I think part of the conversation is if Google et al are engaging on new kinds of monopoly practices that are not covered by the law. For example, the balance between the amount of information contributed by users in a platform like Twitter or Facebook vs. API restrictions to consume this information.

At the same time, antimonopoly laws don't help to create competition in, for example, the search engine space. Google innovation is unique in this aspect and even companies with virtually infinite budgets like Microsoft can't compete.

For a specific example: Lets say you are a PhD computer vision team and you want to create a machine vision API, where anyone can upload an image and return a list of objects in that image. Your product will be less capable than the Google Vision API or the Microsoft Vision API to start with, simply because they have orders of magnitude more data and personnel they can train their networks with. Now lets say you have a novel technique that is faster or needs less data. Ok, well more than likely Google or Microsoft would come to you with $XXM and offer to buy you out. You say no, we're gonna win this space. Well unfortunately, Google/MSFT personnel go to the same vision conferences, recruit equally if not more competent personnel and they just recreate your solution, even if slightly less capable. They also have a much bigger sales staff so they can always beat you at BD and support etc...

So the question in my mind is: Is the bar to compete unreasonably high such that you effectively need hundreds of millions of dollars to compete?

I use Machine Learning examples because they are literally where the biggest competitive gains and fastest growth are happening - also what I know the best in this case.

However I see this happening in almost every market, home furnishings (the market we play in most) is seeing the same thing, but with logistics/fulfillment instead of machine learning - such that the big retailers/manufacturers can take advantage of more efficient fulfillment than smaller ones can.

Yes, I think it is so high that we are launching startups in denial and the startup community, in general, is very naive. At the same time being [partially] irrational can be an advantage.

In this context, capital is critical. For example, Uber received almost $ 9B in investments, Spotify $ 1.6B and obviously those investments are oriented to having an advantage that goes much beyond building an app.

I don't think you can obtain an advantage creating a better vision API while you don't have access to the amount of information that companies like Google have. Sooner than later your algorithm will be used by anyone. You can find an apparent contradiction of what I am saying with what Google initially achieved with PageRank but nobody spent too much attention to mimic it at that time while Google was quietly running low cost server for years without revenue. When a few ones started to pay attention it was late: Google has a crawling at hyperscale and moved much beyond PageRank.

Looking at your company (http://www.pair3d.com) I found pretty exciting what you are building. I don't know about AI, AR, or your competitors but probably you have build something that goes beyond just using vision algorithms, and at the same time you have some kind of temporal advantage. I think you should focus on having an advantage at the distribution stage engaging companies to use your platform. Google et al can copy what you did but I don't think they are laser focused in the way you are doing it. Please take it with a grain of salt. You never asked for advice nor I know about your company, but I think it is good to talk using concrete examples.

That's exactly what we're doing. Thanks for the well thought out response.

What happened is that the landscape shifted out from under many of the incumbents...from hardware to software. The one survivor (I won't consider Oracle because Altman didn't), was the one given an excuse to change its very profitable business practices by anti-trust prosecution. Not that it happened fast by internet standards. But super tankers don't turn on a dime and Wall Street beholden oil tankers probably turn even more slowly.

I guess my big take away is that AFGAM is more of a concern if the world is a finite pie and less so if it isn't.

The infinite pie theory is for all practical purposes impossible. Growing pie theory works on beyond lifetime scale, but not for decade level scale. So it does matter because a startup can't last for two decades to try and gain market share against an incumbent.

Anyway, Amazon is still the underdog going up against Walmart which twenty years on has really good technical chops and turns over a lot more goods and has phenomenally good logistics...think about who is going to benefit more from autonymous trucks. While it has the overhead of a vehicle fleet it is not at the mercy of UPS and Fedex and for a lot of goods, ordering online and picking up at the local Walmart is about as good as door to door shipping.

Anyway, in some ways Netflix is more likely to be crushed by Walmart than Amazon...and Disney is even more likely than either of them. As internet purchases and bandwidth increasingly become commodities, companies with long horizons become more formidable particularly in markets that matter...the players in the mobile market are Deutsch Telecom and ATT not plucky entrepreneurs. The big car makers are moving on self-driving cars and they already have down the really hard part of making cars...software can be copied.

Edit: Amazon, Facebook, Google, Apple, Microsoft

And for most founders, taking that offer is a rational move, even when the company is quite mature. For a non-diversified human, a 100% chance of $50m is worth enormously more than a 10% chance of $500m.

What fascinates me is the strategy, pioneered by Facebook, of pushing that logic as far as it will go, and spending hitherto-inconceivable amounts of money to acquire any social networking or communication product that could possibly compete with it, long after they achieve product/market fit but before they're so big as to be a sure thing. I remember when $1bn for Instagram was a lot...

If you don't sell, then the big companies don't get bigger. Unfortunately in that case if you do get big enough despite that, they will just target you and take a chunk of your market (Instagram Stories vs Snapchat) so that you can't dominate.

Look, it's brilliant and exactly what every firm should do. The question is really, if it's a net positive to have 5 companies run everything* in consumer technology. I don't know but history seems to not like that.

ML in its current form - efficiency optimization across a ton of different vertical industries - is also a reaction to decreasing productivity. If you can't grow your revenues, you need to shrink your costs, and ML/analytics do that well.

Once we see a new technology or infrastructure driver that jumpstarts productivity growth (distributed electricity, refrigeration, plumbing/running water, public roads, combustion engines, more recently the internet and mobile phones) startups will be able to once again flourish and will be able to return to the public markets with their own company/brand rather than look for acquihire exits as many do in the current climate.

Something that makes sense to me would be some form of basic human sleep/energy acquisition optimization. We spend 1/3 of our time asleep - is there a way to use VR/AR to enhance productivity during this time, or to avoid it altogether without killing us? It seems unnatural, but many facets of our current lifestyle would have seemed unnatural to someone 50, 100, 200 years ago.

It isn't that you can't dominate, but the thing is, you need to have a product that others can't copy.

I think he says what he means here; these largest companies "have powerful advantages that are still not fully understood by most founders and investors." And it's compounding. Given that, I think being a bit more cautiously considerate and unpresumptuous about these issues as well as on any views or insights one might have makes a lot of sense.

Basically this is a shitshow still in progress. There are startups that in hindsight had a too early, under-valued liquidity event. And there are many many startups that turned down acquisition offers never to see that many zeros offered again (I think Pebble just sold under these conditions). Whoops.

Also don't forget how a startup's risk profile and internal planning changes dramatically when/if an acquisition offer is turned down. Too early is not the problem here. If Amazon or FB wants to buy you and you say no, well good luck. Competitor risk is a real business cost and it can weigh you down no matter how fast you may be growing. If anything I see uneasiness more than complaint as the playing field has warped where hills are now mountains only giants can cross. The rulebook has changed too, startups aren't even playing the same game as large tech companies anymore. As the letter suggests, startups can play at the fringes and margins, then maybe try to sneak into grown-up business land but this is increasingly, laughably unlikely if you model it. What is more likely is startups miscalculating how big and lucrative their idea is and they are screwed, or they miscalculated how fast and easy Google can crush their prospects once Lar and Serge notice what you're up to. The options are lose now or lose later, with only the rarest of startups as exceptions unless someone or something steps to address the underlying structural realities.

If you are a self-driving car start-up today with $120MM in funding, you still have no data on how humans drive or how pedestrians walk. That's something that is going to be really important. And Apple and Google (both working also on self-driving cars) have a massive amount of data on drivers.

Put something together here: https://medium.com/safegraph/where-should-machines-go-to-lea...

That assertion may be correct regarding purchases, browsing habits (especially for Google) and places you visit via their respective map applications, but it doesn't compare to the staggering amount of raw data [0] that Tesla has on drivers.

[0] https://qz.com/694520/tesla-has-780-million-miles-of-driving...

The competition is too high now to be friendly.

Our antitrust laws have been focused primarily about the anticompetitive aspects of owning a whole market, but has missed the boat so far on the anticompetitive aspects of vertical integration. Google simply doesn't have to compete in many cases because it's stronger products promote and protect it's weaker products.

Users tend to either be Google people or Apple people or Amazon people, and that trend is only getting larger.

This is simply historically false; vertical integration has been a central focus of antitrust law from day one (it was a key consideration in the drafting of the 1890 Sherman Act and the 1914 Clayton Act.)

Also, most of what you describe is participating in multiple markets that aren't part of a vertical supply chain, so while it is integration, it's not vertical integration.