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.
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.