kind of
37:13balance that new reality if it is a new
37:15reality around how companies stay
37:16private with how you think about
37:17building management teams and other
37:19issues around the company yes I think
37:21this gets back to probably the one of
37:23the more neglected parts of company
37:26building which is like what is the
37:27company culture what does it believe
37:29what's our way of doing things you know
37:31when we come to work every day what is
37:33how do we prosecute an opportunity and
37:36the kind of philosophy onboarding
37:40training into that culture and so forth
37:42and so you kind of have to develop a
37:44philosophy like what kind of in place
37:46you want how do you want them to behave
37:49how do people contribute as we're
37:50getting close to wrap it up here what
37:52would be one piece of advice that you
37:54might give either from a management
37:55perspective from a go to market
37:56perspective what would be a takeaway for
37:58people listening to this podcast from a
38:00management perspective I think the the
38:02most common mistake that founders make
38:04is they make decisions based on
38:06management decisions and organizational
38:08design decisions based on very kind of
38:12proximate perspective so what's my
38:14perspective what's a person I'm talking
38:16to his perspective what's my HR person's
38:20perspective without like taking the time
38:22to go okay like how does everybody in
38:24the entire company see this decision and
38:26how will they see it once it's made is
38:28it motivating people in the way that I
38:30think it will and let's look past the
38:32person I'm talking to feeling good about
38:35what I'm saying and really make this for
38:37the long-term health of the organization
38:38yep single biggest strategic piece of
38:41advice we just see across all of our
38:42companies literally people just need to
38:44raise prices people need to charge more
38:46for their products and services the good
38:47news was you have all these new founders
38:49with many different backgrounds who have
38:50come and many of them have never run
38:51companies before run sales forces before
38:53and so they have these extremely
38:55sophisticated views on things like
38:56products and design and engineering and
38:58then I think in some cases relatively
39:00naive views on on how to actually
39:02prosecute a campaign to be able to get
39:04the world to use your product and so the
39:06temptation we see from many founders is
39:08to have a one-dimensional view of what I
39:10call one-dimensional view of the
39:11relationship between price and volume
39:13which is if I price my product cheap
39:15then I could I sell more of it because
39:17the assumption is just that people just
39:19make purchase decisions based on cost
39:20and so you just you drive down you drive
39:22down prices you drive up volume and by
39:23the way a lot of the history of the tech
39:24industry like the chip industry is drive
39:26down prices drive up volume but a lot of
39:28startups really suffer from from having
39:30that view instead we encourage companies
39:33to adopt what I call kind of the two
39:34dimensional view which is the advantage
39:36of raising prices actually there's a
39:38couple advantages so one big advantage
39:39if you raise prices you can afford a
39:41bigger sales and marketing effort a lot
39:43of companies have prices that are
39:44actually too low to be able to mount the
39:47sales marketing campaign required to get
39:48people to ever actually buy the product
39:49and I call this the - hungry to eat
39:51problem I'm not selling enough but I'm
39:55not selling enough because I don't have
39:56the sales marketing coverage required to
39:57actually get the product out there and I
39:59don't have that because I'm charging too
40:00little as a consequence I'm not selling
40:02any despite my low prices the other
40:04really interesting thing is that for a
40:05very large number of products it turns
40:07out if you charge higher prices the
40:08customers take the product more
40:09seriously they impute more value into it
40:11when they're making their purchase
40:12decision and then once they've purchased
40:13they've made a bigger commitment to it
40:15and particularly anybody selling
40:16anything to businesses businesses will
40:18take something that they had to pay a
40:20lot of money for a lot more seriously
40:21than something that they didn't have to
40:22pay very much money for so you can get a
40:24much higher level of engagement and
40:26stickiness and actually use of your
40:27product if you charge more going through
40:29this this this definitely has felt like
40:30swimming upstream for the last several
40:31years we see some glimmers that more
40:33folks are trying to figure this out okay
40:35well that's all we have time for I think
40:37this is the first time I've actually had
40:38all you guys together on the podcast
40:40since we did our fifth anniversary
40:41podcast a couple years ago kind of
40:43amazing how much has changed even in
40:44that short amount of time so thank you