The other thread in this narrative is that many big companies — including startups that were nurtured on AWS and then grew — are finding the hybrid cloud model attractive. This involves keeping some workloads on public clouds like AWS, Microsoft Azure or Google Cloud Platform and others in-house on a company’s own servers. And for workloads that will remain in public cloud, companies would be fiscal dopes if they did not spec out AWS competitors; if only to wring pricing advantages from AWS. Starting a few years ago, this is exactly how big Microsoft Office shops wielded Google Apps to wrangle concessions on their Microsoft enterprise licenses. What’s old is new again.

It feels like this competition is just starting, and that there will be downward price pressure once more people are using these services.

To that point:

So this is all strikingly similar to textiles. Amazon.com has to keep on spending massive amounts on capex, and such investments only keep it in the same place competitively speaking. And over time, with the massive ongoing improvements in equipment, the past investments won't have time to be paid back and will become obsolete quickly.

It gets worse. The business is also similar to a pure paper pulp maker

Providing cloud capacity can be even worse than a textile business. Here, I'll introduce pure paper pulp makers. There were several of these for many years, including Aracruz in Brazil. These producers were not integrated: they produced paper pulp but not paper. Most of the paper production worldwide, however, was integrated, in the sense that paper markets mostly also had pulp operations.

This meant that making paper pulp was a deeply cyclical business - when demand for paper diminished a little, demand for paper pulp diminished a lot, since integrated paper makers first cut outside supplies unless the price was very low.

So how is the cloud becoming similar? It is becoming similar with the emergence of hybrid clouds, where customers run part of the necessary capacity themselves and buy capacity from public clouds for peaking needs. I have explained this phenomenon in my previous article "How Amazon's AWS Can Attract Ugly Economics".

This is already happening with AWS' largest customers, including Zynga, Dropbox and Netflix. At the same time, since AWS needs to compete with baseload capacity, its prices cannot be the "peaking" prices which such usage would demand. Furthermore, the peaks can be large and AWS needs to have the capacity to serve them.

http://seekingalpha.com/article/2351305-how-do-you-like-them-textiles?uprof=17

I believe that Zynga, Dropbox, and Netflix are the exception.

I believe that most of AWS customers are like the customers for IBM Global Services: they are global companies with a lot of dollars to spend who care more about maintainability than saving costs.

Are those not big exceptions?

No, they're a rounding error compared with the Fortune 500 IT budgets.

I also believe government contractors, universities, and science lags are bigger customers than them.

Well, the CIA for one!  My argument is invalid!

You got it. Startups have cachet because they're cutting edge but they're not where the big money is.